An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Immune thrombocytopenia (ITP) is an autoimmune disorder characterized by low platelets and an increase in bleeding risk,1-4 with a reported incidence rate varying between 2 and 4 per 100,000 adults per year according to various epidemiological studies around the world.5 Severe bleeding, such as in the gastrointestinal tract or in the brain, is seen in approximately 10% of adults.6,7 First-line therapy is a short course of corticosteroids;8 alternatively, IV immunoglobulin (IVIG) and anti-D immunoglobulin may be used in patients who are unresponsive or have a contraindication to taking corticosteroids.8 Recommendations for subsequent-line therapies are based on evidence of low certainty.8 Therefore, a systematic review was undertaken at the request of public drug plans to determine the efficacy and safety of subsequent-line therapies — that is, rituximab, eltrombopag, romiplostim, fostamatinib, avatrombopag, and splenectomy — in adult patients with ITP who have failed first-line treatments.
The research protocol was developed a priori, by engaging with patient groups, and the systematic review used robust methodology. A total of 15 publications met the final inclusion criteria, reporting findings from 10 unique trials. The study populations in the trials were considered generalizable to most ITP patients and the concomitant use of other ITP medications was allowed. A narrative synthesis was conducted due to clinical heterogeneity between studies and methodological constraints that precluded a network meta-analysis (NMA).
Findings from 4 randomized controlled trials (RCTs) showed that, compared to placebo, eltrombopag generated a long-lasting platelet count response (at least 6 weeks and up to 6 months) and improved health-related quality of life (HRQoL), in addition to reducing bleeding events and the need for rescue medication. In 3 RCTs, romiplostim increased platelet count response, improved ITP symptoms and HRQoL, and reduced the need for rescue medication compared to placebo or standard of care. Interpretation of the effectiveness of avatrombopag, rituximab, and fostamatinib was inconclusive due to the limited availability of evidence, which was 1 single trial for each intervention, with varying risks of bias.
Due to a scarcity of evidence, conclusions could not be drawn for the clinically meaningful outcomes of emergency department visits, hospitalization, and treatment-free remission.
An economic evaluation could not be conducted due to significant clinical data gaps, including the methodological limitations precluding assessment of comparative treatment efficacy in an NMA. Therefore, the comparative cost-effectiveness of treatments for ITP is unknown.
No head-to-head study, or study with comparisons to splenectomy, was identified from the literature search. Methodological consideration prevented an NMA from being conducted. As such, direct comparisons of effectiveness between ITP treatments cannot be made. Similarly, there was insufficient evidence to inform the efficacy or sequence of subsequent-line therapies in subgroups of adult patients with ITP or the role of splenectomy in the sequency of subsequent-line therapies.
The narrative review was limited by availability of evidence. The risk of bias across the 10 studies ranged from low9-11 to high.12-15 An overall limitation of the included trials was the small sample size, which limited the level of precision and affected the certainty in the findings.
Considering that the evidence of a clinical benefit was restricted to eltrombopag or romiplostim, public drug plans may want to consider revisiting reimbursement criteria for ITP. CADTH suggests that interested jurisdictions request an implementation advice panel to suggest strategies (e.g., discontinuation criteria) to manage the clinical and economic uncertainty identified in the report.
ITP is an autoimmune disorder characterized by low platelets and an increase in bleeding risk due to increased platelet destruction and impaired platelet production.1-4 It was previously called “idiopathic thrombocytopenic purpura,” but is no longer considered to be an idiopathic disease.1,4 The specific cause of ITP is unknown but may be a mix of genetic and environmental factors.16 In addition, not all patients will experience bleeding symptoms such as purpura (hemorrhages in the skin and mucous membranes).16 Symptoms associated with ITP are excessive bruising, superficial bleeding that appears as reddish spots on the skin (petechiae), bleeding from the gums or nose, blood in urine or stool, and unusually heavy menstrual flow; however, ITP may present with no signs or symptoms.
Primary ITP is defined as isolated thrombocytopenia — that is, peripheral blood platelet count less than 100 × 109/L — in the absence of other causes or disorders that may be associated with thrombocytopenia.1 The disorder falls into 1 of the following 3 disease groups according to disease duration:
Spontaneous remission may occur when there is an improved platelet count in the absence of ongoing or recent therapy, which occurs in 45% of adults at 6 months and 49% of adults at 12 months. Among those with chronic ITP, 30% of adults achieve spontaneous remission at 24 months.17
ITP has a reported incidence rate varying between 2 and 4 per 100,000 adults per year according to various epidemiological studies around the world.5 Bleeding symptoms are often present, including severe bleeding (such as in the gastrointestinal tract or the brain) in approximately 10% of adults.6,7 At least 1 comorbid condition is reported in more than 30% of patients.7
First-line therapy for newly diagnosed adult patients, according to both the American Society of Hematology 2019 Guidelines for Immune Thrombocytopenia8 and the Updated international consensus report on the investigation and management of primary immune thrombocytopenia,18 is a short course (approximately 6 weeks) of corticosteroids. For patients with ITP who are unresponsive or have a contraindication to taking corticosteroids, other first-line therapies include IVIG and anti-D immunoglobulin.8 The international standardized definition1 of a platelet response is a platelet count of 30 × 109/L or higher with at least a 2-fold increase in platelet count from baseline and an absence of bleeding, while a platelet count of 100 × 109/L or higher would typically be considered a complete response as long as there is an absence of bleeding.1
Patients who do not respond to first-line treatment within the first 3 months, or who are corticosteroid-dependent, will have a disease duration that falls within the persistent ITP category and may benefit from second-line treatment options.19 These include thrombopoietin receptor agonists (TPO-RA) romiplostim and eltrombopag, as well as rituximab.19 After 12 months of ongoing, active disease, patients are considered to have chronic disease and, in addition to second-line treatment options, become candidates for splenectomy.8 The choice of treatment will be based on patient preferences, such as whether they prefer a durable response, or if they would rather avoid long-term medication or avoid surgery.
Refractory ITP is characterized by nonresponse to splenectomy, or relapse after surgery, and either severe ITP or a high risk of bleeding that requires continued subsequent therapy.1 Treatments may include 1 of the following: TPO-RAs (i.e., eltrombopag, avatrombopag, or romiplostim), spleen tyrosine kinase inhibitor fostamatinib, or rituximab.18
There are other second-line treatment options, including azathioprine, cyclophosphamide, cyclosporine, danazol, dapsone mycophenolate mofetil, and the vinca alkaloids; however, guidelines such as that of the American Society of Hematology do not provide recommendations on these drugs due to limitations in the current evidence, mainly the small sample sizes of studies and heterogeneous patient populations.8 A brief summary of the reported efficacy of immunosuppressive drugs without American Society of Hematology guideline recommendations can be found in Table 1.
Immunosuppressive Drugs Without Guideline Recommendations.
It is unclear which treatments are most appropriate for adult patients with chronic ITP who have failed first-line treatments. The federal, provincial, and territorial public drug plans have requested an evaluation on the comparative evidence of available treatments for adults with ITP who have already received first-line therapies (i.e., corticosteroids, IVIG, and/or anti-D immunoglobin) and the place in therapy of splenectomy.
CADTH welcomed feedback from the patient community on the proposed project entitled Treatment of Adult Patients With Chronic ITP After Failure of First-Line Therapies.” The Platelet Disorder Support Association encouraged its patient community to contribute feedback. Twenty people living with chronic ITP shared their experiences and expectations of ITP therapies, in addition to the information provided by the Platelet Disorder Support Association. Many contributors were in their 50s, 60s, and 70s, with a few contributors in their early 30s. Many had lived with ITP for more than 10 years, with a range of 12 months to 55 years. Most contributors identified that they were from Ontario or British Columbia.
Several contributors described living with the fear of serious or fatal bleeds and the exhaustion associated with low platelets. “Having ITP and living with the fear of bleeding is a very stressful and unpredictable situation. I am not able to work full-time owing to the fatigue associated with ITP. I cannot travel or participate in overly vigorous activities because of the fear of a sudden drop in platelets and bleeding.”
Contributors described a wide range of experiences of therapies: corticosteroids, including prednisone and dexamethasone; IVIG; rituximab; TPO-RAs, including romiplostim and eltrombopag; immunosuppressants, including cyclosporine and azathioprine; and/or splenectomy.
Contributors described trying multiple therapies over the years. “After trying prednisone — did not respond; IVIG which did respond; then splenectomy which did not work; IVIG only lasted a few weeks; rituximab, only last a few weeks; cyclosporine which proved to be the best medication for controlling ITP for the next 11 years. Now on Revolade for the past 4 years, which has worked well.”
Temporary and/or waning effectiveness were concerns shared by contributors. “IVIG treatments which are a lifesaver but temporary.” Time spent in hospital or spent receiving IV therapy was another concern. “During the first six months, I was admitted to the hospital on 5 separate occasions with platelet readings of between 0k and 3k, accompanied by various indications (e.g., unexplained bruises, blood blisters, slight nose bleeds and bleeding gums), with an average stay of 3-5 days in each case.” “I opted for a watch and wait approach because I hated the side effects of prednisone, and the amount of school I was missing for IVIG treatments every 3-4 weeks was not sustainable.”
Many contributors described the side effects they experienced using corticosteroids. “I had terrible side effects from the steroids including manic-like episodes, weight gain, heart racing, insomnia, and eye disturbances. I have had to take bone health supplements and have regular bone scans to monitor bone loss due to long term use of steroids.” One person described being unable to work due to side effects of prednisone, while others described permanent side effects of “brain fog” and osteonecrosis, as a consequence of steroids. A few contributors described a willingness to use steroids only if platelets were extremely low, for example: “I do respond to steroids and have used prednisone but will only use them in an emergency—platelets at 10 or less—as I have a great deal of difficulty dealing with the side effects.”
One person shared their experience of 28 years of remission following removal of their spleen. When ITP returned, use of prednisone and dexamethasone raised platelets levels temporarily until an accessory spleen was found and removed, allowing the person to return to remission. Other contributors shared that splenectomy had not helped them. One person described experiencing portal vein thrombosis, resulting in 2 emergency surgeries and an intensive care unit stay; deep vein thrombosis; bowel resection; permanent daily incontinence; reduced immune system response; and a reliance on blood thinners as a consequence of splenectomy that did not offer remission, to access rituximab, which was also not effective. For this person, only prednisone has been effective.
Outside of Alberta, patients with chronic ITP must try splenectomy to be eligible for public funding of second-line drug treatments. Many contributors described how they faced an uncertain disease path, uncertain benefit of splenectomy in the short term and long term, and knowledge that the removal of their spleen was permanent, required surgery, and increased their risk of infection.
I have been highly resistant to the idea of a splenectomy for a number of reasons, including uneven record of success in treating ITP; surgical risks including long term effects of general anaesthetic; availability of alternative noninvasive medications.
Once the spleen is gone, it is gone forever.
I ended up having a splenectomy that, within a few months, indicated that it had failed. In hindsight I would not have had a splenectomy. But I felt I did not have a choice, as I could not access the more expensive treatments like Rituxan or the TPOs without having first “failed” three other lines of treatment.
Given the variability and unpredictability of the disease, some respondents described refusing splenectomy and the potential for remission, to accept the risks of bleeding and to treat only serious episodes with first-line therapy. “As there is no cure for ITP, I decided that I would rather live with what I know than a whole new batch of side effects with treatment that may or may not help bring my platelets up.”
A desire to see different types of therapies accessible to patients in Canada — including avatrombopag, fostamatinib, eltrombopag, romiplostim, rituximab, and oral veltuzumab — was expressed by contributors. “I understand this is a rare condition but, to me, it feels like there is really no treatment for me because what is available is so very inadequate.”
One contributor asked for public funding of TPO-RA medications. Another emphasized that health care providers should explain to patients and their families the full range of treatments available for ITP, even if not all are publicly funded. “As other new treatment advances are made, more favourable medications should constantly be re-evaluated in terms of coverage and accessibility. For now, I remain in a position where my only next treatment option is the same one offered over 100 years ago.”
Many contributors expressed a desire to avoid surgery (splenectomy) and the side effects associated with the long-term use of steroids. “It seems like an inhumane choice to force surgery to remove an organ before access to other treatments, especially when splenectomy is less often successful the older a patient is and does not have a high percentage of success in the ITP population as a whole.”
Several sought public funding of the use of indium white blood cell screening, to gauge the likelihood that splenectomy would be successful for the individual, before considering surgery.
The collection of patient perspectives highlighted the unmet need for treatment options for adult patients with chronic ITP after failure of first-line therapies.
CADTH undertook a Health Technology Assessment (HTA) to review the effectiveness and cost-effectiveness of treatments for ITP in adults after failure of first-line treatments. A secondary objective was to determine the place of splenectomy in therapy.
The following policy questions were addressed with this project:
The project addressed the following research questions. Details on the specific interventions and outcomes are included in Table 2.
To inform the conduct of this HTA, a preliminary informal scoping of the existing literature was conducted. A protocol was written a priori and was followed throughout the study process. The project is an HTA that includes clinical and economic reviews. Stakeholder feedback was solicited at key steps. The key elements of the project, including the population of interest, the interventions and comparators, and the main outcomes of interest, are outlined in Table 2. The protocol was registered (registration number: CRD42021276565) in PROSPERO.20 No deviations in the protocol occurred.
The project scope was posted on the CADTH website for 10 days and feedback from the patient community was reviewed and summarized. The collection of patient perspectives enabled the research team to consider the evidence found in the literature alongside an understanding of the wider experiences of patients and family caregivers. In addition, the draft list of included studies in the review, the statistical analyses plan, and a draft of the report were also posted on the CADTH website and all stakeholders (e.g., patient community, clinician groups, industry) were given the opportunity to provide written feedback.
This clinical evaluation was designed as a systematic review and an NMA to answer the first and second research questions. The systematic review and NMA of the primary studies focused on the clinical effectiveness, comparative clinical effectiveness, safety, and comparative safety of ITP medications, as well as on the clinical effectiveness, comparative clinical effectiveness, safety, and comparative safety of splenectomy in ITP. The systematic review was conducted following standard methodology and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement,21 with consideration of relevant elements of PRISMA-NMA.22 The NMA was conducted in accordance with the Professional Society for Health Economics and Outcomes Research (ISPOR) guidelines for NMAs.23
An information specialist developed and conducted a literature search for clinical studies using a peer-reviewed search strategy according to CADTH’s PRESS Peer Review of Electronic Search Strategies checklist.24 The complete search strategy is presented in Appendix 1.
Published literature was identified by searching the following bibliographic databases: MEDLINE via Ovid, Embase via Ovid, and the Cochrane Central Register of Controlled Trials (CENTRAL) via Ovid. All Ovid searches were run simultaneously as a multifile search. Duplicates were removed using Ovid deduplication for multifile searches, followed by manual deduplication in EndNote. The search strategy comprised both controlled vocabulary, such as the National Library of Medicine’s MeSH (Medical Subject Headings), and keywords. Search concepts were developed based on the elements of the PICOS (Population(s), Intervention(s), Comparator(s), Outcome(s), Study Design(s)) framework and research questions. The main search concepts were rituximab, fostamatinib, TPO-RAs (romiplostim, eltrombopag, avatrombopag), or splenectomy and ITP. Clinical trials registries with results were searched: the US National Institutes of Health’s clinicaltrials.gov, WHO’s International Clinical Trials Registry Platform (ICTRP) search portal, Health Canada’s Clinical Trials Database, and the European Union Clinical Trials Register.
CADTH-developed search filters were applied to limit retrieval to HTAs, systematic reviews, meta-analyses, or NMAs, RCTs, or controlled clinical trials. Systematic reviews were included in the literature search to inform the initial scoping of the project. Retrieval was not limited by publication date but was limited to the English or French language. Conference abstracts were excluded from the search results.
The initial search was completed on July 23, 2021. Regular alerts updated the database literature searches until July 26, 2022. The clinical trials registries search was updated before the completion of the stakeholder feedback period. Studies meeting the selection criteria of the review and identified in the alerts before the completion of the stakeholder feedback period were incorporated into the analysis of the final report. Any studies that were identified after the stakeholder feedback period are described in the discussion, with a focus on comparing the results of these new studies with the results of the analysis conducted for this report.
Grey literature (literature that is not commercially published) was identified by searching sources listed in relevant sections of Grey Matters: A Practical Tool For Searching Health-Related Grey Literature, which includes the websites of regulatory agencies, HTA agencies, clinical guideline repositories, systematic review repositories, patient-related groups, and professional associations. Google was used to search for additional internet-based materials. These searches were supplemented by screening articles identified by clinical experts and by industry following the stakeholder feedback period.
Studies were included if they met the eligibility criteria, including the specific population, intervention, comparators, and study designs, presented in Table 2. Drug regimens eligible for inclusion in the review were those that have been approved by Health Canada for ITP or were considered of clinical relevance based on expert advice or international clinical practice guidelines. The inclusion criteria were informed by the informal scoping review of the existing literature, patient engagement, stakeholder feedback, and consultation with the clinical experts. Studies were excluded if they did not report outcomes of interest. The language of publications was limited to English and French, with no limitation on publication date, and conference abstracts were excluded.
Selection Criteria.
The patient population included in the review consisted of adults with ongoing, active ITP who did not respond to first-line treatments. Relevant subgroups included duration of disease as persistent ITP (between 3 and 12 months) or chronic ITP (more than 12 months), and previous failure of at least 1 second-line therapy. Patients were eligible for inclusion in the studies if they were adults with a diagnosis of ITP that was either newly diagnosed, persistent, or chronic, depending on the study inclusion criteria. Platelet count was typically required to be less than 30 × 109/L. It was standard for patients to be excluded from the studies if they had significant comorbidities such as other blood disorder, history of thrombosis, cardiovascular disease, or active malignancy.
All currently available subsequent-line therapies for ITP were considered potentially relevant interventions and comparators, including rituximab, eltrombopag, romiplostim, fostamatinib, avatrombopag, and splenectomy. Comparators also included placebo and/or standard of care (IVIG; immunosuppressants, including azathioprine, cyclosporine, cyclophosphamide, mycophenolate mofetil, danazol, dapsone; and corticosteroids). Recognizing that rescue therapies are sometimes given in combination with these drugs, no distinction was made between monotherapy or combination therapy for any of the included treatments.
It should be noted that standard of care was included as a background treatment, or as an add-on treatment to any of the included second-line therapies because the clinical expert noted that standard of care is generally given to all patients. IVIG, immunosuppressants (azathioprine, cyclosporine, cyclophosphamide, mycophenolate mofetil, danazol, dapsone), or corticosteroids were considered as standard of care.
In consultation with a clinical expert, outcomes were selected to assess the clinical effectiveness and safety of the interventions, including platelet count response, bleeding events, emergency department visits, hospitalization, HRQoL, symptoms (e.g., fatigue), treatment-free remission, need for rescue medication (e.g., IVIG, corticosteroids, platelet transfusions), reduction or discontinuation of corticosteroids, and mortality. Various other harms outcomes were considered important for clinical decision-making and were therefore included in the protocol. Any reasonable measurement tool or definition used by study authors for the outcomes presented, as well as any relevant follow-up time point, was considered and critically appraised when needed.
Published phase III and phase IV RCTs were eligible for inclusion.
Two reviewers independently screened titles and abstracts of all retrieved citations in duplicate (i.e., literature searches of academic databases, grey literature searches, and clinical trial database) against the eligibility criteria (Table 2). Exclusion by both reviewers was required for a record to be excluded at the title and abstract level. Full-text articles that were judged to be potentially relevant by at least 1 reviewer were retrieved for the second level of screening. The same 2 reviewers independently examined all full-text articles against the eligibility criteria, and consensus was required for inclusion in the review. Discrepancies between reviewers were resolved by discussion. Reasons for exclusion at the full-text level were documented. The list of final included studies was posted publicly on the CADTH website and feedback was solicited from stakeholders. Studies later identified via monthly database search alerts and semiannual grey literature search alerts meeting the selection criteria of the review were incorporated into the analysis.
Risk of bias in the individual trials was assessed using the Cochrane risk of bias 2 (RoB 2) tool at the level of the reported result.24 Risk of bias was appraised for the effect of assignment to the intervention (i.e., intent-to-treat [ITT]). Risk of bias assessments were completed by 2 reviewers independently in duplicate for each trial. The RoB 2 tool allowed for the assessment of 5 sources of bias or “domains”: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome, and bias in selection of the reported result. Each question within each domain was answered with a “yes,” “probably yes,” “probably no,” “no,” or “no information.” Afterward, a judgment of “low risk of bias,” “high risk of bias,” or “some concerns” was assigned for each domain, with the rationale for each decision included in the comments box field.
For each unique RCT, we assessed the risk of bias of the original primary publication, with additional details sought from supporting literature (e.g., published protocol and ClinicalTrials.gov records). Disagreements were resolved by consensus. The risk of bias of each included study and how it affects the study findings was described narratively. Other critical appraisal points were noted by reviewers, such as assessment of the generalizability of the findings.
Data were extracted by 1 reviewer by the use of piloted and standardized data abstraction forms, and the extracted data were checked for accuracy and completeness by a second reviewer. Any disagreements were resolved by consensus. Modification to the extraction sheet was conducted after a pilot of 5 studies.
The original, primary publication for each included RCT was used for data extraction, with supplementary data obtained from companion reports and ClinicalTrials.gov records to address the research questions. In situations where multiple publications for a unique RCT were available (e.g., supplemental online appendices, companion publications of specific outcomes, or populations from the original study), the most recent data for each outcome were extracted, with preference given to published records.
The following relevant information was extracted, where available:
After the conclusion of data extraction, a feasibility assessment was conducted for addressing the posed research questions, which included evaluating sources of methodological and clinical heterogeneity between the included studies. Study design, patients’ baseline characteristics, treatment characteristics, as well as outcome definitions were compared between studies. A qualitative assessment of feasibility was determined through close collaboration between the reviewers, methodologists, and clinical experts. Because the NMA was deemed infeasible, an alternate NMA was conducted with simplified assumptions. Ultimately, it was concluded that the NMA was not informative; therefore, a narrative summary was reported. Refer to the prespecified statistical analysis plan for this report provided in Appendix 2 for further details of the planned analysis.
A total of 1,421 citations were identified in the literature searches. Following screening of titles and abstracts, 124 studies were identified as potentially relevant and retrieved for full-text review. A total of 206 reports were retrieved from other sources (i.e., grey literature, handsearch, and search alerts). Of these 330 potentially eligible reports, 15 reports9-15,25-32 presenting data from 10 unique studies met the inclusion criteria and were included for review. The report selection process is outlined in Figure 1. A list of included and excluded reports, with details describing the rationale for those excluded, is presented in Appendix 3 and Appendix 4, respectively.
A total of 15 publications met the inclusion criteria and were included in the systematic review, reporting results from 10 individual RCTs. Peer-reviewed publications were selected for inclusion if they met the prespecified review protocol criteria. Study characteristics are shown in Appendix 6.
One study was performed in patients who received corticosteroids without achieving a sustained response or who relapsed after discontinuation,32 and 3 studies included patients who had insufficient response or relapse to any previous therapies for ITP.9,25,27 While this was the main population of interest for the review, it limited the number of drug comparisons that could be evaluated and used to inform decision-making. Additional trials included patients who had experience with prior ITP therapies, without specifically requiring treatment failure, and these were included in the review as well.
Four included studies were double-blind (DB), placebo-controlled RCTs comparing eltrombopag to placebo over 6 weeks9,25,27 or 6 months10 One included study was a small DB RCT evaluating the use of avatrombopag compared with placebo over 26 weeks.12 Three included RCTs compared romiplostim to either placebo in a DB setting11,13 or standard of care in an open-lavel (OL) setting.14 Two included studies evaluated interventions other than TPO-RA: 1 DB RCT evaluated the use of rituximab compared with placebo over 4 weeks32 and 1 DB RCT compared fostamatinib to placebo over 24 weeks.15 In addition to the interventions and comparators, the concomitant use of various ITP medications was allowed in all trials. These medications included, but were not limited to, corticosteroids, azathioprine, and danazol.
Flow Chart of the Selection Process.
All the studies included in the systematic review reported platelet count response as a primary or major outcome; however, the outcome measures used to assess platelet response varied substantially across the studies. The most frequently used was the proportions of patients achieving response. There were notable differences observed in thresholds and definitions for platelet count response across studies, including some that were not consistent with the international standardized definition1 of a platelet count of 30 × 109/L or higher with at least a 2-fold increase in platelet count from baseline and an absence of bleeding, while a platelet count of 100 × 109/L or higher would typically be considered a complete response, as long as there is an absence of bleeding.1 In addition, there were differences among all of the studies as to the conditions to be observed for patients to be considered responders, such as the time needed to be spent above the threshold value (ranging from at least once during the study to throughout the entire study follow-up), as well as criteria related to the use of rescue medication.
Nine studies reported bleeding as an outcome for either efficacy or harms assessment.9-12,14,15,25,27,32 Definitions of bleeding and reporting methods varied across the studies. Other frequently reported outcomes included the need for rescue medication (e.g., IVIG and corticosteroids), reduction or discontinuation of concomitant ITP medications, and various safety outcome measures. Three studies reported findings on HRQoL or ITP symptoms using various scales: the Functional Assessment of Chronic Illness Therapy (FACIT) – Fatigue subscale;10 the Functional Assessment of Cancer Therapy – Thrombocytopenia 6-item questionnaire (FACT-Th6);10 the Motivation and Energy Inventory – Short Form (MEI-SF);10 the Short Form (36) Health Survey, Version 2 (SF-36v2);10 EQ-5D;11 and the Immune Thrombocytopenic Purpura Patient Assessment Questionnaire (ITP-PAQ).11,14
No data were reported for the outcomes of treatment-free remission, emergency department visits, or hospitalization.
In the 4 studies comparing eltrombopag to placebo, eltrombopag was administered at a dosage ranging from 12.5 mg to 75 mg orally once daily, with the dose to be adjusted based on platelet count response. The mean age of the patients ranged from 41 to 58 years across treatment groups at baseline. The median platelet count ranged from 9,500/µL to 21,000/µL; approximately half of the trials’ populations had a platelet count of 15,000/µL or less. The proportions of patients using concomitant ITP medication at baseline was 82% in 1 trial,25 and ranged from 42% to 55% in the other studies. All studies included a portion of patients who had a splenectomy; however, the proportions varied substantially from 1 trial to another.
One study evaluated the superiority of avatrombopag administered orally once daily, with the dose to be adjusted based on platelet count response and need for concomitant ITP medications, over placebo for 26 weeks. The mean age of the patients ranged from 41 to 46 years across treatment groups at baseline. Median platelet count at baseline was not reported; however, approximately half of the patient population had a platelet count of 15,000/µL or less, while the other half had a platelet count between 15,000/µL and 30,000/µL.12 In this study, 45% of patients had concomitant ITP medication at baseline and 33% of patients had a prior splenectomy.12
In the 3 studies comparing romiplostim to placebo or standard of care, romiplostim was administered at a dosage ranging from 1 to 3 mcg/kg subcutaneously once weekly, with the dose to be adjusted up to a maximum between 10 and 15 mcg/kg based on platelet count response. In 1 study, the mean age of the patients at baseline was 59 years in the romiplostim treatment group and 48 years in the placebo treatment group;13 in the 2 other romiplostim trials, the median age of the patients ranged from 52 to 58 years across treatment groups.11,14 In 2 of the 3 studies, baseline mean or median platelet count was less than 20,000/µL;11,13 however, 1 study included patients with a numerically higher median platelet count (33,000/µL in the romiplostim treatment arm and 27,000/µL in the placebo arm).14 The proportions of patients using concomitant ITP medication at baseline varied substantially from 1 trial to another. Two studies reported the proportions of patients who had a splenectomy, ranging from 42% to 51% across treatment groups.11,13
High-Level Study Characteristics.
Outcomes Assessing Platelet Count Response.
Additional Outcomes.
This section highlights a summary of studies included for each treatment comparison. Additional study characteristics are shown in Appendix 6.
One study (Ghanima et al.; n = 112)32 evaluated the superiority of rituximab over placebo. Rituximab was administered at a dosage of 375 mg/m2 IV every week for 4 weeks and patients were followed for a total duration of 78 weeks. The median age of the patients was 46 years at baseline. The median platelet count ranged from 16,000/µL to 21,000/µL across treatment groups.
One trial evaluated the superiority of fostamatinib administered at a dosage of 100 mg orally twice daily over placebo for 24 weeks (Bussel et al.; n = 150).15 The median age of the patients was 54 years at baseline. The mean platelet count ranged from 16,052/µL to 19,818/µL across treatment groups.
The detailed risk of bias appraisal of all the included trials is available in Appendix 7. The key limitations — that is, those having an impact on the interpretation of the findings — are summarized in this section for each treatment comparison.
Eltrombopag was compared to placebo in 4 DB RCTs of a minimum of 6 weeks’ duration. Of these, Yang et al. (n = 155)9 and Cheng et al. (n = 197)10 were both rated as having a low risk of bias and, therefore, were considered well conducted. Bussel et al. (n = 114)27 was rated as having some concerns for overall risk of bias; the main issue with this study was that the differences in the incidence of adverse events (AEs) may have jeopardized blinding and have affected subjective outcome measurements.
Tomiyama et al. (n = 23)25 was rated as having some concerns for overall risk of bias. The small sample size of the study limits precision and affects confidence in the findings. Patients in the eltrombopag group had a higher baseline platelet count compared to placebo (mean platelet count of 21,000/µL versus 9,500/µL, and proportions of patients with platelet count of 15,000/µL or lower of 20% versus 75%, respectively). This is indicative of the failure of the randomization process, resulting in less severe ITP in patients randomized to eltrombopag, and may bias the results in favour of active treatment. Bleeding was not defined and, therefore, multiple and inconsistent definitions could have been used by assessors, again affecting precision and confidence in the findings.
Of note, 2 of the 4 trials9,25 were performed exclusively in Asia; it is possible that standard of care is different in other countries, which would affect generalizability of the results to the Canadian population.
Avatrombopag was compared to placebo in 1 DB RCT of 26 weeks’ duration. Jurczak et al. (n = 49)12 was rated as having an overall high risk of bias. The small sample size of the study limits precision and affects confidence in the findings. In addition, there were high discontinuation rates in both treatment arms, especially due to inadequate therapeutic response. Differences between groups in terms of reduction of concomitant ITP medications from baseline and AEs may have jeopardized blinding and affected subjective outcome measurements.
Romiplostim was compared to placebo in 3 RCTs with substantial heterogeneity in methodology (e.g., DB versus OL setting) and follow-up duration. Of these, Kuter et al. (n = 125)11 was a DB RCT of 24 weeks’ duration rated as having a low risk of bias.
The other 2 studies were rated as having an overall high risk of bias. Shirasugi et al. (n = 34)13 was a DB RCT of 12 weeks’ duration and small sample size, limiting precision and affecting confidence in the findings. Kuter et al. (n = 234)14 was an RCT of 52 weeks’ duration with an OL design; considering that the definition of treatment failure, the primary outcome in the trial, included 1 subjective component (requirement for a change in therapy, which was based on clinical decision-making), it may be subject to bias in favour of active treatment due to the OL study design. In both studies, features suggestive of failure of the randomization process were observed, leading to imbalances between treatment groups in terms of disease severity. In Shirasugi et al.,13 a higher proportion of patients in the placebo group had concomitant ITP medication at baseline compared to romiplostim (83% versus 59%, respectively), suggesting more severe ITP in patients randomized to placebo and biasing the results in favour of active treatment. In contrast, in Kuter et al.,14 a higher proportion of patients in the romiplostim group had concomitant ITP medication at baseline compared to placebo (13% versus 6%, respectively), suggesting more severe ITP in patients in the romiplostim group, and may bias the results against active treatment. Shirasugi et al.13 was also subject to blinding issues, as differences between groups in terms of AEs were observed, hence affecting subjective outcome measurements. The trial was performed exclusively in Japan; it is possible that standard of care is different in other countries, which would affect generalizability of the results to the Canadian population.
Rituximab was compared to placebo in 1 DB RCT of 78 weeks’ duration. Ghanima et al. (n = 112)32 was rated as having some concerns for overall risk of bias. Not all hospital staff were blinded to treatment allocation, because it was communicated to local hospital pharmacies for infusion bag preparation; in addition, differences between groups in some AEs may also have jeopardized blinding, affecting subjective outcome measurements. The definition of treatment failure, the primary outcome in the trial, was a composite outcome that was amended during the study at a timing that was unclear to include being a candidate for splenectomy. Although this decision seemed to be based on clinical considerations, this may be perceived as an issue due to the lack of details surrounding it reported in the publication.
Fostamatinib was compared to placebo in 1 DB RCT of 24 weeks’ duration. Bussel et al. (n = 150)15 was rated as having an overall high risk of bias. Discontinuation rates were high, especially due to inadequate therapeutic response, in both fostamatinib and placebo groups. Differences between groups in some AEs may have jeopardized allocation concealment, but this would only affect the subjective outcome measurements.
Ten RCTs met the inclusion criteria. A network diagram is shown in Figure 2 and additional details of the feasibility assessment are reported in Appendix 5. Several issues were identified by the feasibility assessment both in terms of methodological and clinical heterogeneity of the network. As a result, the data in the network as described in the prespecified statistical analyses plan was considered infeasible for pooling and analyzing using NMA methodology. Alternative networks for analysis were also considered but had similar limitations for feasibility of conducting an NMA. For these reasons, a narrative synthesis of the evidence was conducted for the studies included in the network.
Network for Platelet Count Response Using the Most Encompassing Threshold (20 Thousand or More Platelets).
To mitigate the lack of informative results from the NMA, a narrative synthesis was conducted. The narrative review was conducted due to heterogeneity across the publications; as such, there were several differences in the way that outcomes were defined, measured, and reported. There was no additional restriction set other than the selection criteria for inclusion in the systematic review. The outcome measures that were considered by the review team to be the most relevant to decision-making were extracted from the peer-reviewed publications. This section lays out the body of evidence, as well as the results and findings, for each treatment comparisons that were available in the 10 studies included in the systematic review — that is, a total of 5 placebo or standard of care comparisons of ITP drugs used as subsequent-line treatment in patients who have had experience with prior ITP therapies.
Detailed outcome results for studies included in the narrative review are presented in Appendix 6 to Appendix 9.
Among the studies identified for inclusion in the NMA, 8 studies reported the proportion of patients splenectomized at baseline. Only 1 study reported cohort-specific outcomes for splenectomized and nonsplenectomized groups.11 Four studies reported splenectomy as an outcome on its own or within a composite outcome.9,32 No studies were identified to draw conclusions on the place of splenectomy in second-line therapies.
The intended economic evaluation was to address the following research questions:
Given the findings from the CADTH clinical review, conclusions regarding the comparative clinical effects of treatments could not be drawn. As such, a primary economic evaluation could not be conducted due to both evidence gaps and limitations with the existing clinical evidence. Specifically, the disease trajectory of patients with ITP across different treatments could not be modelled due to limited and heterogeneous treatment-specific response data; heterogeneity across studies in the definition of treatment response thresholds, bleeding, and other adverse outcomes; and limited information on the time-dependent risk of bleeding events across treatments. Further, both baseline characteristics for patients with ITP and the measurement of outcomes of interest on standard or care treatment within a Canadian context were lacking.
Because CADTH conducted a scoping exercise for the economic evaluation, Appendix 11 provides a review of existing published decision-analytic models in this clinical area, an in-depth discussion of the limitations of the current evidence that precludes the conduct of an economic evaluation but can help inform areas for further research, and, a proposed structure for the “ideal” decision-analytic model that would be able to address the above research questions, highlighting the potential model inputs required and data gaps that would need to be resolved.
The aim of this HTA was 2-fold: to determine what treatment(s) should be used in adult patients with ITP who have failed first-line treatment, and to establish the place in second-line therapy of splenectomy in adult patients with ITP. The project scope and research protocol were informed by engaging with patient groups to better understand the challenges associated with ITP and current treatments. A total of 15 publications met the final inclusion criteria, reporting findings from 10 unique trials on the use of subsequent-line ITP treatments. No study with comparisons to splenectomy could be identified from the literature search.
Patients were eligible for inclusion in the studies if they were adults with a newly diagnosed, persistent, or chronic ITP, with a platelet count of less than 30 × 109/L. Four RCTs compared eltrombopag to placebo over a minimum of 6 weeks, 1 RCT compared avatrombopag to placebo over 26 weeks, and 3 RCTs compared romiplostim to placebo or standard of care. Two RCTs evaluated interventions other than TPO-RA, including 1 comparing rituximab to placebo over 4 weeks and 1 comparing fostamatinib to placebo over 26 weeks. Trials permitted concomitant use of other ITP medications including corticosteroids, azathioprine, and danazol. The primary efficacy outcome was platelet count response; however, there were several differences across the included trials in the way that platelet count response was defined, measured, and reported.
A feasibility assessment determined that an NMA of RCTs was not feasible due to clinical heterogeneity between studies and methodological constraints. An alternate NMA with simplified assumptions based on the limitations of the network was conducted and results were not deemed informative. For these reasons, conclusions from the NMA could not be drawn. Rather, a narrative synthesis of the evidence was conducted for the included studies.
The risk of bias across the 10 studies ranged from low9-11 to high.12-15 An overall limitation of the included trials was the small sample size, which limited the level of precision and affected the certainty in the findings. The study populations were generalizable to most ITP patients, bearing in mind that there were likely differences in local background standard of care treatments across the various countries where the studies were performed. Baseline characteristics between treatment groups were balanced in most cases, with few trials having baseline imbalances that could have introduced bias.13,25 Study discontinuation rates were significant in some studies.12,15 Finally, the concomitant use of other ITP medications, likely to be a significant confounding factor, was not always reported in publications, which could affect the interpretation of the findings.
In patients with prior ITP treatment, 4 RCTs found that eltrombopag was superior to placebo at generating a long-lasting platelet count response (at least 6 weeks and up to 6 months), reducing bleeding events, and improving HRQoL. Eltrombopag was associated with a reduced need for rescue medication and concomitant ITP medications compared to placebo, including corticosteroids. Eltrombopag appeared to be well tolerated in the studies and no strong safety signal was identified.
In patients who have experience with prior ITP therapies, 3 RCTs found that romiplostim induced a platelet count response compared to placebo or standard of care. Romiplostim also reduced the need for rescue medication and was associated with improvements in HRQoL compared with placebo and standard of care. Although romiplostim was generally well tolerated, there were inconsistencies in the harm profiles across studies.
In patients who have experience with prior ITP therapies, avatrombopag was assessed in a single, small, placebo-controlled study that had a high risk of bias, which contributed to a high level of uncertainty and limited the conclusions that could be drawn from this trial.
In patients with primary ITP who failed a short course of corticosteroid therapy, 1 RCT with a moderate risk of bias evaluated treatment failure (i.e., the absence of platelet response) in those receiving either rituximab or placebo. However, the a priori defined threshold for statistical significance for treatment failure was not met, limiting the conclusions that could be drawn from the trial.
In patients who have experience with prior ITP therapies, 1 RCT with a high risk of bias found that fostamatinib was associated with a higher proportion of patients with stable platelet response compared to placebo. However, the magnitude of the treatment response appeared low and discontinuation rates were high.
The systematic review was undertaken at the request of public drug plans to determine the efficacy and safety of therapies in adult patients with ITP who have failed first-line treatments. In consulting with patient groups, patients with ITP expressed a desire for more accessibility to additional therapies for ITP and a desire to avoid splenectomy.
A narrative synthesis of the trials identified in the systematic review found that eltrombopag and romiplostim increased platelet count response and demonstrated an improvement in symptoms and quality of life compared to placebo. Interpretation of the effectiveness of avatrombopag, rituximab, and fostamatinib were inconclusive due to the limited availability of evidence (i.e., 1 single trial for each intervention with varying risks of bias). Due to a scarcity of evidence, conclusions could not be drawn about the outcomes of time to platelet response, time to treatment failure, emergency department visits, hospitalization, and treatment-free remission.
No head-to-head study was identified, and the trials included in the systematic review were limited to comparisons to placebo or standard of care. In addition, methodological consideration prevented an NMA from being conducted. As such, direct comparisons of effectiveness between ITP treatments cannot be made. Similarly, there was insufficient evidence to inform the efficacy or sequence of subsequent-line therapies in subgroups of adult patients with ITP or the role of splenectomy in the sequency of subsequent-line therapies.
A de novo economic model could not be conducted due to various clinical data gaps; therefore, the comparative cost-effectiveness of treatments for ITP remains unknown.
The systematic review was developed using robust methodology. The research protocol was developed a priori and registered with the PROSPERO database, and a detailed scoping plan was posted publicly for stakeholder input. Input from those with lived experience was used to inform the research protocol. The literature search was comprehensive and was also publicly posted for stakeholder feedback.
Evidence collection, data extraction, and evaluation of the quality of the studies were completed by a single reviewer, with verification by a second. Conflicts in data collection were adjudicated by a third reviewer. Heterogeneity across trials was assessed and the analytical approach for the NMA was aligned with ISPOR guidelines and employed a standard methodology.
Following an NMA feasibility analysis, it was determined that an alternate and more simplified NMA be conducted. The number of trials that contributed to the NMA was limited and results reported were highly heterogeneous. Due to the limited size of the network, it was not possible to adequately account for the level of heterogeneity, and as a result, the variation around estimated effects were extremely wide and generally not informative. Thus, we were unable to use the findings of the NMA to draw conclusions for the report. The narrative review was based on limited availability of evidence, including single trials with low sample sizes for avatrombopag, rituximab, and fostamatinib.
Given that an economic model could not be conducted due to various clinical data gaps, the comparative cost-effectiveness of treatments for ITP is unknown.
Public drug plans may consider revisiting reimbursement criteria for eltrombopag and romiplostim when used in ITP. CADTH suggests that interested jurisdictions request an implementation advice panel to suggest strategies (e.g., discontinuation criteria) to manage the clinical and economic uncertainty identified in the report. Outlining what specific discontinuation criteria should be considered is beyond the scope of this report.
Patients with ITP expressed a desire to avoid surgery (splenectomy) and the side effects associated with long-term steroid use. Patients also voiced a desire to see different types of therapies accessible to patients in Canada, including avatrombopag, fostamatinib, eltrombopag, romiplostim, and rituximab.
To determine what treatment(s) should be used in adult patients with ITP who have failed first-line treatments, a systematic review of the efficacy of treatments was undertaken at the request of public drug plans. A narrative review of 15 publications, reporting findings from 10 RCTs, found that eltrombopag and romiplostim increased platelet count response and improved ITP symptoms and HRQoL compared to placebo. Interpretation of the effectiveness of avatrombopag, rituximab, and fostamatinib was inconclusive due to limited evidence (i.e., 1 single trial for each intervention, with varying risks of bias). Of note, recently, CADTH recommended that fostamatinib not be reimbursed for the treatment of thrombocytopenia in adult patients with chronic ITP who have had an insufficient response to other treatments.33 Avatrombopag for the treatment of thrombocytopenia in adult patients with chronic ITP who have had an insufficient response to a previous treatment is currently under CADTH Reimbursement Review.34
Due to the scarcity of evidence, conclusions could not be drawn about the outcomes of emergency department visits, hospitalizations, or treatment-free remission for any of the interventions. No direct treatment comparisons were identified, and methodological consideration prevented an NMA from being conducted; therefore, direct comparisons of effectiveness between ITP treatments could not be made. Similarly, there was insufficient evidence to inform the efficacy or sequence of subsequent-line therapies in specific subgroups of adult patients with ITP or the place in therapy of splenectomy. The comparative cost-effectiveness of treatments for ITP is unknown; further research is needed to compensate for clinical data gaps to inform an appropriate and relevant economic evaluation.
adverse event
double blind
health-related quality of life
health technology assessment
immune thrombocytopenia
intent to treat
IV immunoglobulin
network meta-analysis
open label
randomized controlled trial
serious adverse event
thrombopoietin receptor agonist
Note that this appendix has not been copy-edited.
Interface: Ovid
Databases:
Date of search: July 23, 2021
Alerts: Monthly search updates until July 26, 2022
Search filters applied: Systematic reviews; meta-analyses; network meta-analyses; health technology assessments; guidelines; overview of reviews; randomized controlled trials; and controlled clinical trials.
Limits:
Syntax Guide.
Produced by the US National Library of Medicine. Targeted search used to capture registered clinical trials.
[Search -- Studies with results | ITP or immune thrombocytopenia]
International Clinical Trials Registry Platform, produced by WHO. Targeted search used to capture registered clinical trials.
[Search terms -- ITP or immune thrombocytopenia]
Produced by Health Canada. Targeted search used to capture registered clinical trials.
[Search terms -- ITP or immune thrombocytopenia]
European Union Clinical Trials Register, produced by the European Union. Targeted search used to capture registered clinical trials.
[Search terms -- ITP or immune thrombocytopenia]
Search dates: August 9 to 15, 2021, July 26, 2021.
Keywords: ITP or immune thrombocytopenia
Limits: none
Relevant websites from the following sections of the CADTH grey literature checklist Grey Matters: A Practical Tool for Searching Health-Related Grey Literature were searched:
The complete search archive of sites consulted for this report is available on request.
Note that this appendix has not been copy-edited.
The aim of this HTA to inform 2 policy questions:
The objective is to review the available evidence on the use of therapies for ITP in adults with the objective of determining their comparative effectiveness.
The project will address the following research questions. Details on the specific interventions and outcomes are included in the selection criteria (See Protocol Table 2: Selection criteria).
The study population are adult patients with ongoing active immune thrombocytopenia who have failed first-line treatments. Among the trials that meet the eligibility criteria (See Protocol Table 2: Selection criteria), data will be grouped according to the following patient characteristics and used for subgroup analyses.
For studies including more than 1 of these subpopulations, with no subgroup analysis reported, the study will be considered to belong to 1 of these subgroups if 80% of more of the enrolled patients belong to that subgroup.
Intervention and comparators of interest will include all drugs specified here or belonging to the overall drug category specified here:
We will include studies reporting the standard doses of the treatment approved by Health Canada and doses considered of clinical relevance based on expert guidance or international clinical practice guidelines. All data on dosages and treatment duration for each included treatment will be reported.
In cases where there are variations in the dose or dosing schedule of 1 treatment across different studies, we will seek published literature to inform on the effect of such variations. In cases where no published literature is satisfactory, the clinical expert opinion on whether these differences are substantial enough to be a treatment effect modifier, in which case they will be analyzed as separate interventions. However, if dose and dosing schedule differences do not represent a substantial treatment difference in the opinion of the clinical expert, then they will be pooled together as a single intervention. In cases of a pooled treatment with variations in the dose or dosing schedule, a subsequent sensitivity analysis will be conducted to determine the effect of the pooling.
Any follow-up period without active intervention or open-label extension phase where all patients receive the same intervention will not be eligible for analysis.
The proportion of patients with complete response, response, or no response as defined by the authors. We will also consider the mean difference in platelet count.
Time from starting treatment to time of achievement of complete response or response.
The proportion of patients with bleeding events.
Number of emergency room visits for ITP-related clinical problem per patient.
Number of hospitalizations for ITP-related clinical problem per patient.
The assessment of quality of life using any relevant metric will be eligible for analysis if an aggregate score is provided that represents the change from baseline to the end of treatment in the intervention groups. Domain-specific scores will not be eligible for analysis. Disease-specific measures will not be pooled with generic measures. Total scores will be converted to standardized mean differences and analyzed as a continuous outcome.
The proportion of patients with fatigue as reported by the authors or mean difference in fatigue scores.
The proportion of patients with treatment-free remission
The proportion of patients who are given rescue medication, i.e., IVIG or corticosteroids.
The proportion of patients who reduce or discontinue their corticosteroid use.
Proportion of patients reported to have any AEs.
Proportion of patients reported to have any serious AEs. Serious AEs as defined by the authors of the study or when a patient dies, has a life-threatening event, is hospitalized, or experiences disability or permanent damage.
Proportion of patients reported to have withdrawn from the study due to AEs.
Proportion of patients who die during the study.
Risk of bias in the individual trials will be assessed using the Cochrane risk of bias 2 (RoB 2) tool.24 Risk of bias assessments will be completed in duplicate for each trial. The RoB 2 tool can be used to assess bias arising from the randomization process, due to deviations from intended interventions, due to missing outcome data, measurement of the outcome, and bias in selection of the reported results. For each of these domains, a judgment of high risk of bias (the study is judged to be at high risk of bias in at least 1 domain or some concerns for multiple domains), some concerns (the study is judged to raise some concerns in at least 1 domain, but not at high risk of bias for any domain) or low risk of bias (the study is judged to be at low risk of bias from all domains) can be made. An assessment will be conducted separated for each outcome, and an overall risk of bias will be reached using the signalling questions algorithm included in the tool. We do not anticipate including any crossover or cluster RCTs. If we find any crossover or cluster trials, we will only use estimates that account for the correlation between participants in cluster trials and within participants in crossover trials. For each trial, the estimate regarding the effect of randomization i.e., ITT (as opposed to the effect of adhering to the intervention i.e., per protocol) will be considered. The results of these assessments will be displayed using the robvis tool for visualizing risk of bias. Overall risk of bias will be used as the basis for sensitivity analyses in which we will drop the studies at high risk of bias and examine how this affects the results.
In addition to searching for unpublished and ongoing studies, we will investigate small-study effects using Harbord’s test for binary outcomes and Egger’s test for continuous outcomes. 35 We will also inspect the network funnel plot for asymmetry.
The evidence network for each outcome will be displayed graphically as a network plot. The lines will indicate direct comparisons and the thickness of the line will reflect the number of studies in each comparison. The vertices (“nodes”) will represent individual interventions. Sizes of nodes will be directly proportional to the total sample size of participants contributing data to the node across all studies involving that intervention.
A standardized excel extraction sheet is piloted to abstract data from included studies. Modification to the extraction sheet will be conducted after a pilot of 5 studies. The data extraction will cover information pertaining to each study overall publication characteristics, study design, intervention, comparators, patients’ characteristics, outcomes definitions, outcomes results, and AEs.
The primary publication for each unique study included will be used for data extraction, except where multiple publications for a single primary study are found. Multiple publications for a unique study (e.g., substudies, reports of additional outcomes or longer follow-up) will be used if they include data relevant to our research question. The data will be entered as coming from the same study, but the related articles will be referenced as needed.
Where reported, patients’ baseline characteristics and clinical efficacy outcomes will be abstracted from the ITT population in each study while AEs will be abstracted from the per-protocol i.e., safety population. The ITT approach implies the availability of a complete dataset i.e., no missing data. If an ITT analysis is not reported in the paper, we will use the per-protocol analysis. We will conduct sensitivity analyses to determine if the analytical strategy used in the trial affects the effect estimates. Trial definitions for intention to treat, modified intention to treat, and per-protocol data will be noted and compared for consistency.
For dichotomous outcomes, we use the number of patients with the event and the total number of patients in each arm. For continuous outcomes, we will extract the mean, standard deviation, and number of participants in each arm. For time to event outcomes, we will extract the mean/median time to event, standard deviation and the number of participants per-arm (arm-based approach). If only comparative effect estimates are reported, the NMA can be conducted with generic measures of effect such as the log of the odds ratio and hazard ratio and the corresponding standard errors (contrast-based approach).36 For continuous outcomes, the standardized mean difference and its standard error can be used.
Screening, data extraction and risk of bias assessments will be conducted by 1 reviewer and verified by a second reviewer using a standardized data extraction form developed a priori, which will be piloted and modified as necessary.
Prior to commencing the NMA, a qualitative feasibility assessment will be undertaken to assess clinical and methodological heterogeneity in the available data. Trials included in the same evidence network will be compared to identify any potential heterogeneity based on differences in participants, interventions, comparisons, outcomes, or study design. If substantial heterogeneity exists in certain comparisons or subsets of studies and the data are not meaningful to pool, narrative summaries of findings will be reported, including the report on the types of trials required to fill the knowledge gap. Quantitative analyses will not be conducted where there is insufficient data.
Where potential clinical or methods-related heterogeneity is identified and the degree of heterogeneity is considered large, and if we have data on potential effect modifiers, we will use meta-regression (if there are 10 or more studies) to adjust for these sources of heterogeneity.37 If meta-regression is not possible, we will perform subgroup analyses or sensitivity analyses. Our process for assessing the feasibility of a NMA is outlined in Appendix 5. If the feasibility assessment is positive, we can continue to the next steps in this analysis plan. If feasibility assessment is negative, then we may summarize why the network is not sufficient to support a valid NMA.
Provided the NMA is deemed feasible, the identified network will be evaluated and tested for homogeneity.38 For each study in the network, the patient population, definitions of interventions, definitions of outcomes, inclusion and exclusion criteria, and designs will be described and examined for similarity across the network. Patient baseline characteristics for potential effect modifiers identified by clinical experts will be examined for homogeneity across studies.
We will observe the logical structure of the graphs i.e., whether direct evidence is similar to indirect evidence, and evaluate the fitness of the statistical models. We will also measure the confidence intervals of the difference between the direct and indirect comparisons. We will use the node-splitting method, which will allow us to assess consistency locally within each closed loop of the network. We will also run an overall (global) test for inconsistency using the Wald test. 39
We will use sensitivity analyses to determine the role of effect modifiers and meta regression to adjust for identified effect modifiers.
The NMAs will be conducted under a Frequentist framework using the multivariate meta-analysis approach.39,40 Random-effects models will be used as the primary approach when feasible due to the anticipated clinical and methodological heterogeneity across studies. For sparse networks,41 fixed effects models will be considered when the available network for a given outcome is insufficient for estimating a random-effects model, i.e., the confidence intervals are inaccurate. For each NMA we will use the design-by-treatment model (global test) to assess network coherence, and the side-splitting method to assess loop specific coherence,40 computed as the difference between direct and indirect evidence. If there is large incoherence, we will use an inconsistency model to conduct the NMA.
The multivariate random-effects meta-analysis is estimated using restricted maximum likelihood that will be applied to arm-based or contrast-based data.36 The results will be presented as network plots, contribution plots, and league tables.42
If an NMA is not feasible. We will conduct a pairwise random-effects meta-analysis, comparing all available treatments to placebo or to the most common (i.e., reported in most studies) available comparison or the most relevant standard of care. If this alternative approach is also not informative a narrative synthesis will be conducted.
For binary outcomes, odds ratios and 95% confidence intervals will be reported. In cases when events are rare, the Peto odds ratio will be used. For continuous outcomes measured on the same scale, mean differences and 95% confidence intervals will be reported. For continuous outcomes measured on different scales, standardized mean differences and 95% confidence intervals will be reported. If insufficient data are found for conducting meta-analyses, narrative descriptions will be provided.
For the network or pairwise meta-analysis, we will conduct subgroup analyses comparing outcomes in trials belonging to the following subgroups:
We will conduct additional sensitivity analysis by excluding studies with high risk of bias or studies with missing data to a relevant outcome. We will also explore the effect of outliers and influential trials by running models with and without them. All analyses will be conducted using Stata 16.0 (StataCorp, College Station, Texas, US). All comparisons will be 2 tailed using a threshold P ≤ 0.05.
Note that this appendix has not been copy-edited.
Note that this appendix has not been copy-edited.
List of Excluded Studies During Data Extraction.
Note that this appendix has not been copy-edited.
Prior to conducting an NMA, we performed a qualitative feasibility assessment to determine if the data were sufficiently similar to pool from a clinical and a methodological perspective. Clinical heterogeneity is concerned with the homogeneity in participants, interventions, comparisons, and outcomes, while methodological heterogeneity is concerned with the design and analysis.
While all the included studies were all phase III or IV two-arm parallel group randomized trials, we identified some design concerns. Three of the included studies were OL.14,15,43 The study durations ranged from 6 to 78 weeks. These differences on their own would not have precluded an NMA.
We noted some differences in how data were analyzed across some outcomes. For example, in some studies the same outcome was reported as hazard ratio and in others only figures were reported. In some instances, percentages were reported without numerators or denominators, thus precluding the estimation of variances.
Participant homogeneity was assured based on the inclusion criteria. All the studies met the inclusion criteria which included adult patents with ongoing active ITP who had failed first-line treatments.
Intervention and comparator homogeneity was ensured based on the inclusion criteria. All the included studies used either rituximab (100mg/m2/week, 375mg/m2/week), Eltrombopag (12.5mg once daily, 25mg once daily, 50mg once daily), Romiplostim (1mcg/kg qw, 3mcg/kg qw), Fostamatinib (100 mg bd) or Avatrombopag (20 mg once daily) compared to placebo or to each other. Placebo groups were uniform across studies.
There was considerable outcome heterogeneity. For example, platelet count response was defined at different thresholds (≥ 20,000, ≥ 30,000, ≥ 50,000 and ≥ 100,000 platelets/µl). Some studies reported more than 1 threshold, but even when we had multiple studies with the same threshold, the definitions were different. These differences included the number of measurements and the timing of the measurements. Three examples are shown below:
“Platelet counts ≥ 50 x 109 on at least 4 of the 6 clinic visits occurring every 2 weeks during weeks 14–24 inclusive”
“Proportion of patients with platelet counts ≥ 50 000/μL and at least twice the baseline amount”
“At least 1 platelet count ≥ 50 x 109 within first 12 weeks”
This heterogeneity was present for other outcomes such as bleeding, in which some studies used WHO-defined grades and others did not.
We attempted to pool all the studies with a threshold of ≥ 20,000 and ≥ 30,000 and found that the threshold impacted the relative effects of treatments (e.g., at higher thresholds, romiplostim was less likely to be better than placebo).
In addition to the concerns with heterogeneity noted above, the data were sparse i.e., we had too few comparisons per study and any attempts to create smaller more homogenous groups further aggravated the issue, leading to very wide confidence intervals. These wide confidence intervals produced results (most active treatments were not statistically significantly better than placebo) that were contradictory to the findings from the individual studies.
Further, consistency could not be estimated since there was limited direct evidence for the same comparison (almost all comparisons were treatments versus placebo) i.e., there were no closed loops.
Based on this, we concluded that the available data were not meaningful to pool and have presented narrative summaries of the findings. This decision is in line with recommended guidance for addressing feasibility in NMA and our statistical analysis plan.
Note that this appendix has not been copy-edited.
Details of Included Studies — Eltrombopag as Intervention (N = 4 Studies).
Patient Characteristics — Eltrombopag as Intervention (N = 4 Studies).
Patient Disposition — Eltrombopag as Intervention (N = 4 Studies).
Four DB RCTs evaluated eltrombopag compared to placebo.9,10,25,27 Detailed outcome results are presented in Table 8.
The use of eltrombopag was associated with higher proportions of responders compared with placebo at week 6 in 3 trials (60% versus 0% respectively, p value not reported;25 58% versus 6% respectively, P < 0.001;9 and 59% versus 16% respectively, P < 0.001;27) as well as throughout 6 months in another trial (79% versus 28%, respectively; P < 0.0001).10 These were the primary outcome results in the studies, where response was mainly defined as a platelet count ≥ 50 000/µL, with slight variations across trials.
Other outcome measurements for platelet count response were included as secondary or exploratory outcomes with comparative data reported in 2 of these 4 studies.10,27 The use of eltrombopag was associated with a higher proportion of responders compared with placebo, that is, at any time over the 6-week period in 1 trial,27 and at ≥ 75% of the monthly assessments during the 6-month follow-up in the other trial.10
Comparative data for any bleeding events were reported in 3 studies according to the WHO bleeding scale.9,10,27 The use of eltrombopag was associated with fewer bleeding events compared with placebo in all 3 trials (OR 0.28; 95% CI 0.13 to 0.59; P < 0.001;9 39% versus 60% respectively, P = 0.029;27 79% versus 93% respectively, P < 0.00110).
HRQoL was assessed in 1 study using various scales.10 The use of eltrombopag was associated with greater improvements from baseline to week 26 compared to placebo in the FACT-Th questionnaire and in some of the SF-36v2 scales (physical role, vitality and emotional role subscales, as well as in the mental component summary score), supporting an improvement in HRQoL among patients treated with eltrombopag versus placebo. Greater improvements from baseline were also observed with eltrombopag compared with placebo in the SF-36v2 physical component summary, the MEI-FS and the FACIT-Fatigue questionnaires; however, statistical significance was not reached for these subscales. Detailed results are shown in Appendix 6. Caution is advised however when interpreting nonsignificant findings to draw conclusions regarding difference between groups in HRQoL.
Two studies reported data pertaining to the use of rescue medications.9,10 In both trials, eltrombopag was associated with a smaller proportion of patients necessitating rescue treatments compared with placebo (9% versus 34% respectively, OR = 0.13 with 95% CI 0.05 to 0.37, P < 0.001;9 18% versus 40% respectively, OR = 0.33 with 95% CI 0.16 to 0.64; P = 0.00110). Rescue medications were defined similarly as either a new ITP medication, an increase in the dosage of a concomitant ITP medication, platelet transfusion, or splenectomy.9,10
Two studies assessed the reduction or discontinuation of concomitant ITP medication;9,10 however, comparative data were reported in only 1 of the 2 studies.10 Eltrombopag was associated with a reduction in the proportions of patients with reduction or discontinuation of ITP baseline treatments compared with placebo (59% versus 32%, respectively; OR = 3.10 with 95% CI 1.24 to 7.75; P = 0.02). The publication indicates that the drugs most frequently tapered or discontinued in both treatment groups were corticosteroids.
No data were reported to inform on the effectiveness of eltrombopag versus placebo for the outcomes of time to platelet response, emergency department visits, hospitalization, patient-reported specific symptoms of ITP and treatment-free remission.
There was inconsistency across the 4 studies in the proportions of patients with at least 1 AE, which ranged from 25% to 92% across treatment groups. In 2 studies, a numerically higher proportion of patients in the eltrombopag group experienced AEs compared to patients in the placebo group.25,27 The proportions of patients experiencing AEs in the 2 other studies were relatively similar between treatment arms.9,10
Two studies reported the proportions of patients experiencing serious adverse events (SAEs),9,25 which were less than 10% of patients per treatment arm. Only 1 trial reported the nature of the events; in this study, 1 patient receiving eltrombopag experienced a transient ischemic attack.25
Discontinuations of study treatment due to AEs were reported in 3 trials.9,10,27 The proportions of patients who discontinued were less than 10% per treatment arm and differed between treatment arms only in 1 trial (9.6% with eltrombopag versus 6.6% with placebo).10
Mortality was assessed as a harm outcome in the studies and therefore, is reported as such in this report. Three studies reported mortality data;9,10,27 overall, 2 patients in placebo groups died over the study period. The causes of death were undetermined or not individually reported.
No data were reported to specifically inform on the comparative effectiveness of eltrombopag versus placebo for the outcome of infection, except for patient-reported AEs of some types of infections.
Summary of Key Efficacy Outcomes — Eltrombopag as Intervention (N = 4 Studies).
Summary of Key Harms Outcomes — Eltrombopag as Intervention (N = 4 Studies) .
Note that this appendix has not been copy-edited.
Details of Included Studies – Avatrombopag as Intervention (N = 1 Study).
Patient Characteristics — Avatrombopag as Intervention (N = 1 Study).
Patient Disposition — Avatrombopag as Intervention (N = 1 Study).
One DB RCT evaluated avatrombopag compared to placebo.12 Detailed outcome results are presented in Table 13.
The use of avatrombopag was associated with a higher cumulative mean number of weeks of platelet response compared with placebo at week 26 (12.0 ± 8.75 weeks in the avatrombopag group versus 0.1 ± 0.49 week in the placebo group). The cumulative number of weeks of platelet response was the primary outcome in the trial, where response was defined as a platelet count ≥ 50 000/µL in the absence of rescue therapy. The use of avatrombopag was associated with a higher proportion of responders at day 8 compared to placebo (66% versus 0%, respectively; difference in response rate 65.63, 95% CI 49.17 to 82.08; P < 0.0001).
Comparative data for bleeding events were reported according to the WHO bleeding scale. The proportions of patients with any bleeding event were 44% with avatrombopag and 53% with placebo (p nonsignificant).
The proportions of patients who received rescue medication were 22% with avatrombopag and 12% with placebo (p nonsignificant).
The proportions of patients who reduced or discontinued concomitant ITP medication were 33% with avatrombopag and 0% with placebo (difference in reduction rate of 33.33, 95% CI 9.48 to 57.19; p nonsignificant).
No data were reported to inform on the comparative effectiveness of avatrombopag versus placebo for the outcomes of time to platelet response, emergency department visits, hospitalization, HRQoL, patient-reported specific symptoms of ITP and treatment-free remission.
A numerically higher proportion of patients in the avatrombopag group experienced AEs compared to patients in the placebo group (97% versus 59%, respectively).
Nine patients in the avatrombopag group experienced SAEs. There was 1 SAE in the placebo group, the nature of which was not reported (28% versus 6%, respectively).
Three patients (9%) in the avatrombopag group discontinued study treatment due to AEs; the nature of the events was not reported. There was no withdrawal due to adverse event in the placebo group.
Mortality was assessed as a harm outcome in the study and therefore, is reported as such in this report. No patient died over the course of the study.
No data were reported to specifically inform on the comparative effectiveness of avatrombopag versus placebo for the outcome of infection, except for patient-reported AEs of some types of infections.
Summary of Key Efficacy Outcomes — Avatrombopag as Intervention (N = 1 Study).
Summary of Key Harms Outcomes — Avatrombopag as Intervention (N = 1 Study).
Note that this appendix has not been copy-edited.
Details of Included Studies — Romiplostim as Intervention (N = 3 studies).
Patient Characteristics — Romiplostim as Intervention (N = 3 studies).
Patient Disposition — Romiplostim as Intervention (N = 3 studies).
Three RCTs evaluated romiplostim compared to placebo.11,13,14 Detailed outcome results are presented in Table 18.
The 3 romiplostim trials used different outcome measurements as primary end point for platelet count response (number of weeks in response, response rates and incidence of treatment failure), as well as different definitions and thresholds for platelet response.
In 1 12-week DB RCT, the use of romiplostim was associated with a higher number of weeks of platelet count response compared with placebo (median 11 weeks, IQR 9 to 12; versus 0, IQR 0 to 0, respectively; P < 0.0001).13 Platelet count response was defined as ≥ 50 000/µL in the publication. The use of romiplostim was also associated with a higher proportion of patients with platelet count increase ≥ 20 000/µL compared with placebo (96% versus 25%, respectively; P < 0.001).13
One 24-week DB RCT evaluated the platelet count response rate during at least 8 weeks, which was higher with romiplostim compared with placebo (49% versus 2%, respectively; P < 0.01).11 In this study, platelet count response was defined as ≥ 6 of the last 8 weeks of treatment with weekly platelet counts ≥ 50 000/µL and no rescue medication. The use of romiplostim was also associated with a higher proportion of patients with ≥ 4 weekly platelet count response at any time compared with placebo (83% versus 7%, respectively; P < 0.0001).11
The third study was a 52-week OL RCT with incidence of treatment failure (i.e., patients who did not have a platelet response) as primary outcome, which was lower in patients receiving romiplostim compared with standard of care (11% versus 30%, respectively; between-group difference of 0.31; 95% CI 0.15 to 0.61; P < 0.001).14 Treatment failure was defined in the study as a recurrent platelet count ≤ 20 000/µL despite optimal dosage, a major bleeding or withdrawal due to adverse event. The use of romiplostim was also associated with a higher median platelet count between week 2 and end of trial compared with standard of care (between-group difference of 2.3 (95% CI 2.0 to 2.6); P < 0.001).14
The 3 studies reported some data for bleeding events; however, definitions for bleeding as well as comparative data were not consistently provided. The proportions of patients with any bleeding event were 36% after 12 weeks with romiplostim versus 83% with placebo (no statistical analysis was performed) in 1 trial;13 57% with romiplostim versus 61% with placebo in the second trial (p nonsignificant);11 and 52% with romiplostim versus 53% with OL standard of care in the third trial (no statistical analysis was performed).14.
HRQoL was assessed in 2 studies using the ITP-PAQ questionnaire11,14 and in 1 study using the EQ-5D.11 The use of romiplostim was associated with greater improvements from baseline to week 24 compared to placebo in the some of the ITP-PAQ scales (symptoms, bother, activity, psychological, fear, social quality of life, and women’s reproductive health – menstrual symptoms) as well as in the ED-5D index score.11 At week 52, romiplostim was associated with greater improvements from baseline compared to OL standard of care in the following ITP-PAQ scales: symptoms, bother, activity, fear, overall quality of life, and social quality of life.14 Standard of care treatments were at the discretion of the investigators and based on local practices or therapeutic guidelines.14
Greater improvements from baseline were also observed with romiplostim compared with placebo at 24 weeks in the EQ-5D VAS score and in the ITP-PAQ fatigue and overall quality of life scales,11 as well as at 52 weeks in the ITP-PAQ fatigue scale;14 however, statistical significance was not reached.
The studies11,13,14 reported some data pertaining to the use of rescue medications. All results are presented in Appendix 6. One trial performed a between-group comparison with sufficient power to show a difference between treatments. In a 24-week DB RCT, romiplostim was associated with a reduction in the proportions of patients necessitating rescue treatments compared with placebo (22% versus 60%, respectively; P < 0.0001).11 Rescue medications were defined in the trial as an increase in the dosage of a concomitant ITP medication or the use of any new drug to increase platelet count.
One study assessed the reduction or discontinuation of concomitant ITP medication.11 More patients in the romiplostim treatment group were able to discontinue a concomitant ITP medication or reduce the dose of at least 1 medication by more than 25% compared with placebo; however, no analysis between treatment groups was reported.
No data were reported to inform on the comparative effectiveness of romiplostim versus placebo or standard of care for the outcomes of time to platelet response, emergency department visits, hospitalization, patient-reported specific symptoms of ITP and treatment-free remission.
AEs were reported in 2 studies.11,13 Almost all patients in these 2 trials experienced at least 1 AE. The most common AEs reported across all studies included headache, nasopharyngitis, fatigue, contusion, epistaxis, arthralgia, and petechiae.
There was inconsistency across the 3 studies in the proportions of patients with SAEs. In 2 studies, the proportions of patients experiencing SAEs were similar between treatment groups and were less than 10% of patients per treatment arm.11,13 In the OL 52-week trial, the proportions of patients with SAEs were much higher, being 23% of patients under romiplostim and 37% of patients receiving standard of care.14 In this study, the most common SAEs reported included thrombocytopenia, pneumonia, peripheral edema, and pulmonary embolism.
Discontinuations due to AEs were reported in 1 study and the proportions of patients who discontinued were similar between groups and less than 10% per treatment arm.11
Mortality was assessed as a harms outcome in the studies and therefore, is reported as such in this report. The 3 studies reported mortality data; however, no death was observed in 1 trial.13 One patient randomized to romiplostim died in each of the 2 other trials, as well as 2 patients in the placebo or standard of care groups over each study period.11,14 The causes of death included cerebral or intracranial hemorrhage, pulmonary embolism, pneumonia, and hepatic failure.
No data were reported to specifically inform on the comparative effectiveness of romiplostim versus placebo for the outcome of infection, except for patient-reported AEs of some types of infections.
Summary of Key Efficacy Outcomes — Romiplostim as Intervention (N = 3 Studies) .
Summary of Key Harms Outcomes — Romiplostim as Intervention (N = 3 Studies) .
Note that this appendix has not been copy-edited.
Details of Included Studies — Interventions other than TPO-RA (N = 2 Studies).
Patient Characteristics — Interventions Other Than TPO-RA (N = 2 Studies).
Patient Disposition — Interventions Other Than TPO-RA (N = 2 Studies).
One DB RCT evaluated rituximab compared to placebo.32 Detailed outcome results are presented in Table 23. It should be noted that the primary outcome for this study, which was treatment failure, did not meet the a priori defined threshold for statistical significance and therefore, formal testing was halted for all secondary outcomes. As a result, any conclusion drawn from this study must consider the increased risk of type I error due to the failed primary end point.
The proportions of patients with overall response in the trial (platelet count ≥ 30 000 /µL) were 73% with rituximab versus 67% with placebo (p nonsignificant); for complete response (platelet count ≥ 100 000 /µL), the proportions were 51% with rituximab versus 39% with placebo (p nonsignificant).
The proportions of patients with any bleeding symptoms were 38% with rituximab versus 50% with placebo (p nonsignificant).
No data were reported to inform on the comparative effectiveness of rituximab versus placebo for the outcomes of time to platelet response, emergency department visits, hospitalization, HRQoL, reduction, or discontinuation of concomitant ITP medication, patient-reported specific symptoms of ITP, need for rescue medication, and treatment-free remission.
A higher proportion of patients in the rituximab group experienced AEs compared to patients in the placebo group (40% versus 24%, respectively). The most common AEs included infections, influenza, throat irritation, headache, bronchitis, and pyrexia.
No data were reported to inform on the comparative effectiveness of rituximab versus placebo for the outcome of SAEs.
No patient experienced withdrawal due to adverse event throughout the study duration.
Mortality was assessed as a harms outcome in the studies and therefore, is reported as such in this report. One patient in the placebo group died over the study period. The cause of death was gastrointestinal bleeding.
The use of fostamatinib was associated with higher proportions of patients with what was labelled as stable platelet response by week 24 compared with placebo (18% versus 2%, respectively; P = 0.0003).15 This was the primary outcome in the trial and was defined as a platelet count ≥ 50 000/µL on at least 4 of the last 6 scheduled visits during weeks 14 to 24. Despite being statistically significant, the proportions of patients with platelet response appeared low in the study. It should be noted however that the median time since ITP diagnosis in the study was approximately 8 years. There was also an improvement with fostamatinib in the proportions of patients with overall response (platelet count ≥ 50 000/µL) at any time compared with placebo (43% versus 14%, respectively; P = 0.0006).
Greater improvements from baseline were observed with fostamatinib compared with placebo in the proportions of patients receiving rescue medication (30% versus 45%, respectively; p nonsignificant).
No data were reported to inform on the comparative effectiveness of fostamatinib versus placebo for the outcomes of time to platelet count response, bleeding, emergency department visits, hospitalization, HRQoL, patient-reported specific symptoms of ITP, reduction or discontinuation of concomitant ITP medication and treatment-free remission.
A higher proportion of patients in fostamatinib group experienced AEs compared to patients in the placebo group (83% versus 75%, respectively). The most common AEs included diarrhea, hypertension, nausea, dizziness respiratory infections, and ALT increase.
A total of 13% of patients in the fostamatinib group experienced SAEs, which is numerically lower than the proportion of patients experiencing SAEs in the placebo group (21%). The most common AEs included thrombocytopenia, epistaxis, and menorrhagia.
The proportions of patients who discontinued due to AEs were 10% in the fostamatinib group and 8% in the placebo group.
Mortality was assessed as a harm outcome in the studies and therefore, is reported as such in this report. One patient in each group died over the study period. The causes of death were myeloma and sepsis.
No data were reported to specifically inform on the comparative effectiveness of fostamatinib versus placebo for the outcome of infection, except for patient-reported AEs of some types of infections.
Summary of Key Efficacy Outcomes — Interventions Other Than TPO-RA (N = 2 Studies).
Summary of Key Harms Outcomes — Interventions Other Than TPO-RA (N = 2 Studies).
Note that this appendix has not been copy-edited.
Critical Appraisal of Included Studies.
Risk of Bias Assessments Per Outcome Within Each Study (Jurczak, Arnold, Yang, Ghanima, Shirasugi).
Risk of Bias Assessments Per Outcome Within Each Study (Kuter 2010, Bussel 2018, Eisai, Cheng).
Risk of Bias Assessments Per Outcome Within Each Study (Tomiyama, Kuter 2008, Kuter 2009, Kuter, 2010, Kuter 2011, Bussel 2009).
Note that this appendix has not been copy-edited.
As part of the development of the economic evaluation for chronic ITP, a literature review of existing decision-analytic models in chronic ITP was conducted. With the goal to determine whether any of the published economic evaluations could directly address the research questions; and, if not, whether details on the model structure could assist the development of a model. This was not a systematic review and hence likely to be nonexhaustive. Conference abstracts and manuscripts published in a different language than English were excluded.
Table 29 below summarizes the main characteristics of the identified studies. None of the identified studies were directly applicable to the economic analysis research questions. Given that the rationale for conducting this review was to further inform the development of a decision-analytic model, less emphasis was placed on the results of the analysis. Instead, the review focused on the model structure.
Summary of ITP Decision-Analytic Models.
Four studies were identified that had developed a decision-analytic model to support either a cost-effectiveness or a cost-utility analysis. All analyses were models and adopted the same national public health care perspective (except for Tremblay 2020 in which the US payer perspective was used). All studies used a cohort state-transition (Markov) model with discrete time cycles. The cycle lengths varied but all models used a lifetime time horizon. Disease progression, modelled by health states, varied across the published models but could be broadly categorized as response, nonresponse, and death.
As part of the scoping process, existing published models were reviewed, and clinical experts were consulted to determine the most appropriate characterization of ITP. The conceptual structure below was the outcome of the exercise, which clinical experts considered an appropriate representation of the natural history of chronic ITP. Of note, the proposed structure below was developed without taking data availability into consideration.
As the model’s target population is adults with ongoing, active ITP after failure of first-line therapy, the baseline characteristics of the target population ideally would be defined in such a way that resemble the characteristics of ITP patients diagnosed across Canadian jurisdictions. Real world administrative data would be a source to inform this set of model parameters, assuming that the population can be adequately identified through administrative data.
The base-case analysis would be conducted from the perspective of the Canadian publicly funded health care system. The model time horizon would be life-long with weekly cycle lengths. As per CADTH guideline, the discount rates for costs and utilities would be 1.5% annually. We assumed that a weekly cycle length would be a sufficient minimum duration to capture recurrent events (e.g., bleeding) or transitions between health states. A lifetime horizon would be appropriate since some of the modelled events (e.g., severe bleeding/intracranial hemorrhage) can be life altering with an increased risk of mortality.
The model would be an individual-level state-transition model, also known as a microsimulation model, with discrete time cycles. A microsimulation approach was proposed so that time dependency (e.g., differential response and bleeding events over time), the effect of past events, and individual heterogeneity is adequately considered in the model. Treatment choice after nonresponse to therapy is dependent on past treatment exposure, hence, treatment history may be considered a past event that needs to be recorded within the model. An example of individual heterogeneity relevant to consider within this decision problem, as per input from clinical experts consulted by CADTH, was the increased risk of bleeding for individuals who already experienced a bleeding episode. Hence, history of bleeding events should be taken into consideration as this would impact an individual’s risk of bleeding and have associated utility and cost impacts.
Using input from the clinical team, the model should capture the following 6 treatments: rituximab, eltrombopag, romiplostim, fostamatinib, avatrombopag, and splenectomy. It remains unclear presently whether IVIG, immunosuppressants (azathioprine, cyclosporine, cyclophosphamide, mycophenolate mofetil, danazol, dapsone) and corticosteroids constitute real world treatment alternatives.
The proposed model structure would have 4 states: nonresponse, response, and ITP-related death and death from all-cause mortality. The model diagram is presented below (Figure 6). As patients transition between states, they can experience bleeding (major and minor), experience a serious infection, switch treatment, or require rescue medication. Only some of these events are included in the model diagram for simplicity. All states lead to both ITP-related death (through bleeding or infection) and non-ITP-related death (death from other causes). These transition arrows are omitted from the model diagram for simplicity. Transitions may be modelled with time and state residence dependencies, where appropriate, to factor time dependencies, the effect of past events, and individual heterogeneity. The relationships among these events are also modelled, as noted in the transition arrows within the Figure 6.
Diagram of the Ideal Model.
The following outcomes would be needed to meaningfully inform the proposed economic model:
Table 30 and Table 31 provide a subset of the probability, utility, and cost input parameters currently available that could be used to develop a chronic ITP decision-analytic model. Of note, provided utility estimates were by platelet count and not categorized by response (where ideally a model could simulate platelet count movement).
Model Input Parameters That Were Successfully Identified and Sourced.
Treatment Costs That Were Successfully Identified and Sourced.
The existing published decision-analytic models informed the fundamental structure of the proposed economic model. A Markov model was considered to be a less effective approach to incorporate the complexities of the course of treatment and disease trajectories, as well as the interrelationships between treatment, AEs, and other clinical outcomes. As such, a microsimulation was felt to allow the flexibility to capture these effects and could represent the ideal modelling approach. The data collected, described in the Model inputs section of this Appendix, provide several core input parameters to this model. However, these data would not be sufficient by themselves to support the development of the model described in the model section of this Appendix.
Given the lack of data at this time, simplifications to the ideal model may be required to what could be feasible, however this should be revisited as data in this space evolves. A simplified model would have the same structure and components as the ideal model except that treatment effects would be time-independent, while bleeding and infection risks would not be treatment or time-specific (given the current lack of information). To overcome these limitations, several data points and input parameter estimates, which we could not successfully source, are required. Time-dependent treatment effects and time-dependent transition probabilities between health states need to come from a robust NMA; where the heterogeneity around the definition of treatment response, both across the studies identified in the NMA, but also in real world clinical practice complicate the comparison of the alternative identified treatment strategies in a common framework. Estimates of population baseline risk of infection, estimates of treatment-specific bleeding, and infection risks are needed, and would ideally be sourced from data reflective of the population of interest (chronic ITP patients in Canada who have failed first-line treatment). More recent and more accurate estimates of health utility associated with treatment response and bleeding or infection outcomes would also be required.
Once data that are reflective of the need of such model are generated and/or collected, we can then inform core components of the economic model, such as the risk of treatment switching or treatment-specific rates of bleeding events. If Canadian patient databases exist, this would ideally be able to be used to inform the baseline characteristics of the model population. If patient data are sufficiently granularly reported (e.g., age, splenectomy at baseline), subgroup analyses of interest, such as splenectomy status, could further be supported in the economic analysis. Another possibility that comes with these data are the ability to model treatment effectiveness more explicitly, by modelling the waning of treatment efficacy and adherence with differential waning across treatments (e.g., Rituximab vs. Fostamatinib). Thus, the effects of treatment switching on treatment effectiveness can also be more explicitly estimated and modelled.
Disclaimer: The information in this document is intended to help Canadian health care decision-makers, health care professionals, health systems leaders, and policy-makers make well-informed decisions and thereby improve the quality of health care services. While patients and others may access this document, the document is made available for informational purposes only and no representations or warranties are made with respect to its fitness for any particular purpose. The information in this document should not be used as a substitute for professional medical advice or as a substitute for the application of clinical judgment in respect of the care of a particular patient or other professional judgment in any decision-making process. The Canadian Agency for Drugs and Technologies in Health (CADTH) does not endorse any information, drugs, therapies, treatments, products, processes, or services.
While care has been taken to ensure that the information prepared by CADTH in this document is accurate, complete, and up-to-date as at the applicable date the material was first published by CADTH, CADTH does not make any guarantees to that effect. CADTH does not guarantee and is not responsible for the quality, currency, propriety, accuracy, or reasonableness of any statements, information, or conclusions contained in any third-party materials used in preparing this document. The views and opinions of third parties published in this document do not necessarily state or reflect those of CADTH.
CADTH is not responsible for any errors, omissions, injury, loss, or damage arising from or relating to the use (or misuse) of any information, statements, or conclusions contained in or implied by the contents of this document or any of the source materials.
This document may contain links to third-party websites. CADTH does not have control over the content of such sites. Use of third-party sites is governed by the third-party website owners’ own terms and conditions set out for such sites. CADTH does not make any guarantee with respect to any information contained on such third-party sites and CADTH is not responsible for any injury, loss, or damage suffered as a result of using such third-party sites. CADTH has no responsibility for the collection, use, and disclosure of personal information by third-party sites.
Subject to the aforementioned limitations, the views expressed herein are those of CADTH and do not necessarily represent the views of Canada’s federal, provincial, or territorial governments or any third-party supplier of information.
This document is prepared and intended for use in the context of the Canadian health care system. The use of this document outside of Canada is done so at the user’s own risk.
This disclaimer and any questions or matters of any nature arising from or relating to the content or use (or misuse) of this document will be governed by and interpreted in accordance with the laws of the Province of Ontario and the laws of Canada applicable therein, and all proceedings shall be subject to the exclusive jurisdiction of the courts of the Province of Ontario, Canada.
The copyright and other intellectual property rights in this document are owned by CADTH and its licensors. These rights are protected by the Canadian Copyright Act and other national and international laws and agreements. Users are permitted to make copies of this document for noncommercial purposes only, provided it is not modified when reproduced and appropriate credit is given to CADTH and its licensors.
About CADTH: CADTH is an independent, not-for-profit organization responsible for providing Canada’s health care decision-makers with objective evidence to help make informed decisions about the optimal use of drugs, medical devices, diagnostics, and procedures in our health care system.
Funding: CADTH receives funding from Canada’s federal, provincial, and territorial governments, with the exception of Quebec.
PROSPERO Registration Number: CRD42021276565
Copyright © 2023 - Canadian Agency for Drugs and Technologies in Health. Except where otherwise noted, this work is distributed under the terms of a Creative Commons Attribution-NonCommercial- NoDerivatives 4.0 International licence (CC BY-NC-ND).
Your browsing activity is empty.
Activity recording is turned off.
See more...
