Lymphodepleting chemotherapy practices and effect on safety and efficacy outcomes in patients with solid tumours undergoing T cell receptor-engineered T cell (TCR-T) Therapy: a systematic review and meta-analysis

Background T cell receptor-engineered T cell (TCR-T) therapy has shown promising efficacy in advanced solid tumours. Lymphodepleting (LD) chemotherapy improves TCR-T cell therapy efficacy but is associated with significant toxicities. Evidence is sparse regarding the optimum LD regimen for TCR-T cell therapy in solid tumours. Methods A systematic review was conducted of interventional, prospective clinical trials describing LD practices prior to TCR-T cell therapy in patients with advanced solid tumours. The objective was to define LD regimens administered prior to TCR-T cell therapy and their effects on specific safety and efficacy outcomes in this patient population. Results Searches returned 484 studies, 19 (231 patients) met the eligibility criteria. Cyclophosphamide (cyclo) 60 mg/kg daily (2 days), plus fludarabine (fludara) 25 mg/m2 daily (5 days) was the most common LD regimen (38% of studies). Higher dose LD regimens were associated with increased pooled incidence rates of febrile neutropaenia compared to low dose (0.64, [95% Confidence interval (CI): 0.50–0.78], vs. 0.39 [95% CI: 0.25–0.53], respectively) but were not significantly associated with higher objective responses (odds ratio: 1.05, 95%CI: 0.60–1.82, p = 0.86). A major shortfall in safety data reporting was identified; determination of LD regimen effects on many safety outcomes was not possible. Conclusion Standard consensus guidelines for the design and reporting of adoptive cell therapy (ACT) studies would facilitate accurate risk–benefit analysis for optimising LD regimens in patients with advanced solid tumours. Supplementary Information The online version contains supplementary material available at 10.1007/s00262-022-03287-1.


Introduction
Patients with advanced solid tumours relapsed or refractory to standard of care therapies have a poor prognosis, with few further treatment options. Adoptive T cell therapy (ACT) has shown promising efficacy in various solid tumour types [1,2]. Since early trials of tumour infiltrating lymphocytes (TILs), T cell therapies have evolved and include chimeric antigen receptor-T (CAR-T) cell and T cell receptor-engineered T (TCR-T) cell therapies [3]. TCR-T cells target a greater variety of antigens than CAR-T cells, which recognise only cell surface antigens, so may have greater efficacy in solid tumours, supported by response rates from recent clinical studies [4][5][6]. To date, no TCR-T cell therapies are licensed, albeit a positive risk benefit balance has been demonstrated in melanoma and other solid malignancies [6][7][8][9].
An essential requirement for ACT efficacy is the in vivo expansion and persistence of T cells once reinfused [1,10,11]. Clinical trial data show that lymphodepleting conditioning (LD) regimens consisting of cyclophosphamide (cyclo) and fludarabine (fludara) prior to cell reinfusion improves T cell in vivo expansion and persistence [12,13].
ACT is associated with specific toxicities, including cytokine release syndrome (CRS) and immune effector cellassociated neurotoxicity syndrome (ICANS) [14]. Neurological adverse reactions are common to CAR-T cell therapy and cyclo/fludara dosing, so when given together, the lack of randomised controlled studies render causality assignment difficult for neurological toxicities [15].
Most safety and efficacy outcome data for LD regimens derive from CD19 CAR-T cell studies in haematological malignancies, which may not apply to TCR-T cell therapy in solid tumours. Differences include the direct access of CAR-T cells to tumour, high local concentrations of immune effector cells implicated in CRS, and neurological toxicity due to 'on target, off tumour' toxicity in haematological malignancies [16]. LD in haematological malignancies results in direct reduction of the malignant cell population. Bone marrow reserve is often lower than for patients with solid tumours, so rates of cytopaenias and bone marrow aplasia may differ for a given LD regimen [12]. Solid tumours inhabit an immunosuppressive microenvironment, the aim of LD being to reduce immunosuppressive cells.
Given the pivotal importance of LD in ACT, lack of randomised controlled clinical trials comparing LD regimens, and scarcity of TCR-T cell trials in solid tumours, leading to a lack of consensus regarding optimal regimens, a systematic review may help fill this knowledge gap. This review will summarise the lymphodepleting practices employed prior to TCR-T cell therapies and effects on safety and efficacy outcomes in patients with relapsed, refractory solid tumours, participating in interventional, prospective clinical studies.

Literature search and eligibility criteria
This review was written as part of a Master of Science (M.Sc.) thesis, planned prospectively with the supervisory committee and registered with the University of Manchester. It is reported per the Preferred Reporting Items for Systematic Reviews and Meta-analyses, 'PRISMA-P' guidelines [17,18] Table 2). References of related and systematic reviews were cross-referenced with the search strategy to include additional relevant studies. ClinicalTrials.gov was searched for updated results or missing methodological details not included in the original articles/abstracts.

Objectives
These were to define LD regimes administered prior to TCR-T cell infusion and effects on safety and efficacy outcomes in patients with solid tumours. LD was defined as a treatment to reduce the population of circulating lymphocytes, prior to infusion of TCR-T cells, including chemotherapy, radiotherapy and/or any other method specified. Highdose LD regimens included total doses of cyclo ≥ 120 mg/kg and fludara ≥ 100 mg/kg 2 . Specific safety outcomes included adverse events (AEs) of cytopaenias, bone marrow aplasia, infections, CRS, neurotoxicity, and Graft-versus-host disease (GvHD). Specific efficacy outcomes included objective response rate (ORR), defined by the Response Evaluation Criteria in Solid Tumours (RECIST) Guideline, v1.1. [19].

Risk of bias
No specific tool exists for risk of bias assessment of singlearm interventional studies; therefore, Grigor and colleagues' risk of bias tool for interventional study designs was used to assess the quality of included studies [4,20] (Supplementary Table 3). Studies that met all eligibility criteria but had a high risk of bias were included in the review but excluded from data analyses per the Cochrane Handbook recommended the best practice to reduce the risk of reporting bias [21].

Data synthesis and analysis
All analyses were conducted on an intention-to-treat population. Heterogeneity across subgroup analyses was tested by I-square (I 2 ) and Chi-square tests. The Chi-square test measures the existence of a significant heterogeneity, whilst I 2 quantifies the magnitude of heterogeneity in the effect size. If both tests showed low heterogeneity, the fixed effect model (Mantel-Haenszel method) was applied. If a high degree of heterogeneity was observed, a random-effect model was applied. A sensitivity analysis was conducted to identify sources of heterogeneity using "one-out" approach [22]. Single studies that lead to a decrease of I 2 < 25% were reported if identified.
The analysis was conducted in STATA 16.0 [48]. The metaprop command was applied to calculate pooled incidence rates of AEs and ORR. To test the impact of the ordered values of the dose level categories of cyclo or fludara on the objective response status, nptrend command was used in STATA to run the Cuzick test.

Characteristics of the included studies and participants
Systematic searches retrieved 484 potentially relevant citations; 433 were excluded based on abstract/title. 51 studies remaining were read in full, and 12 articles (185 patients) and 7 abstracts (46 patients) met the eligibility criteria ( Supplementary Fig. 1). Two articles [23,24] reported all outcome data for cohort 1, and efficacy data only for all 4 cohorts, respectively, for the same study. The four cohorts received different LD regimens. Data for all 4 cohorts were updated in a recent abstract [25]. Hence, both articles and abstract were included. Two abstracts [26,27] reported data for two patient cohorts respectively, separately, receiving different LD regimens, within the same study; hence, both are included. All studies administered autologous T cells. No paediatric patients were included ( Table 1).

Risk of bias and quality assessment of included studies
Risk of bias is summarised across all studies (Supplementary Fig. 2), with each risk of bias item assessment for individual studies presented ( Supplementary Fig. 3). All studies were single-arm, non-randomised, and non-controlled, and therefore had a high risk of interpretation bias. Methods for measurement of safety and efficacy outcomes were not clearly stated in 40% of studies. Every study lacked an independent, blinded assessor for outcomes and showed a high risk of bias regarding patient recruitment. 40% of studies clearly reported AEs, with 95% of studies reporting ORR, demonstrating a strong reporting bias for favourable versus unfavourable outcomes. The follow-up period for data cutoff, reported in 53% of studies, was highly variable, potentially affecting data reporting (AEs resolution, delayed/ chronic AEs, delayed responses, response duration, overall survival).

The pooled incidence of adverse effects
Nine studies reported detailed adverse effects (AEs) data. Pooled incidence rates showed a trend for febrile neutropaenia and Grade ≥ 3 anaemia occurring more frequently among patients receiving high-dose LD regimens (Table 2). Pooled incidence rates of febrile neutropaenia among patients receiving high-dose LD regimens and low-dose LD regimens were 0.64 (95% CI 0.50-0.78) and 0.39 (95% CI 0.25-0.53), respectively (Fig. 1). Significant heterogeneity was observed in the high-dose subgroup and overall analyses of febrile neutropenia (I 2 37.37% and 59.49%, respectively).
Based on the lack of reported safety data, it was not possible to determine the effect of LD regimen on AEs of CRS, neurotoxicity, infection, or GvHD. (Supplementary Tables 5,  6).

Objective response rate (ORR)
The ORR ranged from 0 to 67% (Supplementary Table 7). Using the reported aggregate objective response rates, the pooled ORR for low-LD-dose and high-LD-dose cohorts were 0.31 (95% CI 0.15-0.46) and 0.36 (95% CI: 0.26-0.45), respectively ( Supplementary Fig. 4). We were able to extract the individual cyclo and fludara doses with the corresponding response status for 231 patients. High LD dose was not associated with a higher likelihood of objective response (odds ratio: 1.05, 95%CI: 0.60-1.82, p = 0.86). We tested the impact of ordered dose levels of fludara and cyclo either alone or combined on the objective response (Fig. 2a). Only higher dose levels of fludara were associated with a significantly higher probability of objective responses (p = 0.03) (Fig. 2b). This was not significant with cyclo (p = 0.24) nor combined cyclo/fludara dose levels (p = 0.28) (Fig. 2c,d).

Discussion
This systematic review identified a major shortfall in reporting of data from TCR-T cell clinical trials in patients with solid tumours. Firstly, LD regimens were not sufficiently detailed in 9 of the potentially suitable 51 studies, increasing the potential for bias (missing data). Secondly, reporting of key, life-threatening AEs of CRS, neurological toxicity, infection, and GvHD was frequently inconsistent or missing. For studies split into cohorts receiving different LD regimens, data were reported for the entire study population, not per cohort. Where reported, criteria used for CRS definition and severity were rarely cited. Some authors reported the three components of CRS (fever, hypotension, and hypoxia) as separate AEs [7]. Taken together, the lack of AE data reporting makes the correlation of LD regimens with safety outcome data, including the true rate of CRS for a given LD regimen, difficult. However, a trend for an association between high-dose LD regimens and the incidence of febrile neutropaenia and grade ≥ 3 anaemia was observed, albeit heterogeneity between high-and low-dose LD groups was significant. Dudley and colleagues first demonstrated that a LD regimen containing total doses of cyclo 120 mg/kg and fludara 100 mg/m 2 significantly increased the efficacy of TIL therapy in metastatic melanoma, with manageable toxicity [13]. Adoption of this LD regimen in CAR-T cell therapy demonstrated that higher dose cyclo/fludara use was also associated with improved efficacy [38][39][40] but an increased toxicity burden in both haematological and solid tumours [12]. Consequently, a range of lower cyclo/ fludara doses have been studied as LD in ACT therapy in solid tumours [12,[49][50][51]. For this review, we considered the original standard LD regimens containing total doses of cyclo ≥ 120 mg/kg and fludara ≥ 100 mg/m 2 as ''high dose''. Cyclophosphamide monotherapy used in 2 of the studies was considered a "low-dose" LD regimen. Higher dose fludara, but not higher dose cyclo nor cyclo/fludara, was associated with a significantly higher probability of objective responses (p = 0.03, p = 0.24, and p = 0.285, respectively) in this review, supported by data from two non-Hodgkin's Lymphoma studies, where LD regimes containing fludarabine were associated with improved CAR-T cell expansion, persistence and efficacy, compared with non-fludarabine containing regimes [38,39]. A multivariate analysis of factors affecting progression-free survival (PFS) in a study of CAR-T cells in B cell lymphoma, using low-or highintensity LD, found that specific cytokine concentrations above the median were associated with improved PFS [40]. PFS was improved in patients receiving high-intensity LD and achieving a favourable cytokine profile, compared with those receiving the same high-intensity LD without achieving a favourable cytokine profile [40], suggesting that the effects of LD regimen on specific cytokine profiles are more important in terms of CAR-T cell efficacy than the intensity of LD. It remains to be seen whether this applies to TCR-T cell therapy in solid tumours, as data for cytokine profiles were not collected in this review.
Included studies employed a range of adjunct therapies. The rates of febrile neutropaenia were slightly lower (range 0-0.4) in the 2 studies reporting G-CSF use [23,24] than for studies not reporting G-CSF use, but 2 of the cohorts used lower doses of cyclo with no fludara (100/0 and 50/0, [23]). As a potential routine component of TCR-T cell therapy, reporting of G-CSF use may have been omitted, making it difficult to determine its effect on AE incidence. The addition of IL-2 (9/19 studies) does not seem to have significantly influenced efficacy outcomes in this review, but data are confounded by variations in doses of IL-2 (72,000, 500,000, 720,000 units tds), doses of TCR-T cells and LD therapy, and use of other adjunct therapies (DC and peptide vaccinations). For these reasons, and due to inconsistent AE reporting (CRS, neurotoxicity, GvHD), the effect of adjunct therapies on safety and efficacy outcomes is therefore confounded and difficult to interpret.

Limitations of this review
Peer review of the search strategy may have reduced potential biases. The process of article and abstract selection from systematic searches for inclusion in the review, and quality assessments of the risk of bias for individual studies, were both intrinsically biased, relying on the interpretation and judgement of a single reviewer (KO), with no second reviewer or opportunity for discussion with peers (selection and reporter bias). This review demonstrated a high level of clinical and methodological heterogeneity between and within studies of TCR-T cells in patients with solid tumours, in keeping with findings from similar systematic reviews/metanalyses of studies in CAR-T cells [4,41]. Notable was the number of variables changing concurrently within a study and within small patient cohorts. Patients often had different tumour types, LD regimens, TCR-T cell doses or adjunct therapies, so it was impossible to decipher the effect of changing a single variable on an outcome. Therefore, the interpretation of the effect of LD regimen on safety and efficacy outcomes is heavily confounded by multiple factors, some largely outside study control, including patient characteristics (age, performance status, tumour burden at baseline, tumour and offtarget expression of target antigen, number/nature of prior therapies, comorbidities, concomitant medications, cyclo/ fludara PK), but also TCR-T cell variables (target antigen, cell dose, 'fitness'[phenotype], persistence and expansion of cells, number of infusions of cells), adjunct therapies (IL-2, dendritic cell or peptide vaccination, immune checkpoint inhibitors) and follow-up periods.
Due to the rarity of TCR-T cell therapy, clinical trial designs should focus on key objectives and outcomes, fixing other variables so that confounding is minimised and the interpretation of an effect on an outcome is as robust as possible. Standardisation of clinical trial designs in TCR-T cell therapy and ACT in general, with a consensus for reporting LD regimen, TCR-T cell dose, adjunct therapies, defined safety/efficacy outcomes, and follow-up times, per the master protocols from the Pan American/World Health Organisations' Master Protocol for clinical trials in COVID-19, may allow more valid data pooling and maximise available evidence [42]. Consensus should be reached on CRS definition and grading, neurological toxicities, and GvHD and adopted globally.
The risk of bias in the results was very high, i.e. the risk of over-or under-estimation of the true intervention effect (LD regimen) on safety and efficacy outcomes post TCR-T cell therapy. There are specific challenges involved in the production of robust, quality data for risk-benefit analyses of ACT using randomised controlled clinical trials [43]. TCR-T cell therapy is a bespoke product for patients with relapsed/ refractory life-threatening disease, with no other suitable treatment options. Few patients are suitable for TCR-T cell therapy; their acquisition, manufacture, storage, shipping, clinical delivery, and patient care are highly complex, with a substantial production failure rate [1]. Together with high costs, these factors preclude larger clinical trials, and the lack of suitable alternative therapies prevents the ethical use of comparator arms.
The primary outcome for this review was to summarise the LD dosing regimens prior to TCR-T cell therapy in patients with solid tumours. A more accurate primary outcome would have been a summary of exposures of the chemotherapies used and safety and efficacy outcomes.
Hepatic and renal function affects cyclo and fludara pharmacokinetics (PK), so for a given dose, there is high interpatient variability in systemic exposure, despite adjustments for body weight or body surface area (BSA), potentially influencing outcomes [44][45][46][47]. Sparse PK sampling during LD may allow a more accurate correlation of cyclo/fludara exposure with safety and efficacy outcomes, with optimisation of future LD dosing, and improvement in LD-associated risks and benefits. represented on the y-axis as the natural log of the sum of ranks of responses computed by Cuzick's test. c Dose levels of cyclophosphamide were not significantly associated with a higher probability of objective responses. d Dose levels of cyclophosphamide and fludarabine were not significantly associated with a higher probability of objective responses

Conclusion
To our knowledge, this is the first systematic review of lymphodepleting practices employed prior to TCR-T cell therapy in patients with relapsed, refractory solid tumours, participating in interventional, prospective clinical studies. Such studies are infrequent, extremely heterogeneous, and the risk of bias for outcomes due to publication, methodological, reporting, and interpretation biases in this review is very high. The most commonly used LD regimen (38% of studies/cohorts) prior to TCR-T cell therapy was 60 mg/kg cyclo and 25 mg/m 2 fludara, given daily for 2 and 5 days, (total doses 120 mg/kg, 125 mg/m 2 ), respectively. Febrile neutropaenia and grade ≥ 3 anaemia tended to occur more frequently among patients receiving high-dose LD regimens. Higher-dose fludara (≥ 100 mg/m 2 ) was associated with a significantly higher probability of objective responses, but this did not apply to cyclo nor combined cyclo/fludara at higher doses. Taken together, these data imply that lower doses of cyclophosphamide may be adopted without compromising TCR-T cell therapy efficacy outcomes.
Safety outcome reporting was inconsistent. Standardised TCR-T cell therapy clinical trial designs and data reporting may allow appropriate pooling of data and facilitate the development of robust evidence bases for future optimisation of LD regimens which may reduce morbidity mortality, shorten hospital stay, and reduce intensive care unit admissions, resulting in a longer and better quality of life for patients with advanced solid tumours with no other treatment options.
Author contributions FT proposed the study concept and objectives, KO designed the study, acquired the data and prepared all tables and figures except for table 2 and figures 2, 3b), 3c) and 3d) which were prepared by KS and RG. KS and RG performed all statistical analyses KO drafted the main manuscript text with input from RG and KS. All authors contributed to the analysis and interpretation of data. All authors critically reviewed the manuscript. KO obtained funding. Supervision: FT. KO had full access to all the data in the study and takes full responsibility for data integrity. KO and FT were involved in concept and design. All authors analysed or interpreted the data. KO, RG, and KS drafted the manuscript. All authors critically revised the manuscript. KO obtained funding. All authors were involved in administrative, technical, or material support. FT was involved in supervision.
Funding This work was part of an application for the degree of Master of Science at Manchester University, funded by Owen Clinical Consulting Limited.

Conflict of Interest
The authors declare no conflicts of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.