- WHO: The 2022 recommended influenza strains in southern hemisphere
- New COVID-19 variant: ETA appeared in at least 82 countries and regions
- Tumor microenvironment and drug resistance in immunotherapy
- Can Probiotics prevent Alzheimer’s disease as regulator of brain function?
- The nano antibody from alpaca effectively against COVID-19 and variants
- COVID-19 can’t escape by the new method of virus detection: DNA sensor
Review of two regulatory approved anti-CD19 CAR T cell therapies
- Science: Scientists have broken through CO₂ synthetic starch technology!
- Can you complete mRNA COVID-19 vaccination by eating vegetable salad?
- What COVID-19 vaccines does CDC accept if foreigners enter U.S.?
- WHO chief scientist: Soumya Swaminathan may face death penalty charges?
- Israel new drug for COVID-19: EXO-CD24 can reduce deaths by 50%
Review of two regulatory approved anti-CD19 CAR T cell therapies.
Why do you want to carry out this research?
Anti-CD19 chimeric antigen receptor (CAR) T cell therapy can effectively treat diffuse large B-cell lymphoma (DLBCL), a cancer with limited treatment options and poor outcomes, especially for relapsed or refractory cancers (r/r) Patients with disease.
Axicabtagene ciloleucel (axi-cel) and tisagenlecleucel (tisa-cel) are CAR T cell therapies approved for r/r DLBCL, based on the proven efficacy and controllability in its key clinical trials ZUMA-1 and JULIET, respectively Security.
Since there is no head-to-head trial comparing axi-cel and tisa-cel, this article explores current clinical trial data and real-world evidence (RWE) to assess whether an effective indirect treatment comparison (ITC) can be made.
What can be learned from this research?
The huge differences in study design and patient populations in the JULIET and ZUMA-1 trials hinder a strong and reliable ITC; without substantive and unrealistic assumptions, the ITC method cannot explain such differences. The current real-world data is also too immature to be used for ITC.
It is not possible to draw comparative conclusions from ITC using existing data because there is a significant risk of misleading decisions or limiting patients’ access to these treatments.
Appropriate statistical methods need to be used to obtain more data from ongoing or future actual studies to gain insight into the comparative effectiveness and safety of these two CAR T cell therapies.
Both axi-cel and tisa-cel have shown curative effects and controllable toxicity characteristics in separate single-arm clinical trials and real-world settings, providing patients with promising new treatment options for r/r DLBCL beyond conventional therapies. Since these two CAR T cell therapies are available almost at the same time, understanding their comparative effectiveness is of great interest to patients, clinicians, payers, and other stakeholders to help provide information for clinical decision-making and maximize patient benefits .
Randomized controlled trials (RCT) are the gold standard for evaluating comparative efficacy and safety; however, no head-to-head RCT has been performed on these two treatments. In the absence of direct evidence from RCT, data from real-world studies in independent clinical trials and/or indirect treatment comparisons (ITC) can be used to evaluate comparative efficacy.
However, if there are substantial differences between the studies, ITC should be conducted with extreme caution. For example, due to the heterogeneity of study design and patient populations, data from individual trials may introduce bias, while real data may be limited by the sensitivity of multiple sources of bias, such as lack of quality control and follow-up procedures surrounding data collection And selection bias (ie, the choice of CAR T cell therapy depends on the patient profile based on the assumed product attributes).
Although all non-random comparisons face limitations, it is important to assess whether the severity of the limitations overwhelms any value in explaining the results of the comparison. Comparative validity conclusions based on immature or unreliable data may mislead treatment options, limit patients’ access to effective treatment options, and undermine the allocation of resources in the health and medical care system.
This article summarizes the existing evidence from clinical trials and real-world studies, and discusses the challenges and limitations of potential analysis methods related to r/r DLBCL, and solves whether the effective ITC of axi-cel and tisa-cel may be used for r /r DLBCL problem.
In addition, this article provides forward-looking thinking on future comparative analysis approaches based on other sources of evidence that are not currently available. This article is based on previous research and does not contain any research conducted by any author on human participants or animals.
Evidence from clinical trials
tisa-cel and axi-cel have demonstrated long-lasting clinical benefits and controllable toxicity in their key clinical trials JULIET (ClinicalTrials.gov identifier: NCT02445248) and ZUMA-1 (NCT02348216), respectively. JULIET is a single-arm, global, multi-center phase II trial for tisa-cel, conducted at centers in North America, Europe, Australia, and Japan.
As of July 2019, 167 patients met the clinical inclusion criteria and received leukocyte removal; of these, 115 patients were injected with tisa-cel. ZUMA-1 is an open-label, single-arm, multi-center, phase I-II trial, mainly conducted in the United States (1 center in Israel). As of August 2018, 108 patients had received axi-cel infusion in two phases of the trial (7 in Phase I and 101 in Phase II).
In order to evaluate the feasibility, advantages, and limitations of ITC comparing JULIET and ZUMA-1, we evaluated the similarities and differences of trial design, inclusion process, result definition, and patient population (summarized in Table 1). When performing ITC, it is important to be able to adjust for cross-test differences that have known or suspected effects on patient outcomes.
These may include prognostic factors, which affect the outcome regardless of the type of treatment, or effect modifiers, which have different effects on the outcome of each treatment. It is often impossible to distinguish between prognostic factors and effect modifiers, especially for new therapies.
In addition, the differences between different trials may stem from the design, enrollment process, and result definition, which may or may not be suitable for statistical adjustment.
There are some similarities between JULIET and ZUMA-1. The main reason is that they both recruit patients with refractory DLBCL. They are both open-label single-arm trials. Eligible patients are required to have previously received chemotherapy and the Eastern Cooperative Oncology Group ( The performance status of ECOG) is 0 or 1 at the time of screening, and similar end points are evaluated (for example, overall response rate [ORR] as the primary end point). However, there are several important differences, which are summarized below.
Patient’s journey from screening to CAR T cell infusion
The consideration of the patient journey from screening to eligibility, leukopenia, enrollment, and treatment in JULIET and ZUMA-1 revealed some important differences in trial design (Figure 1). These differences stem from the screening and manufacturing processes required for CAR T cell therapy.
Since CAR T cells are designed to transduce the patient’s autologous T cells, it is necessary to collect the T cells first through leukopenia, and then genetically modify them to fight cancer before being infused into the patient.
The manufacturing process of CAR T cell therapy begins after the leukocyte removal procedure and the processing laboratory receive the cells. From this point of view, it may take 3-4 weeks to send the modified T cells back to the treatment center for infusion.
The two trials used different methods to allocate production capacity to eligible patients. In ZUMA-1, registration is only performed after confirming the manufacturing slot and starting leukocyte removal. In contrast, in JULIET, patient registration has nothing to do with the availability of manufacturing tanks; leukocyte ablation may occur before the confirmed production period, and some patients may wait for the production period after registration.
Bridging treatment after enrollment
The second substantial difference between the trials involves the use of bridging chemotherapy. The JULIET trial allows the use of bridging therapy before lymphatic depletion chemotherapy (LDC) to maintain a poor prognosis for patients while waiting for infusion, and most (90%) patients have received bridging chemotherapy, and bridging therapy is not allowed. According to the trial protocol in ZUMA -1 in. JULIET patients who receive bridging chemotherapy generally have worse outcomes than those who do not receive bridging chemotherapy (for example, 12-month progression-free survival [PFS] rates are 32% and 61%, respectively). While patients are waiting for CAR T cell infusion, bridging chemotherapy is considered important to provide disease control, and most patients who receive tisa-cel or axi-cel treatment in the real world have already received bridging chemotherapy.
Similarly, a real-world study of patients receiving axi-cel reported that compared with patients who did not receive bridging chemotherapy, patients who received bridging chemotherapy before infusion had significantly worse OS and PFS (OS: risk ratio [ HR]=3.34, p<0.01; PFS: HR=1.43, p=0.04). Considering this negative impact on patient outcomes, any comparative analysis of JULIET and ZUMA-1 will be confused by the very different use of their internal bridging chemotherapy (ie, 90% and 0% accept it).
Even if you try to adjust for differences in bridging treatment, the analysis is still unreliable for two reasons. First and foremost, the clinical significance of the lack of bridging therapy in JULIET and ZUMA-1 is different. Specifically, all patients in JULIET can receive bridging therapy, but some patients did not make a clinical decision after enrollment (presumably because they thought it was unnecessary or they could not tolerate it).
In contrast, the ZUMA-1 regimen does not allow bridging chemotherapy, and patients choose to participate in a trial that is not allowed or available, despite any changes in their clinical status after enrollment.
Second, even though the lack of bridging therapy across trials has the same meaning, the proportion of patients who did not receive JULIET treatment is so small that adjustments to this single factor will limit the JULIET population to 10% of its original size, excluding other important factors. Confounding factors are adjusted without the need to extrapolate beyond the observed data.
Investigation of potential analysis methods to adjust for cross-test differences between JULIET and ZUMA-1
In order to further explore and illustrate the limitations of effective ITC, we investigated comprehensive and commonly used analysis methods, trying to adjust the cross-trial heterogeneity between the two key trials. Since JULIET and ZUMA-1 are single-arm tests, anchor-based ITC methods are not feasible.
In this case, the population-adjusted ITC method is usually considered in a non-anchored environment. Without access to patient-level data in the two trials, traditional statistical methods such as propensity score matching cannot be used to adjust for heterogeneity between trials.
This article discusses two analysis methods, namely matched-adjusted indirect comparison (MAIC) and simulated treatment comparison (STC), because both methods are health technology (HTA) assessment agencies (for example, the National Institute of Health and Care Excellence [NICE ]). Specifically, both methods adjust for inter-trial differences in the distribution of variables that affect outcomes, using patient-level data for one treatment and aggregated trial-level data for the other treatment.
(1) The MAIC method applies weights based on propensity scores (estimated by propensity models) to each patient in a trial, and uses patient-level data to match average patient characteristics between trials. Then compare the weighted results among the balanced test population.
(2) The STC method uses patient-level data from a trial to fit a predictive model (ie regression model) of relevant patient characteristics to each result. Then, the model is used to predict the outcome of a population with the same baseline characteristics as other trials; the comparison between the predicted results and the observed results in this population is used as the estimated treatment effect.
The key assumptions of these two methods are that all confounding factors have been included in the adjustment, and the model has been correctly specified (ie, the propensity scoring model of MAIC and the predictive model of STC). This assumption is strong and may not be satisfied in some cases. Before explaining MAIC or STC, it is necessary to consider whether all important known or suspected confounding factors can be resolved by any of these statistical methods.
This review evaluated two MAICs and one STC (referred to as CAR-T predictive model analysis). The OS result was selected for these analyses because it was consistently defined throughout the trial (ie, the time from infusion to death of the patient from any cause), which is different from other response and safety results.
Indirect comparison of matching adjustments Two MAIC analyses were performed using data from JULIET and ZUMA-1 to indirectly compare tisa-cel and axi-cel. (1) An item conducted by the author, using patient-level data from JULIET and summary data from ZUMA-1 to re-weight JULIET patients to match the characteristics of ZUMA-1 patients. After matching, the estimated HR between tisa-cel and axi-cel is 1.90 (95% confidence interval [CI]: 1.28, 2.82). Figure 2 illustrates the OS curve of the adjusted JULIET data and the observed ZUMA-1 data.
Figure 2 Observation of tisa-cel OS, observation of axi-cel OS, and tisa-cel OS1-4 adjusted based on MAIC and CAR-T prediction models.
Another MAIC was performed by Oluwole et al. and used patient-level data from ZUMA-1 and aggregated data from JULIET to match ZUMA-1 patient characteristics with those of JULIET patients. When matched with JULIET patients, the estimated HR between tisa-cel and axi-cel was 2.27 (1.47, 3.45).
These two MAICs are not expected to produce the same results, even if both are valid and interpretable, because the two analyses compare the OS of different matched populations (ie, adjust JULIET patients to match the ZUMA-1 population, and vice versa ). In addition to matching different patient groups, different matching factors were also considered in these two analyses.
For example, the author’s MAIC matches more baseline characteristics (including gender, previous ASCT, and disease) than the MAIC of Oluwole et al. It is not clear how to exclude these factors as potential confounders. In contrast, Oluwole et al. matched the disease stage in MAIC. But not in the author’s article, because at the time of our analysis, there were no published ZUMA-1 data on disease stages.
Table 2 summarizes the detailed list of adjustment variables for the two MAICs.
In addition to the differences in matching baseline characteristics, the bigger problem is that both MAICs are subject to major limitations that cannot be resolved through statistical adjustments due to cross-test differences. As mentioned in the previous section, the time and enrollment of leukocyte ablation, the use of bridging chemotherapy, the LDC regimen, and the treatment regimen received after CAR T cell infusion have systemic differences between different trials and cannot be adjusted. Since these variables may be prognostic factors for the patient’s prognosis, the inability to adjust them will make the results of the two analyses unreliable.
In addition, given that there are multiple substantial and interrelated differences between trials, it is impossible to discern any possible direction or degree of bias.
CAR-T prediction model analysis
The authors also performed CAR-T predictive model analysis to infer the OS of tisa-cel in a hypothetical population similar to the ZUMA-1 trial in terms of patient characteristics, as shown in Table 2). Compared with MAIC, which cannot directly extrapolate beyond the observed data, the CAR-T prediction model method allows the results to be extrapolated to a population that is not well represented by the source population. In particular, this method is used to extrapolate JULIET results to people who have not received bridging therapy and who have received different LDC regimens. Although this extrapolation method brings many additional limitations, it is very helpful here as an exploratory and illustrative tool.
In this CAR-T predictive model analysis, a multivariate Cox model for OS was developed using patient-level data from the JULIET trial. This model is used to predict the efficacy of tisa-cel in a hypothetical patient population with a patient feature set similar to ZUMA-1 through extrapolation, including 0% receiving bridging chemotherapy and 100% receiving fludarabine-based LDC. In this analysis, tisa-cel is associated with a numerically longer OS than axi-cel (tisa-cel and axi-cel, HR [95% CI]: 0.75 [0.48, 1.18]), contrary to the two MAIC results As mentioned above. The adjusted tisa-cel OS curve assumes that tisa-cel has been used to treat the ZUMA-1 patient population. The observed ZUMA-1 OS curve is shown in Figure 2.
The CAR-T prediction model is extrapolated based on factors that may affect OS, especially bridging chemotherapy and LDC, which cannot be explained by the use of MAIC. However, by assuming that the average patient’s predicted outcome is equal to the average of the patient’s predicted outcome, it may cause additional bias. In a non-linear predictive model (for example, the Cox model of OS), it is unlikely to satisfy such assumptions. In addition, only 11 patients in JULIET did not receive bridging chemotherapy. A small number of events introduced additional bias, as these patients may not represent patients who require bridging chemotherapy.
The ITC investigated here is a commonly used analytical method, but it cannot guarantee that any method, regardless of its operational characteristics and general acceptability, can produce effective and reliable comparisons between non-random treatment groups. In the case of comparing JULIET and ZUMA-1, due to the huge differences in trial design and patient population, both methods have great limitations, and these differences cannot be explained by statistical adjustments. It is worth noting that different methods have led to conflicting conclusions, further demonstrating the high risk of drawing wrong conclusions through any comparison of very different experiments.
Real world evidence
In addition to data from clinical trials, real world evidence (RWE) is another potential source of data for comparative analysis. The clinical practice in the real world does not always follow the norms of clinical trials, which may reflect the evolution of clinical practice and a broader and more heterogeneous patient population in the real world.
Regulatory agencies such as the FDA have long used RWE to provide information for trial design, and to monitor safety and evaluate effectiveness after drug approval. RWE is playing an increasingly important role in treatment evaluation, especially in the context of oncology and rare diseases.
Since these therapies were approved, many real-world studies have been conducted on patients receiving tisa-cel or axi-cel in the United States and Europe. Riddle and others retrospectively analyzed data from patients from eight American academic centers who received commercial use of tisa-cel or axi-cel .
The data collection in this study started after the FDA approved tisa-cel. At that time, the center could choose to open tisa-cel or axi-cel. Of the 244 patients who received CAR T cell therapy, 158 received axi-cel treatment and 86 received tisa-cel treatment. More axi-cel patients were treated in an inpatient setting than tisa-cel patients (92% and 37%, respectively), and most of the two cohorts received bridging chemotherapy (61% and 75%, respectively).
The high rate of bridging chemotherapy may indicate that in real-world practice, most patients cannot wait for no bridging treatment in the manufacturing process before receiving an infusion.
The two cohorts described by Riedell et al. have different demographic and clinical characteristics. For example, tisa-cel recipients were older than axi-cel recipients (median age: 67 and 59 years, respectively) and received more pretreatment (86% vs. 73% received ≥ 3 treatments before) ).
Riddle et al. concluded that the efficacy in a commercial setting appears to be similar to the results observed in key clinical trials. On the 90th day after infusion, 64% of axi-cel patients achieved objective remission (OR), 53% achieved complete remission (CR), and 51% of tisa-cel patients achieved OR, of which 42% achieved CR. Due to the short follow-up time (median: 7.6 and 6.2 months for the axi-cel and tisa-cel cohorts, respectively), neither cohort reached the median OS.
When Riedell et al. examined AEs, they found that tisa-cel was associated with fewer CRS and neurological events than axi-cel, which affected the use of medical resources. For example, compared with tisa-cel patients, patients treated with axi-cel have longer hospital stays (median: 16 and 2 days, respectively), and the incidence of metastases in the intensive care unit has increased (39% vs. 7%) ) And more use tocilizumab (61% vs. 15%) and steroids (53% vs. 8%).
In addition, it has been observed that the management and classification of CRS in the real world (using the American Society for Transplantation and Cell Therapy system) is different from clinical trials. Early use of corticosteroids and prophylactic use of tocilizumab in the real world may reduce the recorded incidence of severe CRS in DLBCL patients receiving CAR T cell therapy.
Although it may be expected that deviations from strict clinical trial specifications may lead to poor results, the results reported by Riedell et al. for patients treated with tisa-cel and axi-cel in the real world are similar to those reported in key research clinical trials. However, without controlling for differences in patient populations, these values should not be directly compared, such as differences in bridging chemotherapy rates, which may affect the results.
The International Center for Blood and Bone Marrow Transplant Research (CIBMTR) registry is another source of RWE for tisa-cel and axi-cel, although the follow-up time is still very short at the time of reporting (tisa-cel is 4.5 months, axi-cel is 6.2 months).
The tisa-cel cohort included 116 patients (median age 65 years); 41% had double/triple hit lymphoma, and 27% had transforming lymphoma. The ORR of tisa-cel was 58%, and 40% of patients achieved CR. The OS and PFS rates at 3 months were 79.6% and 61.6%, respectively. Although the efficacy results of tisa-cel in the real world are similar to the efficacy data reported in the JULIET trial, the incidence of CRS (grade 3+ is 4%) and neurotoxicity (grade 3+ is 5%) is low.
The axi-cel cohort included 533 patients (median age 61 years); 36% had double/triple hit lymphoma, and 30% had transforming lymphoma. The ORR rate was 74%, of which 14% of patients reported CRS (grade 3+), and 61% of patients reported neurotoxicity (any grade).
In the real world, the efficacy and safety of axi-cel are similar to those of the ZUMA-1 trial. The additional data focuses on elderly patients (age ≥ 65 years), and the results are similar to those of younger patients (age <65 years). However, it is worth noting that there are differences in the baseline characteristics of patients treated with tisa-cel or axi-cel. For example, patients who received tisa-cel were older than those who received axi-cel (median age: 65 and 61 years, respectively). In addition, compared with patients who received axi-cel, patients who received tisa-cel had more double/triple gene hits (41% and 36%, respectively). The CIBMTR registration plan will follow up 1,500 patients treated for each therapy for 15 years, and will provide more evidence for the comparative analysis between tisa-cel and axi-cel when longer follow-up data are obtained. With the availability of long-term CIBMTR data, analysis methods can potentially be implemented, such as propensity score matching or addition using patient-level data to adjust for differences in patient characteristics.
In addition to the above studies, many other real-world studies have been conducted around the world to evaluate the effectiveness and safety of tisa-cel and axi-cel. These studies will help provide further information about treatment outcomes and clinical practice in real-world settings. At the time of writing, these studies are limited by short follow-up time (reported median follow-up time ranging from 4-7 months) and small sample size (e.g., <100 in most studies). Therefore, the existing evidence from the real-world studies of tisa-cel and axi-cel is still too immature for comparative analysis.
When considering all available evidence from clinical trials and real-world settings, the conclusion is that due to the huge differences in study design and patient populations, the ITC that evaluates the effectiveness and safety of tisa-cel and axi-cel is currently Cannot provide reliable results.
The key trials of tisa-cel (JULIET) and axi-cel (ZUMA-1) have fundamental and irreconcilable differences in trial design and patient populations. Real-world data shows that both CAR T cell therapies have produced lasting clinical benefits.
However, due to the short follow-up time, RWE is still too immature to effectively compare and analyze the two CAR-T cell therapies.
(source:internet, reference only)