Pretreatment pulmonary function testing has limited utility in B-cell lymphoma treated with CD19 CAR T cells

Pulmonary function tests (PFTs) play an important role in hematopoietic cell transplantation (HCT), with routine evaluation in HCT candidates.¹ Forced expiratory volume in 1 second (FEV1), and single-breath diffusing capacity for carbon monoxide (DLCO) are correlated with morbidity and mortality after allogeneic and autologous HCT.^2-7 However, their predictive value in chimeric antigen receptor T-cell (CAR-T) therapy outcomes and toxicity remains uncertain. Nonetheless, PFT abnormalities were used as exclusion criteria in commercial CAR-T trials^8,9 and are common practice as a part of CAR-T pretreatment workup.

To address this gap, we conducted a retrospective analysis of PFT metrics and pulmonary comorbidity definitions, following the Hematopoietic Cell Transplantation–Comorbidity Index (HCT-CI),^10-12 in patients with B-cell lymphoma treated with autologous CD19 CAR-T therapy.

In this retrospective analysis, we included adult patients (aged ≥18 years) with B-cell lymphoma who received treatment at a tertiary medical center with commercially available (axicabtagene ciloleucel [axi-cel] and tisagenlecleucel [tisa-cel]) or point-of-care (POC) anti-CD19 CAR-T therapy¹³ as their third or later line of treatment, between 2017 and 2022.

Patients underwent routine PFT assessment before CAR-T therapy as part of institutional practice, including spirometry, body plethysmography, and DLCO. DLCO was corrected for hemoglobin level measured during hospitalization for CAR-T using the Dinakara equation.¹² Patients were categorized into 3 pretreatment pulmonary comorbidity groups (normal, moderate, and severe) based on HCT-CI definitions for FEV1 and DLCO, expressed as a percent of predicted values (>80%, 66%-80%, and <66%, respectively). Eighteen patients with no measured DLCO were included in pulmonary comorbidity groups by their FEV1 result only, according to HCT-CI definitions.^10,11 Smoking history was defined as a binary variable. DLCO and FEV1 as percent predicted were divided into deciles to better demonstrate their clinical significance.

The research was approved by Sheba Medical Center Institutional Review Committee.

Lymphoma response to CAR-T therapy was assessed using positron emission tomography–computed tomography and interpreted according to the Lugano criteria,¹⁴ relative to the assessment performed before leukapheresis. The overall response rate was defined as the proportion of patients who achieved a complete response (CR) or partial response (PR). One-month response was defined as the disease response on the 28th day after infusion or the earliest possible assessment if positron emission tomography–computed tomography could not be performed due to the patient’s clinical condition. Best day 100 response was defined as the disease best response until day 100 after infusion or the closest possible. Progression-free survival (PFS) was defined as the time from cell infusion to the date of either first documented progression or death due to any cause. Overall survival (OS) was defined as the time from cell infusion to the date of death from any cause. Duration of response was defined as the time from the first response assessment to disease progression, relapse, or death from any cause among patients who achieved a CR or PR.

Cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS) were graded according to the American Society for Transplantation and Cellular Therapy Consensus Grading.¹⁵ Adverse events were graded according to the Common Terminology Criteria for Adverse Events version 5.0.

Categorical variables were analyzed for statistical significance by χ² test or Fisher exact test. Continuous variables were described by interquartile range and summarized by median and range using Kruskal-Wallis rank. Patient characteristics were analyzed, and clinically significant variables were used to calculate the univariable analysis for PFS and OS. Variables meeting P value <.05 were introduced to a multivariable Cox regression for PFS and OS, together with patient, treatment, and disease. A Cox regression was formed to study the correlation of PFTs and response. A logistic regression was performed to study correlation among PFTs and response and toxicity. Sensitivity analysis models, restricted to patients receiving commercial CAR-Ts, were conducted to assess the potential impact of Sheba POC CD19 CAR-T on outcomes. A Spline regression model was used to assess the potential nonlinear correlation between variables. Spline thresholds were graphically demonstrated as the points at which the 95% confidence interval (95% CI) crossed 0. The Kaplan-Meier (KM) method was used for survival description, and the log-rank test was used for survival comparison. The median follow-up was calculated by the reverse KM method. All P values were 2-sided, and a P value <.05 was considered statistically significant. Data were analyzed using R (version 4.2.3).

A total of 192 patients (median age, 60 years [range, 46-69]) were included. Aggressive B-cell lymphoma was the predominant diagnosis (n = 166 [86%]), followed by indolent B-cell lymphoma (n = 18 [9.4%]), and mantle cell lymphoma (n = 8 [4.2%]). Most patients had a good performance status (Karnofsky Performance Scale [KPS] ≥ 90; 78%), although with high-risk disease features (prelymphodepletion lactate dehydrogenase [LDH] was elevated in 108 patients [56%]). Seventy-eight patients (41%) received bridging therapy. POC CAR-Ts were used in 56%, whereas 44% received commercial CAR-Ts (26% axi-cel and 18% tisa-cel).

Forty-five patients (25%) had a smoking history before CAR-T treatment. The median pretreatment measured-to-predicted DLCO and FEV1 were 96% and 92%, respectively. A total of 129 patients (74%) had normal DLCO, and 147 (77%) had normal FEV1. The calculated normal, moderate, and severe pulmonary comorbidity levels, per the HCT-CI, were seen in 123 (64%), 44 (23%), and 25 (13%) patients, respectively (supplemental Table 1).

Patients' toxicity profile is presented in supplemental Table 2. CRS of any grade occurred in 165 patients (86%), among them 32 (17%) had grade 3-5 CRS. One patient, with moderate pulmonary comorbidity, died due to CRS. CRS rates per normal, moderate, and severe pulmonary comorbidity levels were seen in 86% (n = 106), 84% (n = 37), and 88% (n = 22), respectively. There was no statistically significant relation between CRS grade >2 and pulmonary comorbidity level (P = .2). ICANS of any grade occurred in 64 patients (33%), and grade 3 to 4 ICANS was noted in 38 patients (20%). There was no statistically significant relation between ICANS grade >2 and pulmonary comorbidity level (P = .9). Two patients, with normal pulmonary comorbidity level, were admitted to the intensive care unit care because of treatment toxicity.

To determine whether pre–CAR-T pulmonary comorbidity increases the risk of treatment toxicity, we analyzed associations between pretreatment PFT results and the major CAR-T toxicities, CRS and ICANS.

Univariate logistic regression showed that categorized (normal, moderate, and severe) DLCO level, FEV1 level, and pulmonary comorbidity level were not associated with higher risk of high grade (>2) CRS (supplemental Table 3). DLCO as a continuous metric has correlated with CRS grade >2 (odds ratio [OR], 1.34; 95% CI, 1-1.81; P = .046). This correlation, in which the 95% CI encompasses the value of 1, did not persist in a multivariate model (OR, 1.12; 95% CI, 0.95-1.32; P = .2) adjusted for KPS score, LDH, and CAR-T costimulatory domain (supplemental Table 5).

Categorized (normal, moderate, and severe) DLCO level, FEV1 level, and pulmonary comorbidity level, as well as DLCO and FEV1 as continuous metrics, were not associated with elevated risk of high grade (>2) ICANS (supplemental Table 3).

Patient disease response was evaluated. Day 100 overall response rate was 67% (n = 129; CR, n = 100 [52%]; PR, n = 29 [15%]). Progressive disease (PD) or stable disease (SD) were reported in 60 patients (31%). Three patients did not achieve response evaluation (supplemental Table 2).

Univariate logistic regression revealed that abnormal categorizations (moderate and severe) of DLCO level, FEV1 level, or pulmonary comorbidity level were not associated with poor day 100 response (SD/PD) compared with the normal reference levels of DLCO, FEV1, or pulmonary comorbidity, respectively (supplemental Table 3). FEV1 as a continuous metric demonstrated correlation with poor day 100 response (OR, 0.68; 95% CI, 0.46-0.96; P = .037). This correlation did not persist in a multivariate model (OR, 0.85; 95% CI, 0.7-1.01; P = .062) adjusted for KPS score, LDH, and CAR-T costimulatory domain (supplemental Table 4). With a median follow-up of 17.6 months (interquartile range, 8.5-29.7), the median OS for the entire cohort was 23 months (95% CI, 16 to not reached), and the median PFS was 7.6 months (95% CI, 4.2-14).

In univariable Cox regression, pulmonary reserve and comorbidity metrics including pulmonary comorbidity level (P = .6), categorized DLCO (P = .4), categorized FEV1 (P = .3), DLCO as a continuous variable (P = .3), and FEV1 as a continuous variable (P = .079) were not associated with OS (supplemental Table 6). Spline-based models did not identify a nonlinear relationship between FEV1 and DLCO with OS (Figure 1A-B). Lack of association between pulmonary metrics and OS is also illustrated in OS KM estimates (Figure 2A-C).

Similarly, in univariable Cox regression, pulmonary comorbidity and metrics and smoking history lacked an association with PFS, the only exception being FEV1 as a continuous variable (P < .001; supplemental Table 6). These patterns are also reflected in PFS KM curves (Figure 2D-F). FEV1 maintained an inverse relationship with PFS (hazard ratio [HR], 0.87; 95% CI, 0.78-0.96; P = .007) in a multivariable model adjusted for KPS score, LDH, and CAR-T costimulatory domain (supplemental Table 7). Notably, the relationship between FEV1 and PFS was linear. A spline-based model did not identify a nonlinear relationship between DLCO and PFS (Figure 1C-D).

In a sensitivity analysis, FEV1 was associated (HR, 0.82; 95% CI, 0.71-0.96; P = .014) with PFS in a multivariable Cox regression model, restricted to patients receiving commercial CAR-Ts (axi-cel and tisa-cel). However, DLCO, when modeled as a continuous variable, did not correlate with CRS grade >2 in a multivariable model restricted to these patients. Additionally, FEV1 as a continuous variable was not associated with a poor day 100 response in the same patient group (supplemental Tables 8-10).

In this retrospective study, we address the clinical significance of pre–CAR-T pulmonary assessment in predicting treatment outcome and toxicity. To the best of our knowledge, this is the first study evaluating CAR-T outcomes and toxicity in relation to PFTs. We performed an analysis of PFTs done before anti-CD19 CAR-T administration in relapsed/refractory large B-cell lymphoma (LBCL), follicular lymphoma, and mantle cell lymphoma, divided into 3 levels per HCT-CI definitions,^10-12 and as a continuous metric. Univariate and multivariate regressions were performed, adjusted for KPS score, LDH, and CAR-T costimulatory domain. We observed no difference in day 100 response or in toxicity profile (CRS > 2; ICANS > 2) of patients with pretreatment normal, moderate, and severe PFT results. OS was not influenced by pulmonary comorbidity level either. A 10% increase in measured FEV1 was found to be associated with 13% increase in PFS (HR, 0.87; 95% CI, 0.78-0.96; P = .007).

Pretreatment PFT results, FEV1 and DLCO especially, are key features in identifying patients at high risk for developing pulmonary complications and/or mortality after HCT.^1-7 Scoring PFT via the HCT-CI scoring system was identified as beneficial in predicting HCT outcomes and toxicity over other modalities for patients with lymphoma and myeloma, with the likelihood ratio of nonrelapse survival of 23.7. The use of Dinakara equation for adjusting DLCO for hemoglobin is the most predictive for high-risk patients.^2,7,10-12 Patients with low pretreatment PFT results were excluded from certain CAR-T clinical trials,^8,9 whereas low pulse oximetry (<92%) and grade 1 dyspnea were used as exclusion criteria in others.^16,17 However, data reported in studies from real-world setting demonstrated consistent safety and efficacy outcomes, although many patients had comorbidities that would have made them otherwise ineligible for CAR-T clinical trials.^8,9,16-18 

Pulmonary function measurements are performed routinely as part of pre–CAR-T workup for patients with relapsed/refractory lymphomas. Based on our results, the predictive value of these measurements for survival outcomes and toxicities after infusion of CAR-Ts is very limited. Hence, excluding patients from CAR-T treatment due to their PFT anomalies is not recommended.

This study has several limitations. First, the limited cohort size underscores a need for further research in larger populations, and the retrospective design may introduce potential selection and measurement biases. Nonetheless, it is, to our knowledge, the first to address the knowledge gap regarding the clinical role of PFTs for patients undergoing CAR-T therapy, providing a valuable foundation for future studies. Second, our cohort included predominantly patients treated with POC CD19 CAR-Ts, which are not standard of care. Nonetheless, long-term PFS and OS for POC CD19 CAR-T align closely with those reported in the literature for commercial products.¹³ Moreover, results were consistent in multivariable models adjusted for CAR-T product, as well as sensitivity analysis restricted to commercial CAR-T products.

In the 2022 American Thoracic Society and European Respiratory Society guidelines, the use of z scores rather than percent of predicted PFTs was recommended. z scores are a robust metric for PFT evaluation, accounting for age, sex, height, and ancestral background. In this study, we used percent predicted values for FEV1 and DLCO to align with clinical practice, basing our definitions on the HCT-CI scoring system, which uses percent predicted values rather than z scores. This approach reflects the standard metrics used in routine clinical assessments, facilitating straightforward interpretation and clinical applicability. Although we recognize that z scores offer additional precision, our use of percent predicted values ensures consistency with current clinical tools and real-world application.

In conclusion, our study reveals that pretreatment PFTs have limited prognostic significance for patients with B-cell lymphoma treated with CD19 CAR-T therapy. Although FEV1 emerged as the sole PFT measure associated with PFS, we identified no association with OS or severe toxicities. Given the retrospective nature of this analysis and the small sample size, additional studies are required to validate our findings. As such, FEV1 might better associate with negative outcomes. If our findings are confirmed by others, pre–CAR-T evaluation might include spirometry only.

The reported research was supported in part by the National Institutes of Health/National Cancer Institute (NIH/NCI) Memorial Sloan Kettering Cancer Center Support Grant (P30 CA008748). R.S. reports grant support from NIH/NCI (grant K08-CA282987). A.A. discloses the following consultancy: Takeda, Gilead, Novartis, Roche, Bristol Myers Squibb, Abbvie, Eli Lilly.

Contribution: I.S. collected data, performed research, analyzed and interpreted data, performed statistical analysis, and wrote the manuscript; R.S. designed research, analyzed and interpreted data, performed statistical analysis, and wrote the manuscript; A.A. designed research, analyzed and interpreted data, and wrote the manuscript; M.J.S. designed research, analyzed and interpreted data, and wrote the manuscript; and S.F., R.M., I.D., R.Y., N.S.-T., O.I., E.J., M.K., A.N., and A.S. collected data and wrote the manuscript.

Conflict-of-interest: A.A. reports consultancy fees from Takeda, Gilead, Novartis, Roche, Bristol Myers Squibb, AbbVie, and Eli Lilly. The remaining authors declare no competing financial interests.

Correspondence: Abraham Avigdor, Hematology, Sheba Medical Center, Tel Hashomer, Sheba Road 1, Ramat Gan, 5265601, Israel; email: abraham.avigdor@sheba.gov.il.