Response to Methodological Concerns Regarding Eligibility Criteria and Study Selection in a Systematic Review and Meta‐Analysis of CAR T‐Cell Therapy for Hematologic Malignancies

We thank Dr. Xu et al. [1] for carefully reviewing our recent systematic review and meta-analysis evaluating the efficacy and safety of chimeric antigen receptor (CAR) T-cell therapy in hematological malignancies [2]. We appreciate the opportunity to address their methodological concerns regarding the inclusion of three specific clinical trials. We agree with the authors that strict adherence to eligibility criteria is paramount in meta-analyses. However, systematic reviews of complex and multi-dimensional oncology interventions frequently require nuanced clinical judgments to prevent the erroneous exclusion of valid, non-overlapping data. Below, we provide the clinical rationale for our study selection and present comprehensive sensitivity analyses demonstrating that the fundamental conclusions of our meta-analysis remain unaltered and robust. Regarding the inclusion of the study by Zhang et al. [3], 65% of the enrolled patients were aged 2–14 years. This trial including both adults and pediatric patients was retained as the authors of the study explicitly tested for age-related confounding, stating that: “Age was not a predictor for treatment outcome. There were no differences in the rates of CR, OS, or LFS for patients in groups with age 2 to 14 years or older than age 14 years” [3]. Because age did not modify the treatment effect in this trial, extracting the entire cohort enhanced the statistical power without introducing age-related bias. To definitively address the hypothesis that Zhang et al. may have inflated our pooled estimates, we conducted a sensitivity analysis excluding this study. Upon removal, the revised pooled objective response rate (ORR) for the leukemia cohort became 77% from 78%; the complete response (CR) rate shifted from 70% to 68%; and the 1-year overall survival (OS) rate remained 61%. This statistical continuity confirms that the inclusion of this mixed-age cohort did not skew the clinical conclusions (Figure 1A–C). Second, the authors of the letter argue that Gauthier et al. [4] should have been excluded as a “sub-analysis.” The methodological mandate to exclude sub-analyses exists primarily to prevent patient duplication and the double-counting of outcomes. While the individuals in Gauthier et al. may overlap with the parent trial, the therapeutic intervention and the outcome assessment periods are entirely distinct. Gauthier et al. strictly evaluated outcomes starting from the infusion of a second CAR T-cell dose (CART2). Because the efficacy and toxicity profiles were generated from a separate, subsequent therapeutic event, they do not duplicate the CART1 outcome data. Furthermore, the predefined exclusion of retrospective records was specifically intended to omit observational clinical studies or real-world registry studies conducted outside of a controlled clinical trial framework. Thus, excluding this study would have meant erroneously discarding non-overlapping clinical trial data regarding CAR-T re-treatment—a critical and emerging clinical scenario. To further demonstrate the ultimate robustness of our findings, we performed a sensitivity analysis excluding both Zhang et al. and Gauthier et al. Under these highly restrictive parameters, the revised pooled ORR actually became 80%, the CR rate 71% and the 1-year OS rate also rose to 63%. This confirms that our study selection did not artificially impact the reported benefits of CAR T-cell therapy in leukemia (Figure 2A–C). Another point pertains to the inclusion of the ZUMA-12 study [5]. Eligible patients were required to have a positive PET scan after completing two cycles of intensive anti-CD20 and anthracycline-containing chemoimmunotherapy. In a broader sense, such patients can be considered as “functionally primary refractory” or “high-risk”; thus, aligning with the mainly refractory population in our meta-analysis. Moreover, in the original ZUMA-12 publication it is stated that: “Forty-five per cent of patients in ZUMA-12 had either SD (5%) or PD (40%) at study entry, suggesting that a substantial proportion had primary chemorefractory disease, a group historically associated with median OS of approximately 6 months and long-term survival estimates near 10%.” For all these reasons, we opted to include this trial in our meta-analysis. We do, however, thank the authors for identifying a clerical error in our descriptive tables regarding this trial. In the original paper, the “Prior Lines of Therapy (LoT)” was reported as 2, where it should be 1. We do assure, though, that all meta-regression models implemented the correct number of 1 prior line of treatment. Nevertheless, in order to promote transparency and address the authors' concerns regarding trial inclusion, we present a sensitivity analysis excluding ZUMA-12 [5]. First, we examined the impact on the efficacy outcomes, which was negligible, as the ORR shifted from 76% to 75%, the CRR from 55% to 54%, the 1-year OS rate from 65% to 63% and the 1-year progression-free survival (PFS) rate from 46% to 45% (Figures S1–S4). Then, we re-evaluated our meta-regression models. In our original publication, we reported that an increase in prior lines of therapy was significantly associated with lower ORR (β = −0.5280, p = 0.002) and lower CR (β = −0.4048, p = 0.002). These observations were maintained in the sensitivity analysis (ORR; β = −0.5071, p = 0.006 and CRR; β = −0.3710, p = 0.004). Furthermore, the associations between 1-year OS (β = −0.4600, p = 0.004) and PFS rates (β = −0.4324, p = 0.006) and median number of prior lines of treatment were also retained after excluding the record (OS; β = −0.4201, p = 0.010 and PFS; β = −0.3861, p = 0.010). Mathematically, this attenuation of the negative slopes is the expected behavior when a high-leverage data point (e.g., ZUMA-12, ORR = 89%) is removed from the lower extreme of the predictor variable, thereby reducing the baseline intercept and requiring a less aggressive decrement to fit the subsequent, heavily pretreated cohorts (Table 1). Therefore, our clinical message remains intact: earlier utilization of CAR T-cell therapy correlates with superior patient outcomes. Taking into consideration the authors' concern about the inclusion of the NHL sub-cohort from Gauthier et al. [4], we conducted an additional sensitivity analysis of the efficacy outcomes for high-grade lymphoma, excluding both records. This sensitivity analysis resulted in pooled estimates of 76% for ORR, 55% for CRR, and 64% and 46% for the 1-year OS and PFS rates, respectively, perfectly resembling the original results reports (pooled, ORR = 76%; CRR = 55%; 1-year OS rate = 65%; 1-year PFS rate = 46%) (Figures S5–S8). Additionally, we wish to clarify that the inclusion of the contested trials does not represent a deviation from our protocol, but rather the systematic application of operational definitions during the full-text screening phase. Both the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA guidelines explicitly recognize the necessity of this approach [6, 7]. The Cochrane Handbook emphasizes that the implementation of the selection process is complex and inherently “involves judgment” [6]. In alignment with these methodological directives, our semantic criteria (e.g., “sub-analysis,” “retrospective,” or “pediatric”) were operationalized to ensure that clinically valuable, non-overlapping data were not erroneously omitted. We confirm that these operational definitions were applied uniformly to all 894 full-text articles assessed for eligibility. Consequently, we are confident that our search and screening completeness was rigorously maintained. Finally, we remain transparent about the challenges inherent to such comprehensive reviews; as explicitly noted in the limitations section of our original manuscript, the pooling of diverse clinical trials inevitably introduces between-study heterogeneity that must be considered when interpreting the generalizability of these findings. In conclusion, we are grateful to Dr. Xu and colleagues for their meticulous review and dedication to methodological rigor. We believe that the systematic operationalization of the eligibility criteria, combined with the clinical rationale and the validating sensitivity analyses provided herein, fully preserves the integrity, transparency, and reproducibility of our original meta-analysis. All authors drafted and approved the final version. The authors have nothing to report. This article is a correspondence regarding a published systematic review and meta-analysis. No original data from human or animal subjects were collected or generated for this manuscript. Consequently, ethical review and approval by an Institutional Review Board (IRB) or ethics committee were not applicable. The authors declare no conflicts of interest. Data sharing not applicable to this article as no datasets were generated or analysed during the current study. Figure S1: Sensitivity analysis, ORR in high-grade lymphoma Figure S2: Sensitivity analysis, CRR in high-grade lymphoma Figure S3: Sensitivity analysis, 1-year OS rate in high-grade lymphoma Figure S4: Sensitivity analysis, 1-year PFS rate in high-grade lymphoma Figure S5: Sensitivity analysis, ORR in high-grade lymphoma Figure S6: Sensitivity analysis, CRR in high-grade lymphoma Figure S7: Sensitivity analysis, 1-year OS rate in high-grade lymphoma Figure S8: Sensitivity analysis, 1-year PFS rate in high-grade lymphoma Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.