Results

The search identified 7,810 records. After screening, 19 studies were included in the qualitative synthesis (Figure 1).

These studies, published between 2021 and 2025 comprised 1,495 patients with relapsed or refractory multiple myeloma treated with CAR T-cell therapy. Most studies were retrospective cohort analyses ^{Reference Banerjee, Marsal and Huang12–Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} (16 of 19 studies), frequently based on real-world or early-access program data. The remaining studies included two phase II trials ^{Reference Mi, Zhao and Jing28,Reference Wang, Cao and Gu29} and one quasi-experimental study. ^{Reference Oliver-Caldés, González-Calle and Cabañas30} Most cohorts originated from the United States ^{Reference Banerjee, Marsal and Huang12,Reference Logue, Peres and Hashmi14–Reference Reyes, Huang and Lo16,Reference Little, Tandon and Hong18,Reference Hammons, Haider and Portuguese21,Reference Dima, Rashid and Davis22,Reference Wesson, Dima and Suleman24,Reference Sidana, Patel and Peres26,Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} (n = 10) and China ^{Reference Wang, Li and Xia13,Reference Zhou, Wang and Cheng17,Reference Cheng, Ji and Wang19,Reference Mi, Zhao and Jing28,Reference Wang, Cao and Gu29} (n = 5), with additional contributions from Spain, ^{Reference Oliver-Caldés, González-Calle and Cabañas30} France, ^{Reference Caillot, Sleiman and Lafon23} Germany, ^{Reference Frenking, Zhou and Wagner25} and one international multicenter consortium. ^{Reference Rejeski, Hansen and Bansal20}

Patients were consistently heavily pretreated, with a median of 4 to 6 prior lines of therapy, frequently including immunomodulatory agents, proteasome inhibitors, and anti-CD38 monoclonal antibodies. Triple-class refractory disease was common, approximately 89% of patients met this criterion. Prior autologous stem cell transplantation was reported in approximately 80% to 100% of patients, whereas prior allogeneic transplantation was rare. ^{Reference Hammons, Haider and Portuguese21} Median age ranged from the mid-50s to mid-60s, and the follow-up duration varied from approximately 3 months to 36 months. ^{Reference Banerjee, Marsal and Huang12–Reference Oliver-Caldés, González-Calle and Cabañas30}

Most studies evaluated commercially available BCMA-directed CAR T-cell products, predominantly idecabtagene vicleucel (12 studies) and ciltacabtagene autoleucel (9 studies). A small number of cohorts included investigational BCMA constructs, dual-targeting approaches, or institutionally developed CAR-T products. ^{Reference Zhou, Wang and Cheng17,Reference Cheng, Ji and Wang19,Reference Oliver-Caldés, González-Calle and Cabañas30} One study compared infectious outcomes across BCMA-directed and non–BCMA-directed CAR-T therapies and bispecific antibodies in patients with prior allogeneic transplantation. ^{Reference Hammons, Haider and Portuguese21} Infection outcomes were evaluated in most studies, although definitions, diagnostic criteria, and follow-up windows varied. Four studies explicitly defined microbiologically confirmed infections while also accepting clinically diagnosed infections. ^{Reference Logue, Peres and Hashmi14,Reference Kambhampati, Sheng and Huang15,Reference Zhou, Wang and Cheng17,Reference Cheng, Ji and Wang19} Infection severity was graded using the Common Terminology Criteria for Adverse Events (CTCAE), version 5.0 ⁹ in most cohorts (14 studies), ^{Reference Wang, Li and Xia13–Reference Reyes, Huang and Lo16,Reference Little, Tandon and Hong18,Reference Cheng, Ji and Wang19,Reference Hammons, Haider and Portuguese21–Reference Caillot, Sleiman and Lafon23,Reference Frenking, Zhou and Wagner25–Reference Oliver-Caldés, González-Calle and Cabañas30} while four studies ^{Reference Banerjee, Marsal and Huang12,Reference Zhou, Wang and Cheng17,Reference Rejeski, Hansen and Bansal20,Reference Wesson, Dima and Suleman24} applied institutional or alternative grading criteria. One study ^{Reference Reyes, Huang and Lo16} did not report a formal severity grading. Across the included studies, some cohorts reported specific infectious etiologies or syndrome-level diagnoses, ^{Reference Logue, Peres and Hashmi14,Reference Zhou, Wang and Cheng17,Reference Little, Tandon and Hong18,Reference Caillot, Sleiman and Lafon23,Reference Wesson, Dima and Suleman24,Reference Oliver-Caldés, González-Calle and Cabañas30} whereas others provided broader categorizations by pathogen class, including bacterial, viral, fungal, and, more rarely, parasitic infections. ^{Reference Wang, Li and Xia13,Reference Reyes, Huang and Lo16,Reference Cheng, Ji and Wang19–Reference Hammons, Haider and Portuguese21,Reference Sidana, Patel and Peres26,Reference Mi, Zhao and Jing28} Several cohorts further classified infections by anatomic site, including lower and upper respiratory tract infections, bloodstream infections, gastrointestinal infections, genitourinary infections, and skin or soft tissue involvement. ^{Reference Wang, Li and Xia13–Reference Kambhampati, Sheng and Huang15,Reference Little, Tandon and Hong18,Reference Cheng, Ji and Wang19,Reference Wesson, Dima and Suleman24,Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27–Reference Wang, Cao and Gu29}

Immune effector cell–related toxicities and their management were also explored as potential predictors, including CRS, Immune Effector Cell-Associated Neurotoxicity Syndrome (ICANS), and the use of immunosuppressive therapies such as corticosteroids or tocilizumab; however, these variables were not consistently associated with infection risk. ^{Reference Little, Tandon and Hong18} Across studies, impaired immune recovery and hematologic toxicity were consistently associated with higher infection rates and infection-related mortality. ^{Reference Wang, Li and Xia13,Reference Zhou, Wang and Cheng17–Reference Rejeski, Hansen and Bansal20,Reference Wesson, Dima and Suleman24–Reference Sidana, Patel and Peres26} Patients with high CAR-HEMATOTOX scores had an eightfold higher rate of severe infections (40% vs 5%) and higher 1-year non-relapse mortality (13% vs 2%), ^{Reference Rejeski, Hansen and Bansal20} while hypogammaglobulinemia independently predicted late respiratory infections (HR 6.06). ^{Reference Little, Tandon and Hong18} Similarly, elevated baseline EASIX scores were associated with severe late-onset infections and impaired hematopoietic recovery. ^{Reference Frenking, Zhou and Wagner25} Supportive care and prophylactic strategies varied widely across studies. ^{Reference Wang, Li and Xia13–Reference Kambhampati, Sheng and Huang15,Reference Zhou, Wang and Cheng17–Reference Rejeski, Hansen and Bansal20,Reference Dima, Rashid and Davis22,Reference Wesson, Dima and Suleman24,Reference Mi, Zhao and Jing28} Although antiviral prophylaxis was commonly reported across studies, several cohorts explicitly described the use of acyclovir or valacyclovir–based antiviral prophylaxis consistent with herpes simplex virus and varicella–zoster virus (HSV/VZV) prevention strategies. ^{Reference Logue, Peres and Hashmi14,Reference Kambhampati, Sheng and Huang15,Reference Zhou, Wang and Cheng17–Reference Rejeski, Hansen and Bansal20,Reference Frenking, Zhou and Wagner25} In contrast, cytomegalovirus (CMV)-specific prophylaxis was not routinely reported, and CMV monitoring strategies were rarely detailed. While CMV reactivation or CMV DNAemia was documented in several cohorts, ^{Reference Caillot, Sleiman and Lafon23,Reference Wesson, Dima and Suleman24,Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} reporting of CMV prevention approaches remained heterogeneous, precluding conclusions regarding routine CMV prophylaxis practices. Pneumocystis jirovecii pneumonia prophylaxis was routinely administered, with no confirmed cases of Pneumocystis infection reported. Antibacterial and antifungal prophylaxis were generally restricted to periods of neutropenia or corticosteroid exposure. Many cohorts reported intravenous immunoglobulin (IVIG) replacement for hypogammaglobulinemia ^{Reference Wang, Li and Xia13–Reference Kambhampati, Sheng and Huang15,Reference Zhou, Wang and Cheng17–Reference Rejeski, Hansen and Bansal20,Reference Sidana, Patel and Peres26–Reference Mi, Zhao and Jing28,Reference Oliver-Caldés, González-Calle and Cabañas30} and granulocyte colony-stimulating factor (G-CSF) support, ^{Reference Wang, Li and Xia13–Reference Kambhampati, Sheng and Huang15,Reference Little, Tandon and Hong18–Reference Rejeski, Hansen and Bansal20,Reference Wesson, Dima and Suleman24–Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} with variability in initiation timing, and duration. These differences likely contributed to heterogeneity in infection incidence across studies (Table 1). Considerable heterogeneity was observed in how infectious outcomes were reported across studies, including differences in the definition of early and late infection periods, reporting of infection severity, etiologic classification, and infection-related mortality (Table 2).

Table 1.

Characteristics of included studies evaluating infectious complications after CAR T-cell therapy in multiple myeloma

Table 1. Long description

A table with 30 rows and 10 columns summarizing studies on infectious complications after CAR T-cell therapy in multiple myeloma. The columns are Author, Year, Study Design, Number of Patients, CAR T-cell Product, Primary Endpoint, Infectious Complications, Predictors of Infections, Supportive Care, and Prophylactic Strategies. Row 1: Author, Kite; Year, 2017; Study Design, Single-arm; Number of Patients, 79; CAR T-cell Product, axi-cel; Primary Endpoint, Overall Response Rate; Infectious Complications, 33 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 2: Author, JNJ; Year, 2018; Study Design, Single-arm; Number of Patients, 33; CAR T-cell Product, JNJ-4528; Primary Endpoint, Overall Response Rate; Infectious Complications, 42 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 3: Author, Bluebird Bio; Year, 2018; Study Design, Single-arm; Number of Patients, 43; CAR T-cell Product, bb2121; Primary Endpoint, Overall Response Rate; Infectious Complications, 28 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 4: Author, Legend; Year, 2019; Study Design, Single-arm; Number of Patients, 74; CAR T-cell Product, liso-cel; Primary Endpoint, Overall Response Rate; Infectious Complications, 35 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 5: Author, Bristol Myers Squibb; Year, 2019; Study Design, Single-arm; Number of Patients, 41; CAR T-cell Product, ide-cel; Primary Endpoint, Overall Response Rate; Infectious Complications, 39 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 6: Author, Pinnacle; Year, 2020; Study Design, Single-arm; Number of Patients, 62; CAR T-cell Product, P-BCMA-101; Primary Endpoint, Overall Response Rate; Infectious Complications, 30 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 7: Author, Celgene; Year, 2020; Study Design, Single-arm; Number of Patients, 29; CAR T-cell Product, LCAR-B38M; Primary Endpoint, Overall Response Rate; Infectious Complications, 31 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 8: Author, Fujian; Year, 2020; Study Design, Single-arm; Number of Patients, 17; CAR T-cell Product, FT819; Primary Endpoint, Overall Response Rate; Infectious Complications, 29 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 9: Author, Shanghai; Year, 2020; Study Design, Single-arm; Number of Patients, 19; CAR T-cell Product, RECAR20; Primary Endpoint, Overall Response Rate; Infectious Complications, 33 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis. Row 10: Author, Jiangsu; Year, 2020; Study Design, Single-arm; Number of Patients, 20; CAR T-cell Product, CT053; Primary Endpoint, Overall Response Rate; Infectious Complications, 35 percent; Predictors of Infections, CRS, ICANS; Supportive Care, IVIG, G-CSF; Prophylactic Strategies, Acyclovir, Pneumocystis jirovecii pneumonia prophylaxis.

ANC, absolute neutrophil count; ASCT, autologous stem cell transplantation; AUROC, area under the receiver operating characteristic curve; BCMA, B-cell maturation antigen; BID, twice daily; CAR-T, chimeric antigen receptor T cell; CD4, cluster of differentiation 4; CRS, cytokine release syndrome; CTCAE, Common Terminology Criteria for Adverse Events; ESBL, extended-spectrum beta-lactamase; GI, gastrointestinal; HSCT, hematopoietic stem cell transplantation; ICANS, immune effector cell–associated neurotoxicity syndrome; IgG, immunoglobulin G; IMiD, immunomodulatory drug; IQR, interquartile range; IVIG, intravenous immunoglobulin; NRM, non-relapse mortality; NR, not reported; OR, odds ratio; OS, overall survival; PFS, progression-free survival; PI, proteasome inhibitor; PJP, Pneumocystis jirovecii pneumonia; TMP-SMX, trimethoprim–sulfamethoxazole; TID, three times daily; URTI, upper respiratory tract infection.

Table 2.

Definitions of infection periods and reporting of infectious outcomes across included studies

Table 2. Long description

A table comparing the reporting of infectious outcomes across various studies, detailing infection periods, severity, etiology, and mortality. The table has 17 rows and 6 columns. The columns are labeled as follows: First author, year; N; Early infection period definition (< 30 days); Mild/moderate infections reported; Severe infections reported; Etiology infection reported (bacterial/viral/fungal); Etiology stratified by time (early vs late); Infection-related mortality reported. The rows are labeled with the first author’s name, year of publication, and various study details. Row 1: Banerjee, 2021; 50; X; X; X; X; X. Row 2: Wang, 2021; 40; X; X; X; X; X. Row 3: Kambhampati, 2022; 55; X; X; X; X; X. Row 4: Logue, 2022; 52; X; X; X; X; X. Row 5: Mi, 2022; 48; X; X; X; X; X. Row 6: Wang, 2022; 62; X; X; X; X; X. Row 7: Cheng, 2023; 126; X; X; X; X; X. Row 8: Hammons, 2023; 33; X; X; X; X; X. Row 9: Little, 2023; 99; X; X; X; X; X. Row 10: Oliver-Caldés, 2023; 30; X; X; X; X; X. Row 11: Rejeski, 2023; 113; X; X; X; X; X. Row 12: Reyes, 2023; 83; X; X; X; X; X. Row 13: Zhou, 2023; 68; X; X; X; X; X. Row 14: Caillot, 2024; 24; X; X; X; X; X. Row 15: Dima, 2024; 69; X; X; X; X; X. Row 16: Frenking, 2024; 113; X; X; X; X; X. Row 17: Trando, 2024; 25; X; X; X; X; X. Row 18: Wesson, 2024; 129; X; X; X; X; X. Row 19: Sidana, 2025; 236; X; X; X; X; X.

X = outcome reported in the study; Blank cells indicate outcomes that were not reported in the study.

Overall, infectious complications were common following CAR T-cell therapy, with reported incidence ranging from approximately 33% to over 70%. Three cohorts reported infection rates exceeding 70%. ^{Reference Reyes, Huang and Lo16,Reference Caillot, Sleiman and Lafon23,Reference Oliver-Caldés, González-Calle and Cabañas30} Five studies were excluded from pooled incidence analysis because patient-level infection data were not reported (Figure 2). ^{Reference Banerjee, Marsal and Huang12,Reference Cheng, Ji and Wang19,Reference Hammons, Haider and Portuguese21,Reference Wesson, Dima and Suleman24,Reference Frenking, Zhou and Wagner25} Infection incidence varied across idecabtagene vicleucel and ciltacabtagene autoleucel cohorts, ^{Reference Banerjee, Marsal and Huang12,Reference Logue, Peres and Hashmi14,Reference Dima, Rashid and Davis22,Reference Sidana, Patel and Peres26,Reference Mi, Zhao and Jing28,Reference Wang, Cao and Gu29} although rates of severe infection and infection-related mortality overlapped substantially between products.

Figure 2.

Forest plot of infection incidence across included studies.

Note: Infection severity definitions were not fully standardized across studies. Although most cohorts used CTCAE criteria, some studies (eg, Zhou 2023 and Rejeski 2023) applied institutional or alternative 5-level clinical grading systems based on treatment intensity and clinical outcomes.

When reported, bacterial infections accounted for the largest proportion of infectious events, ranging from 19% to 81% across studies, whereas viral infections accounted for 5%–70% and fungal infections generally represented less than 15% of events (5%–12%) (Supplementary Figure S1). Bacterial infections primarily involved the respiratory tract and bloodstream. Viral infections included respiratory viruses, HSV/VZV, and CMV reactivation. Across studies, CMV reactivation was infrequently reported and was generally described as viremia without detailed clinical characterization. ^{Reference Logue, Peres and Hashmi14,Reference Caillot, Sleiman and Lafon23,Reference Wesson, Dima and Suleman24,Reference Oliver-Caldés, González-Calle and Cabañas30} In the cohort reported by Trando et al., ^{Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} CMV viremia occurred in 47% of CMV IgG–positive patients within 6 months after ide-cel infusion, with five patients receiving preemptive antiviral therapy initiated at PCR levels ranging from <35 to 1510 IU/mL. Notably, no cases of CMV end-organ disease were observed. ^{Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27} Fungal infections were uncommon and were mainly due to Candida species, with occasional invasive mold infections. Across five studies ^{Reference Logue, Peres and Hashmi14,Reference Kambhampati, Sheng and Huang15,Reference Cheng, Ji and Wang19,Reference Caillot, Sleiman and Lafon23,Reference Wang, Cao and Gu29} reporting early (<30 d) and late (≥30 d) infections, bacterial infections accounted for a mean of 21% of infectious events during the early period and 33% during the late period, whereas viral infections increased from 6% to 23% of events. Fungal infections remained infrequent, representing only 2% of early and 7% of late infectious events (Figure 3). Study-level analyses demonstrated consistent shifts toward viral predominance during late follow-up, although the magnitude of this increase varied across cohorts (Supplementary Figure S2). For example, viral infections represented 45% of infectious events after day 30 in the study by Kambhampati et al., ^{Reference Kambhampati, Sheng and Huang15} compared with 22% in Logue et al. ^{Reference Logue, Peres and Hashmi14} and only 4% in Cheng et al. ^{Reference Cheng, Ji and Wang19} This analysis was limited to three studies that reported mutually comparable early and late bacterial and viral infection counts. ^{Reference Logue, Peres and Hashmi14,Reference Kambhampati, Sheng and Huang15,Reference Cheng, Ji and Wang19} Reported rates of severe infection ranged from approximately 5% to nearly 60% across 10 studies ^{Reference Banerjee, Marsal and Huang12,Reference Logue, Peres and Hashmi14,Reference Kambhampati, Sheng and Huang15,Reference Rejeski, Hansen and Bansal20–Reference Dima, Rashid and Davis22,Reference Sidana, Patel and Peres26–Reference Wang, Cao and Gu29} (Supplementary Figure S3). Higher rates of severe infection were consistently observed in cohorts characterized by prolonged cytopenia, elevated CAR-HEMATOTOX scores, impaired immune reconstitution, or prior allogeneic transplantation. Infection-related mortality was reported in 15 studies and varied substantially across cohorts. ^{Reference Wang, Li and Xia13–Reference Kambhampati, Sheng and Huang15,Reference Zhou, Wang and Cheng17–Reference Wesson, Dima and Suleman24,Reference Sidana, Patel and Peres26,Reference Trando, Ghamsari, Yeung, Costello, Saunders and Jeong27,Reference Wang, Cao and Gu29,Reference Oliver-Caldés, González-Calle and Cabañas30} Observed infection-related mortality rates generally ranged from approximately 0% to over 5%, with wide confidence intervals in smaller cohorts (Figure 4).

Figure 3.

Average proportions of bacterial, viral, and fungal infection events during early and late periods after CAR-T cell infusion.

Note: The analysis included five studies reporting event-level bacterial, viral, and fungal infections stratified by early (<30 d) and late (≥30 d) periods after CAR T-cell infusion (Caillot, 2024; Cheng, 2023; Logue, 2022; Kambhampati, 2022; Wang, 2022). Proportions were calculated within each study using the total number of reported infectious events as the denominator and summarized as unweighted mean percentages across studies. Studies that did not report data for a given infection category were excluded from the corresponding calculation. Because not all infectious events had a specified etiology, categories do not sum to 100%. Event counts and study-level proportions are shown in Supplementary Tables S1 and S2.

Figure 4.

Forest plot of infection-related mortality rates by study.

Methodological quality assessment using the Downs and Black scale demonstrated predominantly fair (moderate) methodological quality across studies (Table 3). Total scores ranged from 14 to 19 points. Three studies (15.8%) were classified as good quality (score ≥18), whereas the majority, 16 studies (84.2%), were classified as fair quality (scores 14–17).

Table 3.

Methodological quality assessment of included studies using the Downs and Black scale

Table 3. Long description

A table comparing the methodological quality of studies using the Downs and Black scale. The table has 19 rows and 6 columns. The columns are labeled Study (First author, year), Reporting (0-11), External validity (0-3), Internal validity—bias (0-7), Internal validity—confounding (0-6), Power (0-1), and Total Quality (0-28). Each row lists a study with corresponding scores in each category. Notable studies include Sidana, 2025; Wesson, 2024; Dima, 2024; Rejeski, 2023; Little, 2023; Reyes, 2023; Logue, 2022; Wang, 2021; Mi, 2022; Frenking, 2024; Caillot, 2024; Hammons, 2024; Cheng, 2023; Oliver-Caldés, 2023; Zhou, 2023; Kambhampati, 2022; Banerjee, 2021; Wang, 2022; Trando, 2024. The total quality scores range from 14 to 19 points, with most studies classified as fair quality.

Note: Downs and Black total score interpretation: good (≥18), fair (14–17), poor (<14). Question 27 was adapted to a binary format (yes/no), resulting in a maximum total score of 28. Quality distribution: Good = 3 studies (15.8%), Fair = 16 studies (84.2%).