Real-world outcomes for multiple myeloma patients after apheresis for planned chimeric antigen receptor T-cell therapy

Abstract

BACKGROUND

Chimeric antigen receptor T-cell (CAR-T) therapy is highly active in relapsed/refractory multiple myeloma (RRMM), but manufacturing creates an obligatory waiting period between apheresis and infusion. Patients may deteriorate, progress or die during this window. Analyses that begin follow-up at infusion exclude these patients entirely and cannot describe the experience of all patients who enter the CAR-T pathway.

AIM

To describe, in an intention-to-collect fashion, the outcomes of all RRMM patients who underwent T-cell apheresis for planned CAR-T therapy; to characterise the apheresis-to-infusion interval and the reasons for, and correlates of, failure to reach infusion; and to report progression-free survival (PFS) and overall survival (OS) anchored at the date of apheresis. The study was descriptive; it was not designed to estimate the causal effect of CAR-T infusion on survival.

METHODS

Retrospective single-centre cohort of all RRMM patients undergoing T-cell apheresis for CAR-T at Hackensack University Medical Center between 1 January 2021 and 30 April 2024. PFS and OS were calculated from apheresis by the Kaplan-Meier method with Greenwood 95% confidence intervals (CIs) and numbers at risk. Groups were compared with the log-rank test; baseline characteristics with the Wilcoxon rank-sum and Fisher exact tests. Because infusion is a post-baseline, time-dependent event, the association between infusion and survival was additionally examined using a Mantel-Byar time-dependent Cox model and a 60-day landmark analysis. Correlates of non-infusion were examined by univariable logistic regression; multivariable modelling was not performed owing to the small number of events.

RESULTS

Of 99 patients undergoing apheresis, 87 (87.9%, 95%CI: 79.8-93.6) were infused and 12 (12.1%, 95%CI: 6.4-20.2) were not. Median follow-up was 14.0 months [interquartile range (IQR): 8.2-20.7]. Median apheresis-to-infusion interval was 54 days (IQR 46-62; range 37-202; n = 84 evaluable). Contrary to our preliminary report, baseline disease phenotype differed significantly between groups: Extramedullary disease was present in 9/12 (75.0%) non-infused vs 17/87 (19.5%) infused patients (P < 0.001), and plasma cell leukaemia in 3/12 (25.0%) vs 3/87 (3.4%) (P = 0.017). Age, prior lines of therapy (median 5 in both), high-risk cytogenetics (36.4% vs 34.1%) and Revised International Staging System distribution did not differ. The commonest reasons for non-infusion were progressive disease or myeloma-related death (7/12, 58.3%), non-myeloma death (2/12, 16.7%), manufacturing or collection failure (2/12, 16.7%) and infection (1/12, 8.3%). The median time from apheresis to the attrition event was 44 days (IQR 30-72). Median PFS from apheresis was 17.5 months (95%CI: 12.7-21.6) in infused patients vs 1.5 months (95%CI: 0.8-2.3) in non-infused patients; median OS was not reached vs 1.9 months (95%CI: 1.0-6.2) (both log-rank P < 0.001). In the time-dependent Cox model the hazard ratio for death associated with infusion was 0.24 (95%CI: 0.11-0.53), substantially closer to the null than the unadjusted comparison, illustrating how much of the apparent difference reflects the time-dependent nature of the exposure and confounding by disease phenotype.

CONCLUSION

In this descriptive cohort, roughly one in eight patients who underwent apheresis never received CAR-T, and attrition clustered before the median time to infusion and among patients with extramedullary disease or plasma cell leukaemia. These data support shortening manufacturing turnaround, prioritising patients with aggressive phenotypes for expedited slots and intensified bridging, and reporting outcomes from apheresis using analytic methods that respect the time-dependent nature of infusion.

Key Words: Multiple myeloma; Chimeric antigen receptor T-cell therapy; Apheresis; Real-world data; Bridging therapy; Attrition; Immortal-time bias; Time-dependent analysis; B-cell maturation antigen

Core Tip: Among 99 myeloma patients undergoing apheresis for planned chimeric antigen receptor T-cell therapy, 12% progressed or died before manufacturing completed (median 54 days). This attrition disproportionately affected patients with aggressive phenotypes, such as extramedullary disease or plasma cell leukaemia. Because raw survival differences overstate treatment effects due to survival bias during manufacturing, time-dependent analyses are essential. Clinically, these findings highlight the urgent need for shorter manufacturing times, intensified bridging therapy and expedited scheduling for high-risk patients, and standardizing outcome reporting from the date of apheresis.

INTRODUCTION

Chimeric antigen receptor T-cell (CAR-T) therapy has emerged as a transformative therapy in the management of relapsed or refractory multiple myeloma (RRMM). The pivotal phase 2 KarMMa study of idecabtagene vicleucel (ide-cel) demonstrated meaningful response rates and durability in heavily pre-treated RRMM, leading to its United States Food and Drug Administration approval in March 2021[1]. Subsequently, ciltacabtagene autoleucel (cilta-cel) in the phase 1b/2 CARTITUDE-1 trial achieved deep and durable responses, further reinforcing the role of B-cell maturation antigen (BCMA)-directed CAR-T therapy[2]. Both agents target BCMA, which is highly expressed on malignant plasma cells and therefore serves as a validated therapeutic target in RRMM. In addition to providing deeper and more durable remissions for penta-refractory patients who had a historically dismal prognosis[1,2], these therapies offer a potential treatment-free interval, shifting paradigms in myeloma care on several fronts.

Despite these advances, several important limitations remain. First, populations enrolled in pivotal trials are highly selected and may not reflect real-world patients, many of whom present with aggressive disease biology or multiple comorbidities. The International Myeloma Working Group (IMWG) consensus statement highlights the need for real-world data to complement trial findings in CAR-T therapy for myeloma[3], and real-world experience with ide-cel has shown different safety and efficacy profiles compared with the trial setting[4]. Second, the time from apheresis to infusion remains a critical operational bottleneck, and the lag between collection and infusion can result in clinical deterioration or progression in a subset of patients. Real-world series report apheresis-to-infusion rates in the region of 85%-95%, with attrition attributed predominantly to progression or death before product availability[4]. A real-world analysis by Al Hadidi et al[5] found that among patients wait-listed for commercial CAR-T, the median time to undergo apheresis was 3.7 months and the 12-month cumulative incidence of death while on the waitlist was 26%. These observations underscore the need to understand outcomes from the time of apheresis, not only from infusion.

Emerging data suggest that the apheresis and bridging period itself carries prognostic implications: Pre-apheresis haematopoietic reserve, inflammatory markers, T-cell subsets in the apheresis product and bridging therapy have all been associated with post-infusion progression-free survival (PFS) and overall survival (OS)[6,7]. However, many published analyses exclude patients who underwent apheresis but never received an infusion, and thereby describe a survivor-enriched population. This is a classic manifestation of immortal-time bias: Because a patient cannot be classified as “infused” without first surviving the manufacturing interval, any analysis that conditions on infusion status guarantees a period of event-free time to the infused group[8]. Moving the index date to apheresis removes the mis-allocation of that person-time from the analysis clock, but by itself it does not solve the problem, because group membership is still defined by a post-baseline event. Doing so requires that infusion be modelled as a time-dependent exposure, or that a landmark be imposed[9,10]. To our knowledge, no real-world myeloma CAR-T series has combined apheresis anchoring with time-dependent exposure modelling, and we adopt both approaches here - while emphasising that neither can overcome confounding by indication in an observational cohort of this size.

In light of these issues, we performed a descriptive, intention-to-collect analysis of all patients with multiple myeloma who underwent apheresis for planned CAR-T therapy at our centre, regardless of whether they ultimately received an infusion. Our objectives were: (1) To quantify the proportion of patients who did not reach infusion and to characterise the reasons; (2) To describe the distribution of the apheresis-to-infusion interval; (3) To describe PFS and OS anchored at apheresis, with numbers at risk and confidence intervals (CIs); and (4) To explore, without causal claim, whether baseline characteristics were associated with failure to reach infusion. We did not set out to estimate the efficacy of CAR-T therapy, and the comparisons presented between infused and non-infused patients should not be read as such.

Finally, this work has significant implications for patient counselling, therapeutic planning, bridging strategies, and manufacturing workflows. Understanding the “drop-off” from apheresis to infusion, the time-dependency of progression risk during manufacturing, and the real-world success of identifying, collecting, and infusing eligible patients will inform future trial designs, institutional protocols, and health-system optimization. By using an inclusive chart-review approach covering a defined time-period (January 2021 to April 2024) at a high-volume academic center, this study adds to the growing evidence base of real-world cellular therapy in myeloma and highlights the full pathway from collection to infusion and beyond.

MATERIALS AND METHODS

Study design and setting

This was a single-centre, retrospective, descriptive intention-to-collect cohort study conducted at the John Theurer Cancer Center, Hackensack University Medical Center, Hackensack, NJ. The study evaluated real-world outcomes of patients with multiple myeloma who underwent apheresis for CAR T-cell therapy between 1 January 2021 and 30 April 2024. The administrative data cutoff was 12 July 2024. The protocol (Pro2024-0214, Version 1.0, 1 May 2024) was reviewed and approved by the Institutional Review Board. A waiver of informed consent and Health Insurance Portability and Accountability Act authorisation was granted.

Study population

The study population comprised adult patients with multiple myeloma who underwent apheresis for CAR T-cell manufacturing, regardless of whether they ultimately received the infusion. Patients were identified through the institutional electronic medical record (Epic Systems). Inclusion criteria were a confirmed diagnosis of multiple myeloma, apheresis for anti-BCMA or other investigational CAR-T therapy within the defined dates, and availability of records documenting apheresis, treatment and follow-up. Patients were excluded if their records lacked sufficient data to determine key outcomes or if apheresis was performed for a non-myeloma indication. No patient was excluded on these grounds during the study period; all 99 consecutive patients identified were analysed.

Data sources and collection procedures

Eligible patients were identified through the John Theurer Cancer Center CAR-T registry and verified in Epic. Data were collected by trained investigators who had completed institutional research compliance and CITI training. Each chart was reviewed independently, using a standardised abstraction form. Disease response was assessed according to IMWG criteria. Discrepancies were resolved by consensus review of the primary record. Following peer review, a complete re-audit of all baseline variables against the primary record was performed by two investigators independently; discrepancies identified in this re-audit are reported in the Results and in the Limitations.

Efficacy outcomes were analyzed according to the IMWG criteria. Data abstraction followed a standardized data collection form to ensure uniformity and minimize variability between reviewers.

Variables and data elements

Extracted data included demographics; disease characteristics [Durie-Salmon stage, Revised International Staging System (R-ISS) stage, monoclonal protein isotype, cytogenetics and interphase fluorescence in situ hybridisation (FISH)]; presence of extramedullary disease (EMD) and plasma cell leukaemia (PCL); treatment history (induction regimen, response, prior lines, prior autologous or allogeneic transplantation); date of apheresis; bridging therapy; date of infusion; CAR-T product; and outcomes (date of progression, date of death, overall response rate, PFS, OS).

High-risk cytogenetics were defined a priori as the presence of any of del(17p), t(4;14), t(14;16), t(14;20), or gain/amplification of 1q21 on interphase FISH performed on CD138-selected plasma cells, consistent with IMWG and Mayo Stratification for Myeloma and Risk-Adapted Therapy criteria. EMD was defined as soft-tissue plasmacytoma not contiguous with bone, documented on cross-sectional imaging or biopsy at or within 3 months before apheresis. PCL was defined per the contemporaneous institutional standard (≥ 5% circulating plasma cells or ≥ 2 × 10⁹/L). Bridging therapy was defined as any anti-myeloma systemic therapy or radiotherapy administered after apheresis and before lymphodepletion (or, in non-infused patients, after apheresis and before the attrition event).

Data management and confidentiality

All data were entered into a secure, password-protected database stored on institutional, HIPAA-compliant servers. Each subject was assigned a unique study identification code, which replaced identifiable information in the analytic dataset. The key linking study codes to patient identifiers was stored in a separate encrypted file accessible only to authorized study personnel. All identifying information will be destroyed once data extraction is complete. Study records will be retained for five years after study completion.

Statistical analysis

Analyses were performed in R version 4.3.2 (survival, survminer and cmprsk packages). Descriptive statistics summarised baseline characteristics. Categorical variables were expressed as frequencies and percentages with explicit denominators, and continuous variables as medians with interquartile ranges (IQRs) and ranges. Missing data were reported as counts and were not imputed; denominators for each variable exclude missing observations.

Between-group comparisons used the Wilcoxon rank-sum test for continuous variables and the Fisher exact test for categorical variables (the Fisher exact test replaced the χ² test used in our preliminary report, because multiple expected cell counts were < 5).

PFS and OS were estimated by the Kaplan-Meier method with the date of apheresis as the index date for all patients in both groups. OS was measured from apheresis to death from any cause. PFS was measured from apheresis to the earlier of IMWG-defined progression or death from any cause. Patients alive and progression-free were censored at the date of last documented contact or at the administrative cutoff of 12 July 2024, whichever came first. One patient who declined further care and was lost to follow-up was censored at the date of last contact (1.7 months). Median survival times are reported with Brookmeyer-Crowley 95%CIs, and landmark survival probabilities with Greenwood 95%CIs; where the survival function equalled exactly 0.50 over an interval, the median is reported as the midpoint of that interval, and this convention is stated in the figure legends. Numbers at risk are tabulated at 0 months, 3 months, 6 months, 12 months, 18 months, 24 months and 30 months. Median follow-up was estimated by the reverse Kaplan-Meier method.

Handling of the time-dependent nature of infusion. Receipt of a CAR-T infusion is not a baseline attribute: It can only be observed in patients who survive the manufacturing interval. Comparisons of “infused” vs “non-infused” groups from the date of apheresis are therefore susceptible to immortal-time bias even when the index date is apheresis[8]. Kaplan-Meier curves stratified by eventual infusion status are presented for descriptive purposes only. To address the time-dependent exposure, two additional analyses were pre-specified at revision: (1) A Cox model in which infusion was entered as a time-varying covariate, with each infused patient contributing person-time to the “not yet infused” state from apheresis until the infusion date and to the “infused” state thereafter (Mantel-Byar approach[9]); and (2) A 60-day landmark analysis restricted to patients alive 60 days after apheresis, chosen because 60 days approximated the upper quartile of the observed manufacturing interval[10]. Both are reported as exploratory and are not adjusted for confounding by indication.

Correlates of failure to reach infusion were examined by univariable logistic regression, with odds ratios (ORs) and 95%CIs (Wald, with exact P values from the Fisher exact test for sparse tables). A multivariable model was pre-planned but was not fitted, because only 12 non-infusion events occurred and the events-per-variable ratio would have been well below accepted thresholds. All tests were two-sided; P < 0.05 was considered significant. No adjustment was made for multiple comparisons, and all p values other than those for the two pre-specified primary survival comparisons should be regarded as descriptive.

Ethical considerations

This study involves retrospective review of existing medical records and poses no direct risk to participants. The principal risk is potential loss of confidentiality; however, all data handling and storage procedures comply with HIPAA and institutional data security standards. No physical, psychological, or clinical interventions were performed as part of this study. The waiver of consent does not adversely affect participants’ rights or welfare, as the analysis is conducted on de-identified data.

RESULTS

A total of 99 patients with RRMM underwent apheresis for planned CAR T-cell therapy between January 2021 and April 2024 (Table 1). Eighty-seven (87.9%, 95%CI: 79.8-93.6) received a CAR-T infusion and 12 (12.1%, 95%CI: 6.4-20.2) did not. At the 12 July 2024 cutoff, median follow-up was 14.0 months (IQR 8.2-20.7) overall 14.6 months (IQR 9.3-20.9) among infused patients and 1.7 months (IQR 1.1-5.8) among non-infused patients. Among 84 infused patients with internally consistent dates, the median interval from apheresis to infusion was 54 days (IQR 46-62; range 37-202); the cumulative proportion infused was 59% by day 60 and 82% by day 90.

Table 1 Baseline characteristics of patients with relapsed or refractory multiple myeloma who underwent apheresis for planned chimeric antigen receptor T-cell infusion, n (%).

Characteristic	CAR-T infused (n = 87)	Not infused (n = 12)	P value	Test
Age at apheresis, year, median (range)	67.7 (36.7-81.7)	70.2 (42.6-83.7)	0.415	Wilcoxon
Sex	Not evaluable (missing 87/87)	Not evaluable (missing 12/12)	-	-
Plasma cell leukaemia	3/87 (3.4)	3/12 (25.0)	0.017	Fisher
Extramedullary disease	17/87 (19.5)	9/12 (75.0)	< 0.001	Fisher
High-risk cytogenetics1	29/85 (34.1)	4/11 (36.4)	0.88	Fisher
Missing	2	1	-	-
R-ISS stage	-	-	0.19	Fisher
I	11/66 (16.7)	1/11 (9.1)	-	-
II	52/66 (78.8)	8/11 (72.7)	-	-
III	3/66 (4.5)	2/11 (18.2)	-	-
Not evaluable	21	1	-	-
Prior lines of therapy, median (range)	5 (2-19)	5 (4-8)	0.431	Wilcoxon
Prior autologous or allogeneic transplant	81/87 (93.1)	10/12 (83.3)	0.24	Fisher
Bridging therapy after apheresis	33/87 (37.9)	5/12 (41.7)	0.80	Fisher
Apheresis-to-infusion interval, days, median (IQR; range)	54 (46-62; 37-202), n = 842	Not applicable	-	-
Lymphodepleting regimen
Fludarabine/cyclophosphamide	43/87 (49.4)	-	-	-
Bendamustine	44/87 (50.6)	-	-	-
Intended/administered CAR-T product	-	-	0.55	Fisher
Ciltacabtagene autoleucel	45/87 (51.7)	6/12 (50.0)	-	-
Idecabtagene vicleucel	42/87 (48.3)	6/12 (50.0)	-	-

¹High-risk cytogenetics.

²Apheresis-to-infusion interval data were not evaluable/missing for 3 of the 87 infused patients.

CAR-T: Chimeric antigen receptor T-cell; R-ISS: Revised International Staging System; IQR: Interquartile range.

Baseline characteristics

The median age at apheresis was 67.7 years (range 36.7-81.7) in the infused group and 70.2 years (range 42.6-83.7) in the non-infused group (P = 0.415). However, EMD was markedly more frequent among patients who did not reach infusion (9/12, 75.0% vs 17/87, 19.5%; P < 0.001), as was PCL (3/12, 25.0% vs 3/87, 3.4%; P = 0.017). There were no significant differences in high-risk cytogenetics (4/11, 36.4% vs 29/85, 34.1%; P = 0.88; missing in 1 and 2 patients respectively), median prior lines of therapy (5 in both groups; P = 0.431), R-ISS distribution among evaluable patients (P = 0.19; not evaluable in 1/12 and 21/87), prior autologous or allogeneic transplantation (10/12, 83.3% vs 81/87, 93.1%; P = 0.24), or intended CAR-T product (cilta-cel 6/12 vs 45/87; P = 0.55).

Bridging therapy, lymphodepletion and product

Bridging therapy was administered to 33/87 (37.9%) infused and 5/12 (41.7%) non-infused patients (P = 0.80). Bridging regimens included corticosteroid-based combinations, proteasome-inhibitor- and immunomodulatory-drug-based regimens, alkylator-based regimens (bendamustine, cyclophosphamide, mini-carmustine, etoposide, cytarabine, and melphalan), BCMA- or GPRC5D-directed bispecific antibodies, belantamab mafodotin and localised radiotherapy (Table 2 for non-infused patients). All infused patients completed lymphodepleting chemotherapy, with an even split between fludarabine/cyclophosphamide (43/87, 49.4%) and bendamustine (44/87, 50.6%). Products were nearly equally distributed between cilta-cel (45/87, 51.7%) and ide-cel (42/87, 48.3%).

Table 2 Patient-level characteristics and outcomes of the 12 patients who underwent apheresis but were not infused.

ID	Age band (year)	EMD	PCL	High-risk cyto	R-ISS	Prior lines	Intended product	Bridging regimen (category)	Days apheresis → attrition event	Reason not infused	Subsequent therapy	Outcome (months from apheresis)
N-01	60-64	Yes	No	No	II	7	Ide-cel	None	176	Infection (bacteraemia, prolonged admission)	None	Died, 5.8
N-02	80-84	No	No	No	II	4	Ide-cel	None	11	Non-myeloma death (COVID-19, pulmonary embolism)	None	Died, 1.0
N-03	75-79	Yes	No	No	I	6	Ide-cel	Antibody-drug conjugate	70	Non-myeloma death (ischaemic stroke)	None	Died, 2.3
N-04	60-64	Yes	Yes	Yes	II	4	Cilta-cel	PI + BCL2 inhibitor + steroid	34	Manufacturing failure (out-of-specification product)	BCMA bispecific	Died, 6.2
N-05	55-59	Yes	No	No	II	5	Cilta-cel	None	24	Progressive disease + intracranial haemorrhage	None	Died, 0.8
N-06	75-79	Yes	No	Yes	II	4	Ide-cel	Radiotherapy	71	Progressive disease → bowel ischaemia	None	Died, 2.3
N-07	40-44	No	Yes	Yes	III	5	Cilta-cel	None	51	Progressive disease; declined further therapy	None	Lost to follow-up; censored 1.7
N-08	80-84	Yes	No	No	II	5	Cilta-cel	Alkylator + PI + steroid	47	Progressive disease	None	Died, 1.5
N-09	65-69	Yes	No	Yes	II	5	Ide-cel	BCMA bispecific	33	Progressive disease	None	Died, 1.1
N-10	60-64	Yes	Yes	No	III	8	Cilta-cel	None	5	Progressive disease	None	Died, 0.2
N-11	80-84	Yes	No	Unknown	Unknown	7	Ide-cel	None	196	Collection failure (inadequate CD3+ yield)	GPRC5D bispecific	Alive, 11.7
N-12	70-74	No	No	No	II	5	Cilta-cel	None	42	Progressive disease	None	Died, 1.4

Median time from apheresis to the attrition event: 44 days (interquartile range: 30-72; range 5-196). Eight of 12 events (66.7%) occurred within 60 days. Ages are banded and no dates are reported, to protect confidentiality. BCMA: B-cell maturation antigen; EMD: Extramedullary disease; GPRC5D: G protein-coupled receptor class C group 5 member D; PCL: Plasma cell leukaemia; PI: Proteasome inhibitor; R-ISS: Revised International Staging System; COVID-19: Coronavirus disease 2019; BCL2: B-cell lymphoma 2.

Reasons for, and timing of, non-infusion

Among the 12 patients who did not receive an infusion, 7 (58.3%) did not proceed because of rapid disease progression or myeloma-related death before product delivery; the remainder were attributed to non-myeloma death (2/12, 16.7%: Fatal pulmonary embolism in the context of coronavirus disease 2019, and ischaemic stroke), manufacturing or collection failure (2/12, 16.7%: One out-of-specification product and one inadequate cell yield) and infection (1/12, 8.3%: Bacteraemia with prolonged hospitalisation). The median time from apheresis to the attrition event was 44 days (IQR 30-72; range 5-196), and 8 of 12 attrition events (66.7%) occurred within 60 days of apheresis that is, at or before the cohort's median time to infusion. Patient-level detail is provided in Table 2. Notably, 5 of the 12 (41.7%) received bridging therapy, and in 4 of these 5 the disease progressed during or immediately after bridging.

Survival outcomes anchored at apheresis

These analyses are descriptive; see the following section for time-dependent analyses. At last follow-up, documented disease progression had occurred in 36/87 (41.4%) infused patients and a further 6 died without documented progression, giving 42 PFS events (48.3%); all 12 non-infused patients (100%) experienced a PFS event (P < 0.001). Death had occurred in 15/87 (17.2%, 95%CI: 9.9-27.2) infused and 10/12 (83.3%, 95%CI: 51.6-97.9) non-infused patients (P < 0.001).

Median PFS from apheresis was 17.5 months (95%CI: 12.7-21.6) among infused patients and 1.5 months (95%CI: 0.8-2.3) among non-infused patients (log-rank P < 0.001; Figure 1). Estimated PFS at 6 and 12 months was 81.0% (95%CI: 71.2-87.6) and 67.9% (95%CI: 56.9-76.8) in the infused group, vs 8.3% (95%CI: 0.5-31.1) and 0% in the non-infused group. Median OS was not reached in the infused group (95%CI: Not estimable) and was 1.9 months (95%CI: 1.0-6.2) in the non-infused group (log-rank P < 0.001; Figure 2). Estimated OS at 12 months and 24 months was 88.6% (95%CI: 79.6-93.8) and 75.1% (95%CI: 62.6-84.0) in the infused group, vs 10.0% (95%CI: 0.6-35.8) and 0% in the non-infused group. Numbers at risk at each timepoint are shown beneath Figures 1 and 2. Figure 3 shows the patient-flow diagram.

Figure 1 Kaplan-Meier estimates of progression-free survival from the date of apheresis in patients with relapsed/refractory multiple myeloma who underwent T-cell collection for planned B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy (n = 99). Line definitions: The blue line denotes patients who proceeded to chimeric antigen receptor T-cell infusion (infused cohort, n = 87). The purple line denotes patients who underwent apheresis but never received an infusion (non-infused cohort, n = 12). Progression-free survival was measured from the date of apheresis (day 0, index date) to the date of first documented progression by International Myeloma Working Group criteria or death from any cause, whichever occurred first; patients alive and progression-free were censored at the date of last documented disease assessment (data cut-off 12 July 2024). Vertical tick marks on each curve indicate censored observations; the shaded bands, where shown, represent 95%CIs; the table beneath the X-axis reports the number of patients at risk in each group at each 3-month interval. Median progression-free survival was 17.5 months in the infused cohort vs 1.5 months in the non-infused cohort. The X-axis shows time in months from apheresis; the Y-axis shows the estimated probability of remaining alive and progression-free. PFS: Progression-free survival.

Figure 2 Kaplan-Meier estimates of overall survival from the date of apheresis in patients with relapsed/refractory multiple myeloma who underwent T-cell collection for planned B-cell maturation antigen-directed chimeric antigen receptor T-cell therapy (n = 99). Line definitions: The blue line denotes patients who proceeded to chimeric antigen receptor T-cell infusion (infused cohort, n = 87). The purple line denotes patients who underwent apheresis but never received an infusion (non-infused cohort, n = 12). Overall survival was measured from the date of apheresis (day 0, index date) to death from any cause; patients alive at last contact were censored at the date of last follow-up (data cut-off 12 July 2024). Vertical tick marks indicate censored observations; the shaded bands, where shown, represent 95% confidence intervals; the table beneath the x-axis reports the number of patients at risk in each group at each 3-month interval. Median overall survival was not reached in the infused cohort and 1.9 months in the non-infused cohort. The X-axis shows time in months from apheresis and should be extended to the maximum observed follow-up; the Y-axis shows the estimated probability of overall survival. OS: Overall survival.

Figure 3 Patient-flow diagram (Consolidated Standards of Reporting Trials-style) for the intention-to-collect cohort. R/R: Relapsed or refractory; MM: Multiple myeloma; BCMA: B-cell maturation antigen; CAR-T: Chimeric antigen receptor T-cell; IQR: Interquartile range; PFS: Progression-free survival; PE: Pulmonary embolism; COVID-19: Coronavirus disease 2019.

Time-dependent and landmark analyses

In the Mantel-Byar time-dependent Cox model, in which all patients contributed person-time to the pre-infusion state from apheresis until infusion, the hazard ratio for death associated with being in the infused state was 0.24 (95%CI: 0.11-0.53; P = 0.001); the corresponding estimate for the composite PFS endpoint was 0.53 (95%CI: 0.28-1.00; P = 0.05). In the 60-day landmark analysis (92 patients alive at day 60: 87 subsequently or already infused, 5 not infused), median OS measured from the landmark was not reached vs 3.8 months, with an unadjusted hazard ratio (HR) of 0.06 (95%CI: 0.02-0.18).

These estimates should be interpreted with caution and in the following order. The unadjusted Kaplan-Meier comparison is the most extreme and the least trustworthy, because classification as “infused” requires survival through manufacturing. The time-dependent estimate is markedly closer to the null (HR 0.24 vs an unadjusted hazard ratio exceeding 15), and quantifies how much of the apparent survival gap is attributable to the structure of the exposure rather than to the intervention. Even the time-dependent estimate remains confounded: Patients who reached infusion had, by definition, disease that could be controlled for ~54 days, and they had a substantially lower prevalence of EMD and PCL at baseline. For the PFS endpoint, progression occurring during the manufacturing window in patients who were nonetheless subsequently infused was probably under-ascertained in the medical record, which will bias the time-dependent PFS estimate toward benefit. None of these analyses supports a causal estimate of CAR-T efficacy, and we do not offer one.

Correlates of failure to reach infusion

In univariable logistic regression (Table 3), EMD (OR = 12.4, 95%CI: 3.0-50.6; P < 0.001) and PCL (OR = 9.3, 95%CI: 1.6-53.4; P = 0.012) were associated with failure to reach infusion. Age (OR per 10 years 1.24, 95%CI: 0.71-2.16; P = 0.45), high-risk cytogenetics (OR = 1.10, 95%CI: 0.30-4.05; P = 0.88), R-ISS stage III (OR = 4.7, 95%CI: 0.7-31.9; P = 0.13) and the number of prior lines (medians identical; P = 0.431) were not. CIs are wide, and these estimates are unstable and unadjusted for multiplicity; a multivariable model was not fitted because only 12 events occurred. In a sensitivity analysis excluding the two patients whose attrition was attributable purely to manufacturing or collection failure, the association with EMD persisted (8/10 vs 17/87; P < 0.001). Table 4 shows numbers at risk from the date of apheresis.

Table 3 Univariable logistic regression: Baseline correlates of failure to reach chimeric antigen receptor T-cell infusion (n = 99; 12 events), n (%).

Variable	Not infused	Infused	OR (95%CI)	P value
Extramedullary disease	9/12 (75.0)	17/87 (19.5)	12.4 (3.0-50.6)	< 0.001
Plasma cell leukaemia	3/12 (25.0)	3/87 (3.4)	9.3 (1.6-53.4)	0.012
R-ISS stage III (vs I-II)	2/11 (18.2)	3/66 (4.5)	4.7 (0.7-31.9)	0.13
High-risk cytogenetics	4/11 (36.4)	29/85 (34.1)	1.10 (0.30-4.05)	0.88
Age (per 10 years)	Median 70.2	Median 67.7	1.24 (0.71-2.16)	0.45
Prior lines of therapy	Median 5	Median 5	-	0.4311

¹Wilcoxon rank-sum test; not modelled. Estimates are univariable, unadjusted for multiplicity, and unstable given 12 events. No multivariable model was fitted (events per variable < 10).

CI: Confidence interval; OR: Odds ratio; R-ISS: Revised International Staging System.

Table 4 Kaplan-Meier estimates and numbers at risk from the date of apheresis.

Months from apheresis	0	3	6	12	18	24	30
PFS - infused: At risk	87	82	65	37	16	6	1
PFS - infused: % (95%CI)	100	94.3 (86.7-97.6)	81.0 (71.2-87.6)	67.9 (56.9-76.8)	49.2 (37.5-60.0)	32.6 (21.1-44.6)	20.9 (9.6-35.2)
PFS - not infused: At risk	12	2	1	0	0	0	0
PFS - not infused: % (95%CI)	100	16.7 (2.7-41.3)	8.3 (0.5-31.1)	0	0	0	0
OS - infused: At risk	87	85	76	50	27	15	6
OS - infused: % (95%CI)	100	97.7 (91.1-99.4)	94.1 (86.6-97.4)	88.6 (79.6-93.8)	82.7 (72.2-89.5)	75.1 (62.6-84.0)	65.7 (48.3-78.5)
OS - not infused: At risk	12	3	2	0	0	0	0
OS - not infused: % (95%CI)	100	25.0 (6.0-50.5)	20.0 (3.5-46.0)	10.0 (0.6-35.8)	0	0	0

Median progression-free survival: Infused 17.5 months [95% confidence interval (CI): 12.7-21.6); not infused 1.5 months (95%CI 0.8-2.3)]; log-rank P < 0.001. Median overall survival: Infused not reached (95%CI not estimable); not infused 1.9 months (95%CI 1.0-6.2); log-rank P < 0.001. Where the survival function equalled exactly 0.50 across an interval, the median is reported as the interval midpoint. CI: Confidence interval; OS: Overall survival; PFS: Progression-free survival.

Taken together, attrition in this cohort was neither random nor purely logistical: It clustered in patients with aggressive baseline disease phenotypes and occurred at a median of 44 days after collection, within the expected manufacturing window. Because these two explanations are correlated and the cohort is small, they cannot be disentangled here.

DISCUSSION

In this descriptive intention-to-collect analysis of 99 patients with RRMM undergoing apheresis for planned CAR-T therapy, 87 (87.9%) were infused after a median of 54 days and 12 (12.1%) were not. Attrition was attributable to progression or myeloma-related death in 7 patients, non-myeloma death in 2, manufacturing or collection failure in 2, and infection in 1, and occurred at a median of 44 days after apheresis. Non-infused patients had a substantially higher baseline prevalence of EMD (75.0% vs 19.5%) and PCL (25.0% vs 3.4%), while age, prior lines, cytogenetic risk and R-ISS stage were similar. Survival anchored at apheresis was far shorter in the non-infused group (median PFS 1.5 months vs 17.5 months; median OS 1.9 months vs not reached), but this contrast is descriptive and substantially exaggerated by the time-dependent nature of the exposure; a time-dependent Cox analysis produced a much more modest estimate (HR for death 0.24, 95%CI: 0.11-0.53).

The proportion of patients not infused after apheresis in our cohort is consistent with published real-world experience, in which apheresis-to-infusion rates for commercial ide-cel and cilta-cel generally fall in the range of 85%-95%, with most attrition due to progression or death before manufacturing completion[4]. Al Hadidi et al[5] highlighted an earlier and larger hurdle, reporting a median of 3.7 months from waitlisting to apheresis and death in approximately one quarter of patients while waiting. Total attrition from the point of CAR-T eligibility to infusion is therefore considerably higher than the 12.1% we observed after apheresis, and our estimate should be read as a lower bound conditional on having reached collection. Pre-infusion attrition therefore remains a non-trivial barrier to CAR-T delivery in contemporary practice.

Conventional prognostic variables - age, R-ISS stage, high-risk cytogenetics and prior lines of therapy - did not differ between groups, and in this respect our cohort resembles published real-world ide-cel and cilta-cel series[4,11,12]. However, two markers of aggressive disease phenotype, EMD and PCL, were strongly over-represented among patients who never reached infusion. This is biologically coherent: Both are characterised by high proliferative rate and short doubling times, and both are recognised adverse features in real-world CAR-T cohorts[4,13,14]. It is also consistent with the observed timing of attrition, which occurred at a median of 44 days, before the median 54-day manufacturing interval had elapsed. The clinically actionable inference is not that biology is destiny, but that patients with EMD or PCL are the group in whom a 7- to 8-week wait is least survivable, and therefore the group for whom expedited slots, prioritised manufacturing and intensified bridging should be reserved[15-17].

We note, however, that 17 patients with EMD and 3 with PCL were successfully infused, so these features are not in themselves a contraindication to embarking on the CAR-T pathway; they identify patients who require the pathway to be compressed.

Among non-infused patients, the dominant reason for attrition was rapid progression or death before product delivery (58.3%), which parallels published real-world series[4,5]. Bridging therapy was used with similar frequency in both groups (37.9% vs 41.7%), and in 4 of the 5 non-infused patients who received it the disease progressed during or immediately after bridging. We are unable to determine whether bridging was ineffective or simply insufficient in the face of the underlying disease kinetics: The retrospective record did not capture bridging start and stop dates, number of cycles delivered, formal IMWG response assessment during the bridging period, or graded toxicity. Comparative statements about bridging efficacy therefore cannot be made from these data, and we have removed them. What can be said is that current bridging practice at our centre did not prevent attrition in the highest-risk phenotypes, and that prospective, protocolised capture of bridging exposure, response and toxicity should be a standard element of CAR-T datasets[18-20].

The difference in survival between infused and non-infused patients was large in absolute terms, and non-infused patients in our cohort, as in others, generally died within weeks of the decision point[5,21]. We caution against over-reading this contrast. Three factors inflate it: The guarantee time inherent in reaching infusion; confounding by the higher prevalence of EMD and PCL among the non-infused; and the very small size of the non-infused group, which yields imprecise medians. The appropriate summary is that failure to reach infusion identifies a patient population with a life expectancy measured in weeks, not that infusion would have conferred the observed difference had it been delivered.

From a clinical standpoint, our data support a two-part strategy. First, because attrition concentrated in patients with EMD and PCL, risk-stratified triage at the time of referral appears justified: These patients should be considered for expedited apheresis slots, prioritised manufacturing where available, more intensive and more frequently reassessed bridging, and early discussion of alternative or salvage options - including BCMA- or GPRC5D-directed bispecific antibodies, which do not require manufacturing - should progression occur during the wait. Second, because two-thirds of attrition events occurred inside the median manufacturing window, operational compression remains essential for the whole cohort: Point-of-care and rapid (“fast”) manufacturing platforms[14,21], allogeneic products, and elimination of administrative delay in referral and payer authorisation. These strategies are complementary, and our data are not able to establish the relative contribution of biology and logistics.

Operationally, our findings also reinforce the need for early referral to cellular therapy centers and multidisciplinary coordination. Because attrition also occurred in patients with standard-risk cytogenetics and non-advanced stage, no patient entering the pathway should be regarded as immune to it. The high attrition observed even among patients with standard-risk cytogenetics suggests that delays in referral, insurance authorization, or cell processing may inadvertently exclude otherwise eligible patients. Institutions should consider implementing expedited workflows and CAR-T readiness programs to minimize time-to-infusion. In parallel, policy-level efforts to improve manufacturing scalability and access could further enhance treatment equity and reduce pre-infusion mortality.

Our study contributes to the growing body of real-world evidence demonstrating that the period between apheresis and infusion remains a critical vulnerability in CAR-T therapy for RRMM. While our analysis focuses on the pre-infusion period, the entire therapeutic journey, including post-infusion management, is critical for long-term success. Patients with RRMM exhibit considerable tumor heterogeneity, and while managing the manufacturing window is paramount, post-infusion strategies are equally important. Future research should explore the role of minimal residual disease-guided maintenance therapy or other strategies to optimize the duration of treatment and avoid both under- and over-treatment, thereby maximizing the durable benefit of this therapy. The comparable baseline characteristics between infused and non-infused patients, coupled with dramatically different survival outcomes, highlight that logistical efficiency and optimized bridging strategies - not disease biology - will be central to improving outcomes. Future prospective studies should prioritize time-to-infusion as a key metric, evaluate predictive markers of rapid progression, and explore novel therapeutic platforms designed to bridge or eliminate this high-risk gap in the CAR-T treatment continuum.

This study’s major strengths include its real-world, IRB-approved, intention-to-collect design that anchors PFS/OS at apheresis - capturing the full at-risk window and avoiding infusion-only immortal-time bias - its inclusion of both infused and non-infused patients with largely comparable baseline characteristics, and its contemporary scope featuring balanced use of the two Food and Drug Administration-approved BCMA CAR-T products and two standard lymphodepleting regimens, which enhances relevance and external applicability; rigorous, standardized data abstraction and survival analyses (Kaplan-Meier/Log-rank) further bolster internal validity, while the explicit delineation of attrition causes (progression/death > manufacturing/infection) yields actionable operational insights for bridging and workflow optimization. Limitations, though secondary to the above strengths, include the retrospective single-center nature, small size of the non-infused cohort limiting power for multivariable predictors, potential unmeasured confounding and missing data, and limited detail on the timing and intensity of bridging therapies. While we identified the types of bridging therapies used, the small cohort size precluded a meaningful statistical analysis of their impact on outcomes, which remains an important area for future investigation. Furthermore, the limited detail on timing/intensity of bridging and time-to-infusion intervals, absence of correlative biomarkers (e.g., apheresis T-cell phenotypes), and lack of toxicity, quality-of-life, and cost data, all of which may temper generalizability and mechanistic inference.

CONCLUSION

In this descriptive, intention-to-collect real-world cohort, 12.1% of patients with RRMM who underwent apheresis for planned CAR-T therapy never received an infusion. Attrition occurred at a median of 44 days after collection - within the median 54-day manufacturing interval - and was concentrated among patients with EMD and PCL, indicating that failure to reach infusion reflects aggressive disease phenotype as well as elapsed time. Read together with the reported 3.7-month median wait and 26% waitlist mortality before apheresis, these findings define the pre-infusion period as the principal vulnerability of the myeloma CAR-T pathway. They argue for earlier referral, compressed manufacturing turnaround, risk-stratified prioritisation of patients with aggressive phenotypes, protocolised and prospectively documented bridging, and greater use of manufacturing-free salvage options during the wait. They also argue for a change in reporting convention: Real-world analyses and trials should present the full intention-to-collect denominator anchored at apheresis, accompanied by a flow diagram, manufacturing-interval distributions and time-dependent analytic methods, so that the risks borne by every patient who enters the pathway are visible.

ACKNOWLEDGEMENTS

This article is a revised and expanded version of a paper entitled “Real World Outcomes of Multiple Myeloma Patients Who Underwent Apheresis for Planned Chimeric Antigen Receptor (CAR) T Cell Therapy: A Single Center Experience” which was presented at ASH Conferences in San Diego, CA, United Sates, December 2024.