Pretransplant T-cell immune imbalance predicts slow engraftment after autologous hematopoietic stem cell transplantation in lymphoma

Abstract

BACKGROUND

Autologous hematopoietic stem cell transplantation (auto-HSCT) is a cornerstone therapeutic strategy for lymphoma. Slow engraftment after auto-HSCT increases the risks of infection and hemorrhage, prolongs hospital stay, and is therefore a critical determinant of transplant safety and clinical outcomes. To date, clinical evidence on risk factors for slow engraftment after auto-HSCT in patients with lymphoma remains insufficient, particularly regarding the pretransplant immune microenvironment and lymphocyte subsets.

AIM

To explore clinical and immune risk factors and construct a prediction model for slow engraftment after auto-HSCT in patients with lymphoma.

METHODS

We retrospectively enrolled 166 patients with lymphoma who underwent auto-HSCT at two Chongqing centers from July 2022 to June 2025. Patients were divided into slow engraftment (neutrophil engraftment > 10 days or platelet engraftment > 12 days, based on the median times) and early engraftment groups. Baseline clinical features, transplantation parameters, pretransplant lymphocyte subsets, and inflammatory cytokines were collected. Univariate and multivariate logistic regression analyses were used to identify independent risk factors, construct a combined predictive model, and assess its performance.

RESULTS

Multivariate analysis identified advanced age, reinfused CD34+ cell dose < 3.5 × 10⁶/kg, a higher proportion of CD8+ T cells, a lower absolute CD4+ T-cell count, and prolonged peritransplant fever as independent risk factors for slow engraftment after auto-HSCT in lymphoma patients. Patients receiving a high CD34+ cell dose (≥ 3.5 × 10⁶/kg) achieved faster neutrophil and platelet engraftment than those in the low-dose group. Compared with patients with early engraftment, those with slow engraftment showed pretransplant T-cell subset imbalance and elevated interleukin-2 and interferon-γ levels. A prediction model integrating these variables demonstrated good predictive performance and outperformed individual indicators, achieving an area under the curve (AUC) of 0.780 for identifying patients at risk of slow engraftment after auto-HSCT.

CONCLUSION

Slow engraftment after auto-HSCT is associated with clinical and immune factors. Pretransplant T-cell imbalance predicts delayed engraftment, and a combined model improves risk prediction (AUC = 0.780).

Key Words: Autologous hematopoietic stem cell transplantation; Slow engraftment; T lymphocyte subsets; T-cell immune imbalance; Risk factor; Lymphoma

Core Tip: This retrospective study enrolled 166 lymphoma patients undergoing autologous hematopoietic stem cell transplantation. Advanced age, low CD34+ cell dose, prolonged peritransplant fever, and pretransplant Tcell immune imbalance were independent risk factors for slow engraftment. A combined predictive model (area under the curve = 0.780) showed good performance. Pretransplant immune evaluation helps identify high-risk patients and improve transplant safety.

INTRODUCTION

Autologous hematopoietic stem cell transplantation (auto-HSCT) remains a vital therapeutic approach for lymphoma. Despite the widespread use of chimeric antigen receptor T cell therapy and novel targeted agents, auto-HSCT is still recommended by national and international guidelines as the standard of care for patients with chemotherapy-sensitive relapsed/refractory lymphoma and high-risk newly diagnosed lymphoma[1-4]. Although auto-HSCT has demonstrated reliable efficacy and is widely implemented in clinical practice, posttransplant slow engraftment remains a common and urgent clinical challenge. Slow engraftment prolongs hospital stay, increases medical burden, and elevates the incidence of early complications, such as severe infection and hemorrhage. These adverse outcomes compromise long-term patient prognosis and substantially restrict the safety and clinical benefits of auto-HSCT[5-7]. Existing studies have confirmed that a low infused CD34+ cell dose is a well-established risk factor for delayed engraftment (DE) following auto-HSCT[8-10]. In addition, age, baseline peripheral blood parameters, and pretransplant clinical characteristics have also been reported to be closely associated with impaired posttransplant engraftment[11,12].

Accumulating evidence indicates that disrupted immune homeostasis plays an important role in diseases characterized by impaired hematopoietic function in the bone marrow. Imbalanced T-cell subsets, such as reduced regulatory T cells and aberrant activation of cytotoxic T cells, can disrupt the hematopoietic microenvironment, suppress the function of hematopoietic stem and progenitor cells, and thus contribute to the progression of bone marrow failure diseases[13-15]. Moreover, several studies have identified a correlation between natural killer cells and graft failure[16]. The close link between immune cell homeostasis and hematopoietic function suggests that disorders of lymphocyte subsets may also affect hematopoietic recovery after auto-HSCT. Previous studies on allogeneic hematopoietic stem cell transplantation have demonstrated that T-cell depletion strategies for graft-vs-host disease prophylaxis are frequently associated with an increased risk of poor engraftment, indirectly supporting the critical role of T-cell homeostasis in posttransplant engraftment[17,18]. Current research mainly focuses on the effects of baseline clinical features and infused stem cell dose on hematopoietic reconstitution following auto-HSCT. In contrast, studies focusing on overall pretransplant immune status, especially the correlation between peripheral blood lymphocyte subset distribution and slow engraftment, remain scarce. The relevant risk factors and potential early warning biomarkers are still poorly defined, and effective assessment systems and predictive models for early clinical screening are lacking.

Accordingly, this retrospective analysis used clinical data from patients with lymphoma undergoing auto-HSCT. The primary objectives were as follows: (1) To systematically screen and validate independent clinical risk factors for slow engraftment after auto-HSCT; (2) To explore the correlation between pretransplant lymphocyte subsets and the rate of engraftment and clarify the clinical implications of immune imbalance; and (3) To integrate clinical and immune-related indicators to develop a concise and effective combined prediction model for accurate identification of high-risk populations, thereby providing evidence-based references for individualized peritransplant assessment and refined management of patients with lymphoma receiving auto-HSCT.

MATERIALS AND METHODS

Study design and patients

This was a retrospective, observational study that enrolled patients with lymphoma undergoing auto-HSCT at the Second Affiliated Hospital of Chongqing Medical University and Chongqing University Cancer Hospital from July 2022 to June 2025. The inclusion criteria were as follows: (1) A definitive histopathological diagnosis of lymphoma; (2) Receipt of standardized peripheral blood auto-HSCT; and (3) Availability of complete baseline clinical data, peritransplant records, laboratory parameters, and posttransplant follow-up information.

The exclusion criteria were as follows: (1) Patients treated with allogeneic hematopoietic stem cell transplantation; and (2) Individuals with substantial missing key clinical, laboratory, or follow-up data that precluded valid evaluation.

Neutrophil engraftment was defined as a peripheral blood neutrophil count of ≥ 0.5 × 10⁹/L for three consecutive days after transplantation, whereas platelet engraftment was defined as a platelet count of ≥ 20 × 10⁹/L for three consecutive days without platelet transfusion support for seven days[19]. All of the above indicators were uniformly determined based on laboratory test results from our central laboratory. Although no universally recognized definition of DE has been established, several studies define DE as a platelet count ≤ 50000/μL, hemoglobin (HB) ≤ 8 g/dL, or absolute neutrophil count ≤ 1000/mm³ starting on day 30 following autologous hematopoietic cell infusion[20]. The median values of neutrophil and platelet engraftment times in the overall study population were used as cutoffs for grouping. The median time to neutrophil engraftment was 10 days, and the median time to platelet engraftment was 12 days. These cutoffs were derived from the clinical characteristics of the enrolled patients and served as a data-driven stratification criterion. The early engraftment group was defined as neutrophil engraftment within ≤ 10 days combined with platelet engraftment within ≤ 12 days. The slow engraftment group was defined as neutrophil engraftment time > 10 days or platelet engraftment time > 12 days. Given the absence of a unified threshold for slow engraftment, sensitivity analysis was not performed. Median-based grouping is a conventional approach, and no circularity was present in this design. This study focused on intergroup differences in immune profiles. Significant heterogeneity in the composition of pretransplant T lymphocyte subsets was observed between the two cohorts, reflecting distinct T-cell immune homeostasis states. To further explore the impact of infused CD34+ cell dose on engraftment, patients were stratified according to the threshold of 3.5 × 10⁶/kg. The low CD34+ dose group was defined as a reinfused CD34+ cell count < 3.5 × 10⁶/kg, and the high CD34+ dose group was defined as a reinfused CD34+ cell count ≥ 3.5 × 10⁶/kg. The primary endpoint of this study was slow engraftment. The research protocol was reviewed and approved by the Institutional Ethics Committee of the hospital. Given the retrospective and observational design of this study, a waiver of written informed consent was formally approved. All group stratifications were performed based on real-world clinical data without artificial intervention. Grouping outcomes was applied to subsequent baseline comparisons and univariate and multivariate regression analyses.

Data collection

This was a retrospective observational study. Using a uniformly designed data collection form, we systematically reviewed and extracted information from electronic medical records, laboratory information systems, and pathological archives. All clinical and laboratory data for enrolled patients were comprehensively collected. The collected variables were categorized into four dimensions: (1) Baseline demographic and disease characteristics: Age, sex, pathological subtypes of lymphoma, Ann Arbor stage, risk stratification at diagnosis, and number of previous chemotherapy lines; (2) Peritransplant indicators: Specific conditioning regimens, infused CD34+ cell dose, and the duration of fever during transplantation and other key clinical events; (3) Routine laboratory parameters: Complete blood count and biochemical test results at initial diagnosis and before transplantation, such as HB, albumin (ALB), lactate dehydrogenase (LDH), and beta 2-microglobulin (β2-MG); and (4) Lymphocyte subset profiles: Pretransplant peripheral blood flow cytometry data, including the percentages and absolute counts of CD4+ T cells and CD8+ T cells, to evaluate baseline lymphocyte subset distribution. All data were independently entered and cross-checked by two researchers who were uniformly trained. Discrepant records were rechecked and corrected by reviewing original medical documents to minimize selection and entry bias, thereby ensuring data accuracy.

Statistical analysis

Statistical analyses were performed using SPSS 26.0 and GraphPad Prism 8.0. A two-tailed P value < 0.05 was considered statistically significant. According to the Shapiro-Wilk normality test, continuous data were presented as medians with interquartile ranges (IQR, Q1-Q3) or means ± SDs, whereas categorical data were expressed as n (%). For intergroup comparisons between the slow engraftment group and the early engraftment group, Pearson’s χ² test was used for categorical variables. The independent-samples t-test or the Mann-Whitney U test was used for continuous variables according to their distribution characteristics. When comparing engraftment outcomes stratified by the CD34+ cell dose cutoff of 3.5 × 10⁶/kg, the Mann-Whitney U test was applied because both neutrophil and platelet engraftment times did not conform to a normal distribution. Only axis scales and legends were adjusted to improve readability. No data were altered, and all original data points were used for statistical analyses. With slow engraftment as the dependent variable, univariate binary logistic regression was first performed to screen potential risk factors among clinical, laboratory, immune, and inflammatory indicators. Given the severely uneven distribution of patients receiving BEAM and BEAC regimens, the conditioning regimen was not included in the regression analyses to ensure model stability. The variable screening strategy for the multivariate logistic regression model was established as follows: A P-value threshold of 0.1 was set for baseline clinical variables; based on previous evidence and biological relevance, immune parameters related to lymphocyte subsets were preferentially enrolled with a cutoff of P < 0.2 to retain immunologically meaningful variables. This stratified screening strategy was designed to align with the research focus on T-cell immune imbalance. The threshold of P < 0.2 for immune variables helped preserve biologically relevant markers, whereas P < 0.1 for clinical variables reduced redundant factors. No severe multicollinearity was found in the final model. Multicollinearity analysis was conducted for candidate variables identified in univariate analysis. The variance inflation factor (VIF) was calculated to assess collinearity, and variables with VIF values exceeding 5 were excluded to avoid model overfitting and unstable parameter estimates. Although CD8+ T-cell percentage, CD4+ T-cell absolute count, and CD4/CD8 ratio are mathematically correlated, VIF analysis indicated no severe multicollinearity. Considering their distinct biological implications, all three indicators were retained in the analyses. A stepwise logistic regression model using the forward likelihood-ratio method was used to develop the multivariate model. Bootstrap resampling with 1000 iterations was performed for internal validation to evaluate model stability and goodness of fit. Ultimately, independent risk factors were integrated to construct a combined predictive model. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was calculated to compare and quantify the predictive performance of single indicators and the combined model. Sample adequacy for the five-predictor logistic regression model was assessed using the EPV rule. With 97 positive events of slow engraftment among 166 participants, the sample size provided sufficient statistical power and reduced the risk of model overfitting.

RESULTS

Baseline characteristics of patients

A total of 166 patients with lymphoma undergoing auto-HSCT between July 2022 and June 2025 were enrolled in this retrospective study. All participants were recruited from the Second Affiliated Hospital of Chongqing Medical University and Chongqing University Cancer Hospital, including 97 males and 69 females, with a median age of 53 years (range, 16-74 years). According to pathological classification, the cohort consisted of 105 cases of diffuse large B-cell lymphoma, 11 cases of angioimmunoblastic T-cell lymphoma, 6 cases of mantle cell lymphoma, 6 cases of peripheral T-cell lymphoma not otherwise specified, and 38 cases of other lymphoma subtypes. A definitive diagnosis was confirmed by pathological biopsy in all patients. Transplantation indications mainly included consolidation therapy following first-line treatment for high-risk lymphoma and relapsed or refractory lymphoma. Most patients received the BEAM conditioning regimen (bendamustine, etoposide, cytarabine, and melphalan) (n = 151), whereas a small proportion were treated with the BEAC regimen (bendamustine, cyclophosphamide, etoposide, and cytarabine) (n = 15). Hematopoietic stem cells were harvested from peripheral blood in all enrolled patients, with a median infused CD34+ cell count of 4.215 × 10⁶/kg. The median time to neutrophil engraftment was 10 days (range, 8-25 days), and the median time to platelet engraftment was 12 days (range, 5-51 days). Uniform supportive care was administered to all patients throughout the peritransplant period, including standardized antimicrobial prophylaxis, blood product transfusion support, and growth factor administration in accordance with institutional protocols. During the follow-up period, one patient failed to achieve sustained hematopoietic engraftment and ultimately died of cerebral hemorrhage. Based on median engraftment times, patients were stratified into two groups for subsequent comparative analysis: 97 patients in the slow engraftment group (neutrophil recovery > 10 days or platelet recovery > 12 days), and 69 patients in the early engraftment group (neutrophil recovery ≤ 10 days and platelet recovery ≤ 12 days).

Clinical factors between the two groups

All baseline and peritransplant clinical parameters were compared between the slow engraftment group (n = 97) and the early engraftment group (n = 69). Categorical variables were analyzed using Pearson’s χ² test. For continuous variables, the Mann-Whitney U test or independent-samples t-test was selected for intergroup comparisons based on data distribution, with all results presented in Table 1 and Figure 1. Regarding demographic and disease baseline characteristics, no significant differences were observed between the two groups in sex distribution, initial bone marrow involvement, baseline disease stage (stage IV vs non-stage IV), initial risk stratification (high-risk vs non-high-risk), or pretransplant disease remission status (all P > 0.05). Patients in the slow engraftment group were significantly older (Figure 1A) and had distinct baseline LDH (Figure 1B) and HB levels (Figure 1C), with all differences reaching statistical significance (P < 0.05). Regarding pretransplant laboratory indicators and stem cell characteristics, the change in body mass index (BMI) (pretransplant minus baseline) was higher in the slow engraftment group (P = 0.025, Figure 1D). Patients in the slow engraftment group received a significantly lower infused CD34+ cell count (P < 0.001, Figure 1E), accompanied by elevated pretransplant β2-MG levels (P = 0.016, Figure 1F). No remarkable differences in other laboratory markers were identified between the two cohorts (all P > 0.05).

Figure 1 Comparison of significantly different baseline and peritransplant clinical parameters between early and slow engraftment groups. A: Age at transplantation; B: Lactate dehydrogenase level at diagnosis; C: Hemoglobin level at diagnosis; D: Change in body mass index (BMI) (pretransplant BMI - BMI at diagnosis); E: CD34+ cell dose infused; F: Pretransplant beta 2-microglobulin level. ^aP < 0.05, ^bP < 0.01. LDH: Lactate dehydrogenase; HB: Hemoglobin; BMI: Body mass index; β2-MG: Beta 2-microglobulin.

Table 1 Baseline clinical and laboratory characteristics of patients stratified by early vs slow engraftment group after autologous hematopoietic stem cell transplantation, n (%).

Variable	Early engraftment group	Slow engraftment group	P value
Baseline characteristics
Age (years), median (IQR, Q1-Q3)	50 (38-56)	55 (47.5-61.5)	0.0011,a
Sex			0.3922
Male	43 (62.3)	54 (55.7)
Female	26 (37.7)	43 (44.3)
BM involvement at diagnosis			0.7742
Yes	11 (16.4)	16 (18.2)
No	56 (83.6)	72 (81.8)
Ann Arbor stage			0.1332
IV	33 (55.9)	58 (70.7)
Non-IV	26 (44.1)	24 (29.3)
Risk stratification			0.3842
High risk	15 (30.0)	26 (37.7)
Non-high risk	35 (70.0)	43 (62.3)
LDH at diagnosis (U/L), median (IQR, Q1-Q3)	228 (175.5-393.3)	292 (204-546)	0.0391,a
β2-MG at diagnosis (mg/L), median (IQR, Q1-Q3)	2.58 (1.9-4.2)	3.215 (2.3-4.4725)	0.1141
HB at diagnosis (g/L), median (IQR, Q1-Q3)	129 (111.5-136)	119 (102-133.75)	0.0431,a
WBC at diagnosis (× 109/L), median (IQR, Q1-Q3)	6.29 (4.83-8.03)	6.24 (4.72-7.8875)	0.5971
PLT at diagnosis (× 109/L), median (IQR, Q1-Q3)	228 (172.5-293)	239.5 (155-307.75)	0.8311
Transplant-related indicators
CD34+ cells (× 106/kg), median (IQR, Q1-Q3)	5.27 (3.93-8.645)	3.59 (2.64-5.54)	< 0.0011,a
Mononuclear cells (× 108/kg), median (IQR, Q1-Q3)	5.15 (3.44-7.055)	5.01 (3.7-7.64)	0.7291
Stem cell storage time (days), median (IQR, Q1-Q3)	46 (33-70)	50 (30-72.5)	0.9261
Pretransplant indicators
Pretransplant remission status			0.1742
CR	47 (68.1)	56 (57.7)
NR	22 (31.9)	41 (42.3)
Change in BMI (kg/m2), median (IQR, Q1-Q3)	0 (-0.535 to 0.335)	-0.35 (-1.7525 to 0.065)	0.0251,a
Pretransplant nutritional score, median (IQR, Q1-Q3)	2 (2-2)	2 (2-2)	0.1351
Pretransplant albumin (g/L), mean ± SD	27.08 ± 5.38	26.10 ± 5.55	0.2573
Pretransplant prealbumin (mg/L), median (IQR, Q1-Q3)	236 (203-265.32)	236 (214-267)	0.661
Pretransplant globulin (g/L), median (IQR, Q1-Q3)	39.8 (37.6-44.2)	40.92 (37-44.2)	0.7321
Pretransplant LDH (U/L), median (IQR, Q1-Q3)	203.5 (182-269.25)	221 (181-282.5)	0.5241
Pretransplant β2-MG (mg/L), median (IQR, Q1-Q3)	2.465 (1.9-3)	2.715 (2.2125-3.3375)	0.0161,a
Pretransplant HB (g/L), mean ± SD	105.9 ± 18.7	102.5 ± 17.5	0.2323
Pretransplant WBC (× 109/L), median (IQR, Q1-Q3)	3.96 (2.9-5.795)	3.83 (2.855-5.9275)	0.8841
Pretransplant PLT (× 109/L), median (IQR, Q1-Q3)	199 (148-253)	183 (131-215.75)	0.0721

¹Mann-Whitney U test.

²χ² test.

³Student’s t-test.

^aP < 0.05.

All lymphocyte subsets and inflammatory factors were measured at pretransplant baseline. Pretransplant nutritional score was assessed using the Nutritional Risk Screening 2002 criteria. Change in body mass index (BMI) =BMI at pretransplantation - BMI at diagnosis. IQR: Interquartile range; BM: Bone marrow; LDH: Lactate dehydrogenase; β2-MG: Beta 2-microglobulin; HB: Hemoglobin; WBC: White blood cell; PLT: Platelet; CR: Complete remission; NR: Non-complete remission; BMI: Body mass index.

Comparison of pretransplant lymphocyte subset distribution between the two groups

To identify baseline lymphocyte subsets associated with slow engraftment, we compared pretransplant peripheral blood lymphocyte subsets between the slow- and early-engraftment groups. Detailed results are presented in Table 2 and Figure 2. Significant immune imbalance was observed across cohorts in major T lymphocyte subsets. Compared with the early engraftment group, the slow engraftment group exhibited a higher proportion of CD8+ T lymphocytes (P = 0.017, Figure 2A) and a lower absolute count of CD4+ T lymphocytes (P = 0.041, Figure 2B), which collectively contributed to a reduced CD4/CD8 ratio (P = 0.048, Figure 2C). Concurrent abnormalities in these key parameters indicated that impaired T-cell homeostasis already existed in patients with slow engraftment before transplantation. This immune disturbance, characterized by CD4+ T-cell depletion and excessive CD8+ T-cell activation, may create an unfavorable bone marrow microenvironment, thereby delaying hematopoietic recovery following auto-HSCT. These findings provide core evidence for screening promising predictive biomarkers of slow engraftment after auto-HSCT.

Figure 2 Comparison of lymphocyte subsets with significant differences between early and slow engraftment groups. A: Percentage of CD8+ T cells; B: Absolute count of CD4+ T cells; C: CD4/CD8 ratio. Slow engraftment group exhibited elevated CD8+ T cell proportion, reduced CD4+ T cell count, and decreased CD4/CD8 ratio, indicating pretransplant T cell immune homeostasis imbalance. ^aP < 0.05.

Table 2 Comparison of core immune and inflammatory indicators, peritransplant clinical events between early and slow engraftment groups in lymphoma patients undergoing autologous stem cell transplantation.

Variable	Early engraftment group	Slow engraftment group	P value
CD8+ T cell (%), median (IQR, Q1-Q3)	41.68 (29.73-51.58)	47.47 (37.79-57.02)	0.0171,a
CD4+ T cell count (cells/μL), median (IQR, Q1-Q3)	315 (187-426.5)	234.37 (141.76-330.5)	0.0411,a
CD4/CD8 ratio, median (IQR, Q1-Q3)	0.95 (0.57-1.6)	0.79 (0.48-1.27)	0.0481,a
IL-2 (pg/mL), median (IQR, Q1-Q3)	0.74 (0.655-1.0275)	1.335 (0.8-1.75)	0.0031,a
IFN-γ (pg/mL), median (IQR, Q1-Q3)	1.275 (0.88-2.01)	2.715 (1.73-5.52)	0.0051,a
Days with fever during transplantation (days), median (IQR, Q1-Q3)	4 (2-8)	6 (3-11)	0.0061,a
Onset of neutropenia (days), median (IQR, Q1-Q3)	3 (2-4)	2 (1-4)	0.0331,a

¹Mann-Whitney U test.

^aP < 0.05.

All lymphocyte subsets and inflammatory factors were measured at pretransplant baseline. IQR: Interquartile range; IL-2: Interleukin-2; IFN-γ: Interferon-γ.

Inflammatory cytokines and peritransplant events

Further analysis of inflammatory cytokines and peritransplant clinical indicators (Table 2, Supplementary Table 1, Figure 3) showed that patients in the slow engraftment group had significantly elevated levels of nterleukin-2 (IL-2) and interferon-γ (IFN-γ) (P = 0.003 and P = 0.005, respectively; Figure 3A and B). This group also had a longer duration of fever during transplantation and an earlier onset of neutropenia (Figure 3C and D), with intergroup differences reaching statistical significance (P = 0.006 and P = 0.033). No significant differences in other cytokines were detected between the two groups (all P > 0.05). Collectively, elevated IL-2 and IFN-γ levels, prolonged peritransplant fever, and earlier onset of neutropenia were prominent in the slow engraftment group. These findings suggest that excessive inflammatory activation and adverse peritransplant events may jointly impair hematopoietic recovery after auto-HSCT.

Figure 3 Comparison of significantly different pretransplant inflammatory cytokines and peritransplant clinical events between early and slow engraftment groups. A and B: Pretransplant inflammatory cytokines (interleukin-2, interferon-γ); C and D: Peritransplant clinical events (neutropenia onset, fever days). ^aP < 0.05, ^bP < 0.01. IL-2: Interleukin-2; IFN-γ: Interferon-γ.

Differences in hematopoietic reconstitution according to distinct CD34+ cell doses

All 166 patients were stratified into the low-dose group (< 3.5 × 10⁶/kg, n = 57) and high-dose group (≥ 3.5 × 10⁶/kg, n = 109) based on infused CD34+ cell count. Differences in posttransplant hematopoietic reconstitution kinetics were then compared between the two cohorts (Table 3). Given the non-normal distribution of neutrophil and platelet engraftment times, the Mann-Whitney U test was applied for intergroup comparison. The median time to neutrophil engraftment was 10 days (range, 8-20 days) in the high CD34+ cell dose group, which was shorter than 11 days (range, 8-25 days) in the low-dose group (P = 0.002). For platelet recovery, the high-dose group achieved a median engraftment time of 11 days (range, 6-36 days) vs 12 days (range, 5-51 days) in the low-dose group, with a statistically significant difference (P = 0.016). Although the absolute differences in neutrophil and platelet engraftment times were modest, such disparities have critical clinical implications during the high-risk peritransplant window, when the risks of infection and hemorrhage are concentrated. Collectively, these data demonstrate that a higher infused CD34+ cell dose accelerates neutrophil and platelet reconstitution and effectively shortens the overall time to sustained hematopoietic engraftment after auto-HSCT.

Table 3 Association between CD34+ count and hematopoietic engraftment time.

	Total (n = 166)	< 3.5 × 106/kg (n = 57)	≥ 3.5 × 106/kg (n = 109)	P value
Neutrophil engraftment time (days), median (range)	10 (8-25)	11 (8-25)	10 (8-20)	0.0021,a
Platelet engraftment time (days), median (range)	12 (5-51)	12 (5-51)	11 (6-36)	0.0161,a

¹Mann-Whitney U test.

^aP < 0.05.

Univariate logistic regression analysis

To systematically screen for risk factors of slow engraftment following auto-HSCT, univariate binary logistic regression analysis was performed. The dependent variable was defined as the occurrence of slow engraftment. Multidimensional clinical variables were comprehensively included for pairwise analysis, including baseline demographic characteristics, initial disease features, pretransplant laboratory parameters, stem cell infusion indicators, nutritional status indexes, peripheral blood T-lymphocyte subsets, inflammatory cytokine levels, and early posttransplant clinical events. Detailed statistical results are presented in Supplementary Table 2.

Univariate analysis revealed that multiple clinical and immune parameters were significantly associated with slow engraftment. Regarding baseline demographic and disease characteristics, advanced age was a risk factor for slow engraftment, with each 1-year increase in age associated with an increased risk [odds ratio (OR) = 1.044, 95% confidence interval (CI): 1.016-1.073, P = 0.002]. Regarding transplantation-related parameters, patients receiving a low CD34+ cell infusion dose (< 3.5 × 10⁶/kg) exhibited a 3.128-fold higher risk of slow engraftment (OR = 3.128, 95%CI: 1.537-6.366, P = 0.002). Additionally, a per-unit increase in BMI change (a higher BMI value before transplantation relative to the initial diagnosis, indicating less weight loss or net weight gain) was linked to a reduced risk of slow engraftment (OR = 0.825, 95%CI: 0.682-0.998, P = 0.048).

Univariate analyses of baseline biochemical and immune characteristics demonstrated that elevated pretransplant β2-MG acted as a risk factor for slow engraftment (OR = 1.663, 95%CI: 1.066-2.595, P = 0.025). Regarding pretransplant immune cell profiles, a higher proportion of CD8+ T lymphocytes was correlated with an increased risk of slow engraftment (OR = 1.027, 95%CI: 1.007-1.048, P = 0.008). In terms of inflammatory cytokines, elevated pretransplant IL-2 levels (above the population tertile cutoff of 0.83 pg/mL) were significantly associated with slow engraftment, conferring a more than five-fold increase in risk (OR = 5.833, 95%CI: 1.200-28.366, P = 0.029). With respect to early posttransplant clinical events, a longer duration of peritransplant fever corresponded to a higher likelihood of slow engraftment (OR = 1.117, 95%CI: 1.038-1.202, P = 0.003). Moreover, an earlier onset of neutropenia was associated with a greater risk of slow engraftment (OR = 0.825, 95%CI: 0.690-0.986, P = 0.034). All other enrolled variables failed to exhibit a significant univariate correlation with slow engraftment (all P > 0.05).

In summary, this univariate logistic regression analysis identified multiple clinical, immune, and inflammatory indicators significantly associated with slow engraftment after auto-HSCT. All candidate variables intended for multivariate analysis had VIF values below 5, indicating no obvious multicollinearity and supporting valid inclusion in the subsequent regression model. These variables were further incorporated into the multivariate regression model to adjust for confounding factors and screen for independent risk factors for slow engraftment.

Multivariate logistic regression analysis

A multivariate binary logistic regression model was constructed to further identify independent risk factors for slow engraftment. The variable inclusion criteria were established as follows: Conventional baseline clinical indicators with univariate P < 0.1 were initially included. Based on previous research evidence and biological relevance, immune indicators related to lymphocyte subsets were comprehensively evaluated and included using a liberal inclusion threshold of P < 0.2 to retain immune parameters with potential biological implications to the greatest extent. After adjustment for confounding effects across variables, several independent influencing factors were ultimately determined (Table 4).

Table 4 Univariate and multivariate logistic regression analyses of risk factors for slow hematopoietic engraftment after autologous hematopoietic stem cell transplantation.

Variable	Univariate analysis			Multivariate analysis
	OR	95%CI	P value	OR	95%CI	P value
Age (years)	1.044	1.016-1.073	0.002	1.044	1.009-1.081	0.015a
HB at diagnosis (g/L)	0.986	0.971-1.001	0.062	-	-	-
CD34+ cells (× 106/kg) (< 3.5 × 106/kg vs ≥ 3.5 × 106/kg)	3.128	1.537-6.366	0.002	2.633	1.119-6.196	0.027a
Change in BMI (kg/m2)	0.825	0.682-0.998	0.048	-	-	-
Pretransplant β2-MG (mg/L)	1.663	1.066-2.595	0.025	-	-	-
Pretransplant PLT (× 109/L)	0.997	0.993-1	0.066	-	-	-
CD8+ T cell (%)	1.027	1.007-1.048	0.008	1.039	1.010-1.070	0.009a
CD4+ T cell (%)	0.979	0.957-1.002	0.077	-	-	-
CD4/CD8 ratio	0.699	0.477-1.025	0.066	-	-	-
CD4+ T cell count (cells/μL)	0.999	0.997-1	0.136	0.997	0.995-1.000	0.038a
Days with fever during transplantation (days)	1.117	1.038-1.202	0.003	1.119	1.023-1.224	0.014a
Onset of neutropenia (days)	0.825	0.690-0.986	0.034	-	-	-

^aP < 0.05.

Change in body mass index (BMI) =BMI at pretransplantation - BMI at diagnosis. All lymphocyte subsets and inflammatory factors were measured at pretransplant baseline. Multivariate model inclusion criteria: Standard baseline covariates with univariate P < 0.1; for lymphocyte subset variables (core exposure of interest), univariate P < 0.2 was used for entry. OR: Odds ratio; CI: Confidence interval; HB: Hemoglobin; BMI: Body mass index; β2-MG: Beta 2-microglobulin; PLT: Platelet.

Multivariate analysis confirmed that advanced age (OR = 1.044, 95%CI: 1.009-1.081, P = 0.015), insufficient infused CD34+ cell dose (< 3.5 × 10⁶/kg) (OR = 2.633, 95%CI: 1.119-6.196, P = 0.027), elevated baseline proportion of CD8+ T lymphocytes (OR = 1.039, 95%CI: 1.010-1.070, P = 0.009), decreased absolute count of CD4+ T lymphocytes (OR = 0.997, 95%CI: 0.995-1.000, P = 0.038), and prolonged duration of peritransplant fever (OR = 1.119, 95%CI: 1.023-1.224, P = 0.014) were independent risk factors for slow engraftment after auto-HSCT. After adjustment for confounders, including age and CD34+ cell infusion dose, in the multivariate logistic regression model, these immune parameters remained statistically significant. This finding indicates that immune disturbance is an independent risk factor contributing to delayed hematopoietic reconstitution. No residual variables retained an independent statistical association with slow engraftment (all P > 0.05). Univariate analysis revealed no statistical significance for IFN-γ. Moreover, because of the wide 95%CI (1.200-28.366) and the relatively small sample size (n = 38), the accuracy of effect estimation was insufficient, and this indicator was therefore not included in the multivariate model (Supplementary Tables 1 and 2). To quantify the clinical predictive value of relevant indicators, ROC curves were plotted for each independent risk factor and the combined multivariate model (Figure 4). Among single indicators, pretransplant age (AUC = 0.664), CD8+ T-lymphocyte percentage (AUC = 0.623), and days of peritransplant fever (AUC = 0.600) exhibited moderate predictive performance. Notably, the combined model integrating all independent risk factors yielded an AUC of 0.780, which was superior to that of any single variable alone. This combined model demonstrated satisfactory discriminatory capability and reliable clinical predictive efficiency for the risk of posttransplant slow engraftment. Internal validation via bootstrap resampling revealed no substantial deviations in regression coefficients, OR values, or statistical effects of model variables, with stable overall model fit. These findings confirm that the established multivariate logistic regression model is robust and generalizable.

Figure 4 Receiver operating characteristic curves. A: Receiver operating characteristic curves for predictive performance of risk factors for slow hematopoietic engraftment after autologous hematopoietic stem cell transplantation; B: Receiver operating characteristic curves of the multivariate combined model for predicting slow hematopoietic engraftment after autologous hematopoietic stem cell transplantation. ROC: Receiver operating characteristic; AUC: Area under the curve.

DISCUSSION

Delayed graft engraftment after allogeneic hematopoietic stem cell transplantation has been well documented[19,21,22]. However, large- scale and systematic clinical evidence focusing on slow engraftment following auto-HSCT remains limited, especially regarding the regulatory impact of pretransplant immune status on hematopoietic engraftment. This retrospective study enrolled 166 patients with lymphoma undergoing auto-HSCT. By comprehensively analyzing baseline clinical data, stem cell infusion parameters, inflammatory cytokines, and lymphocyte subset profiles, we identified advanced age, low CD34+ cell infusion dose, pretransplant T-cell homeostasis imbalance, and prolonged peritransplant fever duration as independent risk factors for slow engraftment. The combined predictive model constructed from these core indicators demonstrated satisfactory discriminatory power (AUC = 0.78), enabling effective screening of high-risk individuals in clinical practice. Impaired posttransplant engraftment increases the risk of severe adverse clinical outcomes. In the present cohort, one patient with primary engraftment failure ultimately died of cerebral hemorrhage. These findings highlight that early identification of high-risk factors and optimized peritransplant management are critical for reducing transplant-related complications and ensuring transplant safety.

First, the present study demonstrated that an insufficient CD34+ cell infusion dose serves as a powerful independent risk factor for slow engraftment after auto-HSCT. Patients in the high-dose CD34+ group (≥ 3.5 × 10⁶/kg) exhibited significantly faster neutrophil and platelet engraftment than those in the low-dose group (< 3.5 × 10⁶/kg). These findings are consistent with results from previous studies[23-25], further validating the well-recognized role of CD34+ cell count as a core determinant of hematopoietic engraftment. Adequate stem cell infusion is an essential prerequisite for uneventful hematopoietic recovery following transplantation[9,26]. Consistently, multivariate regression analysis confirmed that low CD34+ cell dose remained an independent adverse predictor that could directly modulate posttransplant hematopoietic reconstitution kinetics. Accordingly, optimizing CD34+ cell collection and infusion dosage may reduce the incidence of slow engraftment and improve clinical outcomes. Second, multiple independent clinical risk factors for hematopoietic recovery were identified in the current cohort. Advanced age and prolonged peritransplant fever were closely associated with slow engraftment after auto-HSCT. Advanced age independently increased the risk of impaired hematopoietic recovery, which was consistent with earlier evidence[27]. Prolonged peritransplant fever is commonly accompanied by infection, thereby disrupting hematopoietic recovery in the bone marrow. As a non-modifiable risk factor, advanced age warrants careful pretransplant assessment of hematopoietic reserve and physical tolerance in elderly recipients, as well as individualized conditioning regimens and supportive care strategies. For peritransplant fever, intensified early infection screening and targeted anti-inflammatory management are recommended to constrain excessive inflammation, curtail inflammation-mediated myelosuppression, and ultimately facilitate timely and sustained hematopoietic reconstitution.

Most importantly, the present study confirmed that pretransplant disturbance of peripheral blood T lymphocyte subset homeostasis acts as an independent risk factor for slow engraftment in patients with lymphoma undergoing auto-HSCT, characterized by an elevated proportion of CD8+ T lymphocytes and a reduced absolute count of CD4+ T lymphocytes. Previous studies have demonstrated that aberrantly activated T lymphocytes can mediate hematopoietic suppression, impair the bone marrow microenvironment, and participate in the pathogenesis of multiple disorders with compromised bone marrow hematopoiesis, including aplastic anemia[14,28,29]. Restoration of hematopoietic function following immunosuppressive therapy provides the most direct evidence for the immune-driven mechanism underlying this disease[13,30,31]. In the current study, the distinct pattern of T-cell subset dysregulation observed in the slow engraftment group suggests a shared immune background of hematopoietic suppression with immune-mediated bone marrow injury disorders, which is consistent with previous findings[15,32]. The combined effects of excessive CD8+ T-cell activation and insufficient CD4+ T cells contribute to an immune microenvironment that impedes hematopoietic stem cell engraftment and proliferation, thereby delaying hematopoietic reconstitution. CD4+ T cells are essential for maintaining systemic immune homeostasis and regulating inflammatory responses; their depletion disrupts immune equilibrium and exacerbates aberrant inflammatory reactions[33,34]. Accumulating evidence indicates that CD4+ T cell dysfunction, aberrant CD8+ T cell activation, and subsequent pro-inflammatory cytokine disturbances act synergistically to suppress the activity of hematopoietic stem and progenitor cells[35-37]. Solomou et al[13] demonstrated that a reduction in CD4+ CD25+FOXP3+ regulatory T cells increases the secretion of pro-inflammatory cytokines such as IFN-γ, which synergizes with CD8+ T cells to disrupt the hematopoietic microenvironment[37]. In the early setting of allogeneic hematopoietic stem cell transplantation, conventional T-cell depletion strategies have been shown to elevate the risk of engraftment failure[18,38]. Collectively, these lines of evidence strongly indicate that T-cell immunity plays a key role in regulating hematopoietic reconstitution. Together with our results, patients with slow engraftment presented a distinct immune imbalance characterized by decreased absolute CD4+ T-cell count and elevated proportion of CD8+ T cells, further indicating that pretransplant T-cell dysregulation serves as a critical determinant of hematopoietic recovery following autologous transplantation. Immune parameters remained statistically significant after adjustment for confounding variables, including age and CD34+ cell dose, confirming their independent impact on slow engraftment. In addition, the slow engraftment group exhibited elevated levels of IL-2 and IFN-γ. Excessive production of these pro-inflammatory cytokines can directly suppress hematopoietic function and disrupt the bone marrow microenvironment[39,40], thereby exerting synergistic pathogenic effects together with T-cell immune dysregulation. No significant intergroup differences in functional CD4+ T-cell subsets were observed in the present study, which may be attributable to the limited sample size. Further validation with an expanded cohort is warranted in future investigations. Overall, the major strength of this study is the confirmation that pretransplant T-cell immune imbalance, characterized by an elevated proportion of CD8+ T cells and a reduced absolute CD4+ T-cell count, independently increases the risk of slow engraftment. This adverse effect is not simply confounded by stem cell dose, age, or other baseline factors. These findings highlight the clinical value of routine pretransplant peripheral blood T-cell subset evaluation. Moreover, they provide novel theoretical evidence for high-risk stratification before auto-HSCT, rational optimization of peritransplant immune modulation, and individualized intervention strategies in lymphoma recipients. Univariate logistic regression analysis showed that when stratified by the IL-2 tertile cutoff of 0.83 pg/mL, IL-2 levels above 0.83 pg/mL were significantly associated with slow engraftment (OR = 5.833, 95%CI: 1.200-28.366, P = 0.029). However, because of the wide 95%CI and the relatively small sample size (n = 38), the accuracy of effect estimation was insufficient, and this indicator was therefore not included in the multivariate model. Elevated IL-2 may reflect excessive T cell immune activation, whereas its independent pathogenic role remains to be validated in larger cohorts[39,41-43]. In addition, the proportion of activated CD4+ T cells (CD4+CD25+, %) was excluded from the multivariate regression analysis owing to limited sample availability.

Compared with similar previous studies, the present study incorporated key clinical indicators, including age and the infused CD34+ cell dose. Notably, immune homeostasis parameters, including lymphocyte subsets, as well as inflammatory cytokines such as IL-2 and IFN-γ, were included for the first time. A combined predictive model (AUC = 0.780) was further developed based on these independent risk factors. This integrated approach improves the ability to identify the risk of slow engraftment after auto-HSCT, overcomes the limited predictive accuracy and clinical applicability of single indicators reported in previous studies, and provides a more practical assessment tool for the early clinical screening of high-risk patients. This study has several limitations. First, it was a retrospective, dual-center observational study conducted at two tertiary hospitals in Chongqing, with regional restrictions and no external multicenter validation. Because of the retrospective design, unmeasured confounders such as conditioning regimens and peritransplant supportive care could not be fully balanced, and selection bias was inevitable. Second, the overall sample size was limited. Only single-time-point pretransplant immune and inflammatory markers were detected, without dynamic monitoring data during the peritransplant period. In addition, the follow-up duration was short, and long-term survival data were incomplete. Therefore, the impacts of slow engraftment on tumor recurrence, long-term survival, and other long-term outcomes remained unclear, as did the prognostic benefit of early clinical intervention. Future large-scale, multicenter prospective cohorts are required for external validation. Extended follow-up is needed to clarify the long-term prognostic impact and to evaluate targeted interventions. Serial dynamic immune monitoring and mechanistic experiments should also be combined to clarify how immune dysregulation disrupts the bone marrow microenvironment, thereby providing evidence for optimized immune-targeted therapeutic strategies.

CONCLUSION

Based on the retrospective analysis of clinical data from 166 patients with lymphoma undergoing auto-HSCT, this study identified advanced age, low infused CD34+ cell dose, prolonged peritransplant fever, and pretransplant T-cell immune homeostasis imbalance as independent risk factors for slow engraftment after auto-HSCT. Compared with conventional clinical indicators, the addition of lymphocyte subsets and inflammation-related immune markers further improves the risk assessment system for poor hematopoietic engraftment. The combined prediction model constructed from multiple independent influencing factors exhibits good discriminatory performance, enabling effective early identification of patients at high risk of posttransplant slow engraftment. These findings suggest that, in clinical practice, aside from optimizing stem cell collection dose and preventing peritransplant infection, greater attention should be paid to screening peripheral blood T-cell subsets for pretransplant immune status. This provides reliable clinical evidence for peritransplant risk stratification, individualized intervention, and optimized transplant management in lymphoma patients.