Influence of time-of-day of graft infusion on allogeneic hematopoietic stem cell transplantation outcomes: a validation cohort study

Abstract

The biological clock plays a central role in hematopoiesis and immune regulation, making circadian rhythms an increasingly important factor in immunotherapies and cellular therapies such as allogeneic hematopoietic stem cell transplantation (allo-HSCT). Previous work by Hou et al suggested that afternoon graft infusions were associated with a higher risk of acute graft-versus-host disease and worse survival, highlighting a potential role for infusion timing in shaping transplant outcomes.

To further investigate this, we performed a retrospective analysis of 368 patients who underwent allo-HSCT at Geneva University Hospitals between 2015 and 2024. Observing that infusion timing patterns at our center differed from those reported by Hou et al, we applied a data-driven approach using receiver operating characteristic analysis, which identified 11:17 as the optimal cut-off for overall survival.

Stratification using this threshold revealed that patients receiving grafts before 11:00 had significantly improved 2-year overall survival and lower non-relapse mortality compared with those infused later in the day, with no differences in relapse or engraftment rates.

These findings suggest that infusion timing may be an important, under-recognized factor influencing allo-HSCT outcomes. Prospective clinical trials are needed to confirm these observations and explore their applicability across different clinical contexts.

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is a well-established curative therapy for a broad range of hematologic disorders, yet it remains associated with significant morbidity and mortality. In addition to established determinants of transplant success—such as donor type, human leukocyte antigen matching, conditioning regimens, graft composition and graft-versus-host disease (GvHD) prophylaxis—an emerging body of research highlights a novel and intriguing potential modulator: circadian rhythms.^{1 2} The circadian clock relies on recurring environmental signals—such as light exposure—to regulate daily, ~24-hour cycles in humans in various processes such as sleep, hormone release, metabolism, body temperature, and immune activity. The relevance of the circadian clock on immune responses is an area of intense investigation,³ and increasing evidence indicates its impact on antitumor immune responses.^{4 5} We have recently shown that circadian rhythms influence the efficacy of antitumor immunotherapies, including checkpoint inhibitors and chimeric antigen receptor T cells efficacy in murine models of cancer.⁶ Importantly, these mechanistic preclinical studies appear to also impact patient outcomes in the clinical setting.7,10 Very recently, Hou and colleagues¹ revealed that the timing of stem cell infusion might potentially influence the outcome after allo-HSCT. In murine models, transplantation performed 2 hours after dark onset—corresponding to the active phase in mice—led to significantly lower acute GvHD scores, reduced weight loss, and improved survival compared with transplantation at 5 hours after light onset. These clinical benefits were associated with reduced levels of multiple inflammatory cytokines and chemokines, including interleukin-2, interferon-γ, and the T-cell motility–associated mediators CXCL10, CCL2, CCL3, CCL5, and CCL7. In addition, the nocturnal group exhibited lower percentages and absolute numbers of CD38⁺ activated CD4⁺ T cells 24 hours after transplantation compared with the daytime group. In clinical cohorts, they showed that earlier time-of-day infusion (<14:00) of peripheral blood stem cell (PBSC) grafts was associated with significantly lower rates of grade II-IV and III-IV acute GvHD, as well as improved GvHD-free, relapse-free survival (GRFS), compared with later time-of-day infusion (>14:00). Their findings suggest that the recipient’s immune environment differs depending on the time of day and that the impact of graft timing infusion is governed by the recipient’s circadian rhythm. Recognizing and harnessing circadian biology might offer a simple yet surprisingly powerful opportunity to optimize HSCT outcomes with minimal additional intervention. However, before such provocative and potentially practice-changing results can be applied to the clinics, external validation across centers and countries is required.

To validate the findings of Hou et al, we analyzed a retrospective cohort of 368 patients who underwent allo-HSCT at our institution, the Geneva University Hospitals, in Geneva, Switzerland, between 2015 and 2024. The analysis was restricted to patients undergoing their first transplantation from PBSC as graft source. Patients receiving bone marrow and ex vivo manipulated PBSC were excluded from the analysis. Patients’ characteristics are summarized in table 1. We first stratified patients using the same 14:00 cut-off employed by Hou and colleagues and found no significant difference in 2-year overall survival (OS), progression-free survival (PFS) and GvHD-free, GRFS (online supplemental figure 1A). Similarly, we observed no significant difference in the cumulative incidence of all grade, grade II-IV, grade III-IV acute GvHD (online supplemental figure 1B) and in the 2-year cumulative incidence of non-relapse mortality

Table 1. Patients’ characteristics.

	Whole cohort	Until 11:00	After 11:00
Characteristic	n=368	n=63	n=305	p value
__Age__				0.5
Median (range)	58 (48–66)	56 (43–64)	58 (48–67)
__Sex__				0.6
Female	135 (37%)	21 (33%)	114 (37%)
Male	233 (63%)	42 (67%)	191 (63%)
__Disease__				0.3
AL	6 (1.6%)	1 (1.6%)	5 (1.6%)
ALL	30 (8.2%)	8 (13%)	22 (7.2%)
AML	166 (45%)	24 (38%)	142 (47%)
CLL	6 (1.6%)	0 (0%)	6 (2.0%)
CML	8 (2.2%)	1 (1.6%)	7 (2.3%)
MDS	60 (16%)	9 (14%)	51 (17%)
MPN	17 (4.6%)	2 (3.2%)	15 (4.9%)
MDS/MPN	16 (4.3%)	2 (3.2%)	14 (4.6%)
Lymphoma	47 (13%)	14 (22%)	33 (11%)
Myeloma	12 (3.3%)	2 (3.2%)	10 (3.3%)
__DRI__				>0.9
Low	15 (4.1%)	3 (4.8%)	12 (3.9%)
Intermediate	253 (69%)	42 (67%)	211 (69%)
High	86 (23%)	16 (25%)	70 (23%)
Very high	14 (3.8%)	2 (3.2%)	12 (3.9%)
__Donor type__				0.3
Haploidentical	89 (24%)	16 (25%)	73 (24%)
HLA-identical sibling	72 (20%)	18 (29%)	54 (18%)
HLA-matched unrelated donor	180 (49%)	25 (40%)	155 (51%)
HLA-mismatched relative	1 (0.3%)	0 (0%)	1 (0.3%)
HLA-mismatched unrelated donor	26 (7.1%)	4 (6.3%)	22 (7.2%)
__Congelation__				<0.001
Fresh	231 (63%)	21 (33%)	210 (69%)
Frozen	137 (37%)	42 (67%)	95 (31%)
__Conditioning_MAC_RIC__				0.9
MAC	94 (26%)	15 (24%)	79 (26%)
RIC	274 (74%)	48 (76%)	226 (74%)
__TBI_yes vs no__	126 (34%)	23 (37%)	103 (34%)	0.8
__T_cell_depleted__				>0.9
ATG	244 (66%)	42 (67%)	202 (66%)
No	124 (34%)	21 (33%)	103 (34%)
__CD3_E6_per_kg__	235 (175–315)	231 (184–344)	236 (174–308)	0.5
__CD34_E6_per_kg__	6.60 (5.50–7.30)	6.30 (5.20–7.40)	6.60 (5.50–7.30)	0.3
__CMV serostatus of recipient__				>0.9
Negative	142 (39%)	25 (40%)	117 (38%)
Positive	226 (61%)	38 (60%)	188 (62%)
__CMV serostatus of donor__				0.5
Negative	170 (46%)	26 (41%)	144 (47%)
Positive	198 (54%)	37 (59%)	161 (53%)

AL, acute leukemia; ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; ATG, anti-thymocyte globulin; CLL, chronic lymphocytic leukemia; CML, chronic myeloid leukemia; CMV, cytomegalovirus; DRI, Disease Risk Index; HLA, human leukocyte antigen; MAC, myeloablative conditioning; MDS, myelodysplastic syndrome; MPN, myeloproliferative neoplasia; RIC, reduced-intensity conditioning; TBI, total body irradiation.

(NRM) and relapse between patients transplanted before or after 14:00 (online supplemental figure 1C). These findings contrast with those reported by Hou et al, suggesting that the circadian effect or cut-off timing may not apply uniformly across centers, cities or continents. We hypothesized that the discrepancy might stem from different practices in terms of time of transplantation among centers. Patients included in the Hou et al analysis displayed a slightly positively skewed distribution displaying a peak and a median at 14:00 cut-off selected for their analysis (online supplemental figure 1D). Conversely, the data from our center displayed a bimodal distribution with a first peak at 11:00 and a second one at 14:00 (figure 1A). We hypothesized that such a difference might account for the absence of circadian impact observed in our cohort and the requirement to define an adapted cut-off time for our center. We therefore undertook a data-driven approach to define the optimal time-of-day cut-off and performed a receiver operating characteristic analysis using OS as the endpoint. This analysis identified 11:17 as the optimal threshold for stratification for our cohort (AUC=0.53; sensitivity=0.90; specificity=0.23; online supplemental figure 2). We re-stratified the patients into two groups based on the start time of the infusion: before 11:00 and after 11:00. Comparison of the two groups revealed similar clinical and transplant characteristics, with the only notable difference in the proportion of patients receiving fresh or frozen grafts, frozen grafts being more frequently used in patients transplanted before 11:00 (42 (67%)) then after 11:00 (95 (31%); p<0.001). We then analyzed patients’ outcomes based on this new data-driven cut-off. We observed a significant improvement in 2-year OS among patients who received their infusion before 11:00 a.m (83%, 95% CI 73% to 93%) compared with patients transplanted after 11:00 (65%, 95% CI 60% to 71%; p=0.0078; figure 1B). This was associated with a significant reduction in NRM in patients transplanted before 11:00 (3.2%, 95% CI 0.6% to 10%) compared with patients transplanted after 11:00 (13%, 95% CI 9% to 17%; p=0.02; figure 1C, left panel) while no difference was observed between the two groups in terms of relapse (figure 1C, right panel). Regarding the risk of acute GvHD, we observed a trend not reaching statistically significant differences toward the reduction of all grades, grade II-IV and grade III-IV acute GvHD in patients transplanted before 11:00 compared with patients transplanted later in the day (online supplemental figure 3A–C). No differences were observed in time to neutrophil engraftment (online supplemental figure 3D). Given the difference in terms of congelation status of the graft between the two groups, we performed a multivariable analysis to account for this important variable. The association between later graft administration and reduced OS remained significant (HR, 2.29, 95% CI 1.18 to 4.45; p=0.015) in multivariable regression analyses including the graft congelation status as covariate (figure 1D). After multivariable regression analysis a trend not reaching statistically significant difference was observed for higher NRM in the group transplanted after 11:00 (HR 4.11, 95% CI 0.96 to 17.56; p=0.056; figure 1E, left panel). In agreement with the univariable analysis, no impact of time of day or graft congelation status was observed on risk of relapse (figure 1E, right panel). Taken together, our findings support the central premise proposed by Hou et al that circadian timing of stem cell infusion might influence transplant outcomes but highlight some important differences. The inability to replicate the results reported by Hou et al using the same 14:00 cut-off underscores the need for context-specific analyses. Our identification of 11:00 as an optimal cut-off reflects differences in institutional practices that might impact the potential translation of chronotherapy in the allo-HSCT field and that should be considered in the design of future clinical trials. Moreover, our data-driven cut-off based on OS as an endpoint did not detect any significant difference in GvHD incidence among groups, although a trend was observed. The lack of a detectable difference in GvHD incidence is most likely attributable to the limited number of patients transplanted before 11:00, which resulted in an underpowered analysis, but may also suggest that the impact of time of day on graft infusion might go beyond the risk of GvHD incidence.

Figure 1. Impact of time of day of graft infusion in allogeneic HSCT outcome. (A) Histogram of stem cell infusion start times distribution in our Geneva cohort. (B) Kaplan-Meier curves showing 2-year OS. (C) Cumulative incidence of NRM and relapse in patients infused before or after 11:00. (D) Multivariable analysis of OS performed using the Cox regression model. (E) Multivariable analysis of NRM and relapse performed using the Fine and Gray model. HSCT, hematopoietic stem cell transplantation; NRM, non-relapse mortality; OS, overall survival; ROC, receiver operating characteristic.

Two currently phase 3 randomized clinical trials are ongoing comparing patients undergoing transplantation between 11:30 and 12:30 or between 15:30 and 17:30 in the context of allo-HSCT for malignant (NCT06294678) or non-malignant (NCT06294691) hematological diseases. We believe that our data will contribute to the interpretation of the results of these trials, which should take into account the potential heterogeneity in time windows that might apply to different centers, countries and geographic regions.

In conclusion, our study supports the emerging importance of circadian timing in allo-HSCT and extends previous work by Hou and colleagues. Currently ongoing and future clinical trials will reveal if a low-cost intervention such as scheduling stem cell infusions earlier in the day can significantly improve transplant outcomes.