Time-of-day of first checkpoint inhibitor dose influences clinical outcomes and immune responses in hepatocellular carcinoma

Abstract

Background

Although immune checkpoint inhibitors (ICIs) have long half-lives, preclinical and retrospective clinical studies across multiple tumor types suggest that the time-of-day of ICI infusion may influence therapeutic efficacy by aligning initial drug exposure with circadian peaks in T-cell responsiveness. The immunological basis of this phenomenon and its clinical relevance in hepatocellular carcinoma (HCC) remains unknown.

Methods

We followed patients with advanced HCC receiving ICI therapy at Johns Hopkins from 2021 to 2025, classifying them into a morning (first treatment before 12:00 hours) or afternoon (first treatment after 12:00 hours) group. We assessed clinical outcomes and compared immunological responses from baseline to early-on-treatment by profiling peripheral blood mononuclear cells using cytometry by time-of-flight and plasma cytokines using a 39-plex Luminex assay.

Results

Our cohort included 84 patients, 39 of whom received their first infusion in the morning. There were no statistically significant differences in baseline demographic or clinical characteristics between patients initiating therapy in the morning versus afternoon. The morning group had superior progression-free survival (multivariable HR 0.50, 95% CI 0.30 to 0.84, p<0.01) and higher odds of treatment response (multivariable OR 3.26, 95% CI 1.08 to 10.90, p<0.05), with no significant increase in immune-related adverse events. The timing of subsequent infusions after the first dose had no impact on outcomes. Immunological responses diverged after the initial dose, with morning-treated patients showing reduced interleukin (IL)-6 levels (p<0.01) and greater expansion of cytotoxic central memory CD8+ T cells (p=0.01) as well as cytotoxic effector and effector memory CD8+ T cells (p=0.06).

Conclusions

Morning first-dose infusion of ICIs in HCC was associated with improved clinical outcomes and distinct immune responses, including reduced IL-6 signaling and expansion of cytotoxic central memory CD8+ T cells. These findings suggest that the timing of the initial infusion can imprint an immunological program that shapes subsequent antitumor immunity, providing a mechanistic rationale for strategically scheduling ICI administration.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• It has been suggested in most tumor types that early time-of-day infusions are associated with superior survival and response rates.

WHAT THIS STUDY ADDS

• This study demonstrates that the timing of the first immune checkpoint inhibitor (ICI) infusion in patients with hepatocellular carcinoma (HCC) imprints a distinct early immunological program including reduced interleukin-6 levels and a greater expansion of cytotoxic central memory CD8+ T cells, in association with superior clinical outcomes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study provides crucial mechanistic and clinical evidence to support consideration of shifting toward morning baseline infusions while informing future randomized trial design for improved ICI efficacy in HCC.

Introduction

Immune checkpoint inhibitors (ICIs) have transformed the management of many cancers, including hepatocellular carcinoma (HCC), by restoring antitumor T-cell activity and improving survival.^{1 2} These agents are monoclonal antibodies that have half-lives measured in weeks, and thus, it might be expected that the precise time of infusion would have little to no impact on efficacy.³ Nevertheless, an accumulating body of retrospective clinical evidence across multiple tumor types suggests that patients treated with ICIs earlier in the day experience superior outcomes compared with those treated later.4,11 In addition, the clinical benefit of early day infusions has recently been shown retrospectively in chimeric antigen receptor T-cell (CAR-T) therapy in B-cell lymphomas.¹² Although such studies attempt to control for confounding variables, it remains possible that morning and afternoon patients differ systematically in ways that affect prognosis. Recently, a prospective, randomized clinical trial also demonstrated improved outcomes in patients who received their first four ICI infusions prior to 15:00 hours.¹³ However, at the time of our current report, public concerns have been raised regarding the study design and findings in this study, and thus, the significance of time-of-day treatment on ICI efficacy remains unsettled.

The proposed biological rationale of this time-of-day phenomenon is that early infusion aligns initial drug exposure with circadian peaks in T-cell responsiveness.¹⁴ T cells, like many immune cell populations, follow a circadian rhythm in their trafficking, activation thresholds, cytokine secretion, and effector function.⁹ For example, both murine and human studies have shown daily oscillations in T-cell proliferation, migration into lymphoid tissues, and expression of key homing and activation molecules.¹⁵ These rhythms are driven by cell-intrinsic clock genes such as BMAL1 and PER2 as well as systemic cues including hormones, body temperature, and light-dark cycles.16,18 In addition, the strength of the CD8+ T cell response to antigen presentation varies depending on time-of-day and CD8+ T cell’s innate circadian clocks.¹⁹ The CD4+ and CD8+ T-cell pool also varies in size depending on time-of-day, with naïve and central memory T-cell populations peaking at night and nadiring in the early afternoon.²⁰ As a result, the state of T cells at the moment of initial checkpoint receptor engagement may have profound downstream consequences for the trajectory of the antitumor immune response. Thus, we hypothesize that the timing of initial ICI administration may influence whether T cells enter a more proliferative, cytotoxic program or a less favorable immunological state, providing a mechanistic rationale for the observed clinical benefit of morning ICI infusion.

Despite these emerging insights, important questions remain. First, the clinical impact of infusion timing in HCC has not been well characterized, and it is unclear whether time-of-day affects efficacy and the risk of immune-related adverse events (irAEs). Previous studies have also stratified cohorts based on proportions of early infusions and have varying definitions for early versus late infusions. However, the importance of the first infusion itself on clinical outcomes and early immune activation has not been established. Finally, the immunological consequences of morning versus afternoon ICI administration are largely unexplored. In particular, it is unknown how time of day influences systemic cytokine responses, T-cell expansion, and the differentiation or activation of key T-cell subsets, all of which may underlie differences in clinical outcomes. Elucidating mechanistic underpinnings of these clinical effects is critical to substantiate previous studies, guide optimal timing of therapy, and harness circadian biology for improved immunotherapy strategies.

Methods

Study design and treatment timing definitions

Our study includes adult (aged ≥18 years) patients with advanced/unresectable HCC who were initiated on standard of care ICI-based therapies at Johns Hopkins from May 2021 to May 2025. We prospectively collected clinical data and serial peripheral blood samples from patient at baseline and early-on-treatment time points (month 1 or month 2). Infusion times were retrospectively reviewed; patients who received their cycle 1 day 1 (C1D1) infusion prior to noon were classified as the morning infusion group, while those who received infusions at or after noon were classified in the afternoon infusion group. Infusion times refer specifically to the start time of ICI infusion. In instances where multiple ICIs or combination therapies were given, the time of infusion refers to the start time of the first ICI infusion.

Electronic medical records were reviewed for clinical outcomes and patient demographic data. Primary end points of interest included progression-free survival (PFS), overall survival (OS), and best response according to Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1 criteria. Objective treatment response was defined as those who experienced partial and complete responses. Clinical data were censored on August 1, 2025. irAE development was determined by treating physicians and was graded according to the Common Terminology Criteria for Adverse Events (CTCAE) V.5. Only those that were reviewed to be definitively, likely, or possibly related to ICI therapy were included in our analysis. All clinical data were reviewed by two study members, including at least one board-certified medical oncologist.

Sample processing and cytokine measurements

Blood samples were collected and processed as previously described by our team.21,24 Briefly, peripheral blood samples were obtained in heparinized syringes and processed within 2 hours of collection. Subsequently, blood was transferred into a 50 mL conical tube and spun down in a centrifuge at 1600 r.c.f. for 30 min with the brakes off, and the plasma layer was isolated and stored in 1 mL aliquots at −80 °C. Additionally, peripheral blood mononuclear cells (PBMCs) from the remaining blood were isolated using a LeucoSep tube technique. Blood diluted in equal parts of phosphate-buffered saline (PBS) was added to LeucoSep tubes preloaded with Ficoll-Paque. The tubes were then centrifuged for 5 min at 1500 r.p.m. with the brake on. The PBMC suspension was isolated, and the cells were resuspended in freeze media (90% AIMV and 10% dimethyl sulfoxide) and stored in a cryovial initially at −80 °C until transfer to liquid nitrogen for long-term storage.

The Bio-Plex 200 platform (Bio-Rad, Hercules, California, USA) was used to measure the concentration (pg/mL) of cytokines of interest in plasma, as previously described by our team.21,24 Luminex bead-based immunoassays (Millipore, Billerica, Massachusetts, USA) were performed following the Johns Hopkins Immune Monitoring Core standard operating procedures and concentrations were determined using five-parameter log curve fits (using Bio-Plex Manager V.6.0) with vendor-provided standards and quality controls. The HCYTA-60K panel was used to detect and measure the following 39 cytokines: soluble CD40 ligand (sCD40L), interleukin (IL)-1α, IL-1β, IL-1 receptor antagonist (IL-1RA), IL-2, IL-3, IL-4, IL-5, IL-6, IL-8, IL-9, IL-10, IL-12 (p40), IL-12 (p70), IL-13, IL-15, IL-17A, IL-17E, IL-17F, IL-18, IL-21, IL-22, IL-23, IL-35, B-cell activating factor (BAFF), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon-inducible T-cell alpha chemoattractant (ITAC), tumor necrosis factor-α (TNF-α), interferon-γ (IFN-γ), monocyte chemoattractant protein-1 (MCP-1), macrophage inflammatory protein (MIP)-1α, MIP-1β, MIP-3α, myeloid progenitor inhibitory factor-1 (MPIF-1), regulated on activation, normal T cell expressed and secreted (RANTES), monokine induced by gamma interferon (MIG), interferon gamma-induced protein 10 (IP-10), and vascular endothelial growth factor (VEGF)-A. Concentrations that were outside of the standard curves values were categorized as ‘out of range’ (OOR). For analysis purposes, each cytokine concentration that was OOR below the standard curve values was replaced with the lower limit of the standard curve of the assay. OOR greater than the standard curve values was replaced with the upper limit of the standard curve.

Cytometry by time-of-flight analyses

CyTOF staining, acquisition, and analyses were performed at the Johns Hopkins CyTOF core facility as previously described by our team.21,24 Since our initial publication where 99 patients were analyzed with CyTOF,²¹ 122 additional patients were enrolled in our prospective pan-tumor Johns Hopkins immunotherapy biobank who also had peripheral immune cells profiled and clustered using CyTOF. Among these 221 patients, we focused our subset analysis on the 51 patients who had HCC. Specifically, patient PBMCs were gently thawed in a 37 °C water bath and slowly recovered using a prewarmed Roswell Park Memorial Institute (RPMI) medium supplemented with 10% fetal bovine serum (FBS). 2×10⁶ cells from each sample were then plated in 96-well plates. Cells were rested in the media for 30 min prior to staining and were subsequently washed once in PBS with 2 mM EDTA. Next, cells were incubated for 2.5 min at room temperature in 20 μM Pt (Standard BioTools) and PBS to distinguish viable cells. RPMI medium supplemented with 10% FBS was then added to the cells to neutralize residual platinum. The cells were then washed twice with cell staining buffer (CSB) (Standard BioTools). In order to barcode cells, all samples were subsequently incubated with unique combinations of metal-conjugated anti-CD45 antibodies (Standard BioTools) for 20 min. Afterward, the samples were washed twice with CSB. Samples were then multiplexed and transferred to v-bottom FACS tubes using a 40 μm filter. Next, the tubes were blocked using antihuman FcR block (12 µL used for 1.5×10⁷ cells) for 10 min at room temperature. This was followed by a chemokine stain cocktail (online supplemental table 1) for 10 min in a 37 °C water bath. After removal from the water bath, a surface stain cocktail (online supplemental table 1) was added for 30 min at room temperature. Samples were then washed twice with CSB and fixed/permeabilized using cytofix/cytoperm solution (BD Biosciences) for 30 min at room temperature. These samples were then washed with perm/wash solution (BD Biosciences) and stained using the intracellular cocktail (online supplemental table 1) for 30 min at room temperature. After two additional washes with perm/wash solution, samples were stored in 1.6% paraformaldehyde in PBS at 4°C until the day of acquisition, which was no later than 1 week after. On the day of acquisition, samples were stained with 1:500 ¹⁰³Rh in Maxpar Fix/Perm solution (Standard BioTools) for 30 min at room temperature for cell identification. Samples were washed with PBS once and resuspended in normalization beads (Standard BioTools). The data were acquired on a Helios mass cytometer (Standard BioTools) at the Johns Hopkins University CyTOF core facility. All acquired data were randomized and normalized using CyTOF software (V.7.1.16389.0, Standard BioTools). Resulting Flow Cytometry Standard (FCS) files were then debarcoded by manual gating using FlowJo software (V.10.9.0, BD Biosciences). Cell events were gated using ¹⁰³Rh positivity. Live cells were then gated based on ¹⁹⁴Pt and ¹⁹⁵Pt viability staining. Debarcoding was then completed based on positivity of unique combinations of CD45 barcodes. Each debarcoded sample was exported as an individual FCS file. Samples were normalized in R (V.4.0.2) using the CytoNorm algorithm that used a repeated sample included in each staining batch to normalize all samples based on goal quantiles of mean marker expression. For downstream analyses, each sample was subsampled to a maximum of 50 000 events to reduce computational burden and prevent the largest files from dominating clustering results. Clustering analysis was performed in R using FlowSOM to generate metaclusters that were annotated using the expression profile of markers included within the panel, resulting in the 24 final clusters used in this analysis. Cell-surface expression markers on each cluster were also obtained. Antigens used for CyTOF are also listed in online supplemental table 1. A heatmap demonstrating scaled marker profiles for each resulting final cluster is shown in online supplemental figure 1.

Statistical analysis methods

Survival data were analyzed using the Kaplan-Meier method and comparisons between groups were performed using the log-rank test as well as Cox regression analyses to adjust for covariates. Treatment response and irAE data were analyzed using multivariable and univariate logistic regressions, respectively. For cytokine analyses, early-on-treatment fold change was calculated relative to baseline to account for interpatient variability and to investigate effects of baseline treatment on immune effectors. Non-parametric methods were used to assess group differences for the cytokine analyses given skewing in the distributions; the Wilcoxon rank-sum test was used when assessing statistical differences between two groups. Log₂ transformation to fold change was performed to reduce skewness for visualization purposes. For the CyTOF analyses, the Wilcoxon rank-sum test was used when assessing statistical differences between groups. A Fisher’s exact test was used when assessing statistical differences between two categorical variables. All statistical tests were two-sided unless stated otherwise. For the statistical analyses, p values <0.05 were considered statistically significant. Statistical analysis was performed using RStudio software (V.2025.05.0+496). Only open-source software was used for this study, and no custom code was generated for data analysis of cytokine and CyTOF data.

Results

Study cohort demographics

Baseline characteristics of the study cohort, stratified by C1D1 infusion time (morning vs afternoon groups), are summarized in table 1. Among 84 patients, 39 (46.4%) received their first infusion in the morning and 45 (53.6%) in the afternoon. No significant differences between groups were observed in baseline demographics or clinical characteristics. Mean age was similar (66.2 in morning vs 67.1 years in afternoon group), as were sex distribution (predominantly male; 82.1% vs 86.7%) and racial composition (predominantly white; 64.1% vs 55.6%). ICI regimen type was balanced, with roughly half receiving ICI monotherapy or combotherapy and half receiving ICI plus tyrosine kinase inhibitor (TKI)/anti-VEGF (online supplemental table 2). Disease stage and liver function, as measured by Barcelona Clinic Liver Cancer (BCLC) stage and Child-Pugh class, were also comparable between groups.

Table 1. Baseline patient characteristics by first infusion time.

	Afternoon (n=45)	Morning (n=39)	P value
Age at study
Mean (SD)	67.1 (11.2)	66.2 (11.1)	0.802
Median (minimum, maximum)	68.0 (32.0, 88.0)	67.0 (20.0, 84.0)
Sex
Female	6 (13.3%)	7 (17.9%)	0.779
Male	39 (86.7%)	32 (82.1%)
Race
Black	18 (40.0%)	11 (28.2%)	0.521
Other	2 (4.4%)	3 (7.7%)
White	25 (55.6%)	25 (64.1%)
Prior systemic treatment
No	34 (75.6%)	33 (84.6%)	0.448
Yes	11 (24.4%)	6 (15.4%)
ICI regimen
ICI with TKI/aVEGF	22 (48.9%)	19 (48.7%)	1
ICI(s) only	23 (51.1%)	20 (51.3%)
BCLC stage
A	1 (2.2%)	1 (2.6%)	1
B	10 (22.2%)	9 (23.1%)
C	34 (75.6%)	29 (74.4%)
Child-Pugh class
A	32 (71.1%)	25 (64.1%)	0.814
B	12 (26.7%)	13 (33.3%)
C	1 (2.2%)	1 (2.6%)

aVEGF, antivascular endothelial growth factor; BCLC, Barcelona Clinic Liver Cancer; ICI, imune checkpoint inhibitor; TKI, tyrosine kinase inhibitor.

Statistical tests for significance between groups for continuous variables were done using Wilcoxon rank-sum test. The χ² test was used for categorical variables with values >5. Otherwise, Fisher’s exact test was used to test for differences between groups.

Morning C1D1 infusion predicts improved progression-free survival and treatment response outcomes in HCC

Patients who received morning C1D1 infusions demonstrated superior PFS as compared with those who received afternoon C1D1 infusions (log-rank p=0.029) (figure 1A), with median PFS of 4.7 months as compared with 2.8 months. Morning infusions were also associated with numerically improved OS, but these findings did not reach statistical significance (log-rank p=0.15) (figure 1B). PFS remained superior in the morning infusion group after accounting for clinical and tumor-specific features relevant in HCC, including Child-Pugh score, albumin-bilirubin (ALBI) grade, liver disease etiology, baseline α-fetoprotein (AFP), and BCLC stage (HR 0.50, 95% CI 0.30 to 0.84, adjusted p<0.01) (figure 1C). The morning group also had a significantly higher odds of objective treatment response than the afternoon group in a multivariable logistic regression (OR 3.26, 95% CI 1.08 to 10.90, adjusted p<0.05) (figure 1D). The receipt of ICI+anti-VEGF/TKI was associated with a superior PFS (HR 0.47, adjusted p<0.01) as compared with ICI(s)-only regimens. To further characterize the effect of specific ICI regimens on PFS, we stratified our cohort into three groups for further analysis: ICI+anti-VEGF/TKI, anti-PD(L)1 monotherapy, and combination anticytotoxic T-lymphocyte-associated protein 4+antiprogrammed cell death protein 1 (PD-1)/programmed death-ligand 1 (PD-L1) therapy. Among 42 patients who received ICI+anti-VEGF/TKI, initial morning infusion demonstrated an HR of 0.35 (p=0.02) in a multivariable Cox model for PFS (online supplemental figure 2A). The monotherapy and combination ICI groups also demonstrated trends toward improved PFS for the morning group (HR 0.39 and 0.74, respectively), although the sample size was more limited in these cohorts (online supplemental figure 2B,C). The superior outcomes seen in the anti-VEGF/TKI group are likely driven by the predominance of patients receiving atezolizumab with bevacizumab (online supplemental table 2), which may enrich for patients with fewer comorbidities and preserved performance status and liver function, factors that often contribute to eligibility for bevacizumab.

Figure 1. Morning baseline ICI infusion is associated with improved survival and objective response outcomes. KM curves for (A) PFS and (B) OS by baseline infusion time (morning vs afternoon). Survival differences analyzed using log-rank test. (C) Multivariable Cox model for PFS displaying HRs adjusted for clinical covariates (HRs, 95% CIs). (D) Multivariable logistic regression for objective response showing adjusted ORs (95% CIs). (E) Spline logistic model of predicted response probability across treatment start hour with 95% confidence band. Bottom rug indicates infusion times for specific patients colored by response status. AFP, alpha-fetoprotein; ALBI, albumin-bilirubin; aVEGF, antivascular endothelial growth factor; BCLC, Barcelona Clinic Liver Cancer; C1D1, cycle 1 day 1; ICI, immune checkpoint inhibitor; KM, Kaplan-Meier; NR, not reached; OS, overall survival; PFS, progression-free survival; TKI, tyrosine kinase inhibitor.

To gain a more granular understanding of how time-of-day impacts treatment response, we modeled a spline logistic regression to visualize predicted probability of response per hour, which suggests that late morning is associated with best probability of response and the early afternoon the worst (figure 1E). Based on this, we grouped patients into four time bins—early morning (9:00–10:30 hours), late morning (10:30–12:00 hours), early afternoon (12:00–15:00 hours), and late afternoon (15:00–18:00 hours), and found that using early afternoon as the reference group, late morning infusions were associated with significantly higher odds of response (OR 4.11, 95% CI 1.14 to 17.5, adjusted p=0.038) (online supplemental figure 3). Interpretation of early morning and late afternoon groups is limited due to small sample size.

Non-C1D1 morning infusions do not result in improved progression-free survival

After showing that the initial, C1D1, morning infusion itself is sufficient to predict improved oncological outcomes, we hypothesized that C1D1 is the most important infusion to dictate immune program and outcomes, especially given the long half-life of ICIs. After restricting the cohort to those who received three or more cycles of therapy, we found that additional morning infusions after the initial morning infusion did not result in improved PFS when comparing the two following groups: first, patients who received only C1D1 in the morning and all subsequent infusions in the afternoon (cycle 2 day 1 (C2D1) through cycle 4 day 1 (C4D1)) versus second, patients who received C1D1 in the morning and at least one other morning infusion (C2D1 through C4D1) (figure 2A). In fact, median PFS is improved in the first group (13.9 months vs 7.46 months), although the result was not statistically significant. Similarly, we find that there is no difference in PFS between patients who received all first four infusions in the afternoon compared with those who received C1D1 in the afternoon and at least one subsequent infusion (C2D1 through C4D1) in the morning (figure 2B).

Figure 2. Non-C1D1 morning infusions do not result in improved PFS. In patients treated with at least three cycles of ICIs, the KM method and log-rank test were used to compare PFS between (A) patients who received only C1D1 in the morning and all subsequent infusions in the afternoon (C2D1 through C4D1) versus patients who received C1D1 in the morning and at least one other morning infusion (C2D1 through C4D1) as well as (B) patients who received all four first infusions in the afternoon versus patients who received C1D1 in the afternoon and at least one other morning infusion (C2D1 through C4D1). C1D1, cycle 1 day 1; C2D1, cycle 2 day 1; C4D1, cycle 4 day 1; ICI, immune checkpoint inhibitor; KM, Kaplan-Meier; NR, not reached; PFS, progression-free survival.

C1D1 infusion time is not associated with immune-related adverse events

Given improved response and survival rates in the morning infusion group, we hypothesized that this signals an increase in immune activation that may also be associated with a higher incidence of irAEs. We find that morning treatment is not associated with clinically significant (ie, grade 2 or greater) irAE development (Fisher’s p=1.0) (figure 3A), nor is it associated with all grade irAE development (Fisher’s p=0.38) (figure 3B). However, the morning treatment group did have a numerically greater amount of low-grade, grade 1 irAEs, although this also did not achieve statistical significance. Using multivariable logistic regressions, the morning group had a numerically, but not statistically, greater odds of both grade 2+irAE (OR 1.11, 95% CI 0.44 to 2.80, adjusted p=0.83) (figure 3C) and all grade irAE (OR 1.61, 95% CI 0.68 to 3.88, adjusted p=0.29) (figure 3D) development on treatment. All irAE types and associated grades for both groups are described in online supplemental table 3.

Figure 3. Morning infusions of ICIs are not associated with increased incidence of clinically significant and all grade irAEs. Stacked bar plots of proportion of patients in the afternoon and morning groups who develop (A) clinically significant (ie, Common Terminology Criteria for Adverse Events V.5 grade 2 or greater) irAEs and (B) all grade irAEs. Fisher’s exact test was used to assess statistical differences between groups. Multivariable logistic regressions for (C) clinically significant and (D) all-grade irAE development showing adjusted ORs (95% CIs) for ICI regimen status. ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; TKI, tyrosine kinase inhibitor; VEGF, vascular endothelial growth factor.

Early-on-treatment cytokine and CyTOF analysis reveals time-of-day infusion immune signatures

Of the total cohort, 58 patients had paired baseline and early-on-treatment peripheral cytokines analyzed (online supplemental table 4). We analyzed fold change differences in cytokines between groups from baseline to early-on-treatment to isolate the systemic effects of initial infusion, which are visualized in figure 4A. IL-1beta, VEGF-A, and IL-21 were all significantly expanded early-on-treatment in the afternoon group compared with the morning group. IL-6 demonstrated the greatest between-group difference (mean log₂-fold change of −0.42 in the morning group and 0.83 in the afternoon group, p=0.0042) (figure 4B). All cytokine early-on-treatment fold change values are depicted in online supplemental table 5.

Figure 4. Early-treatment cytokine and immune cell changes to morning versus afternoon baseline ICI infusions. (A) Heatmap showing log₂-transformed fold changes from baseline to early-on-treatment for all analyzed cytokines between morning and afternoon groups. All cytokines with significant early fold change between groups are shown (B). One immune cell population analyzed using CyTOF had a statistically significant difference in log₂-transformed fold change in absolute counts between groups (C). All statistical tests for differences between groups were performed using the Wilcoxon rank-sum test (*p<0.05, **p<0.01). BAFF, B-cell activating factor; CyTOF, cytometry by time-of-flight; G-CSF, granulocyte colony-stimulating factor; GM-CSF, granulocyte-macrophage colony-stimulating factor; ICI, immune checkpoint inhibitor; IFN, interferon; IL, interleukin; IP-10, interferon gamma-induced protein 10; ITAC, interferon-inducible T-cell alpha chemoattractant; MCP-1, monocyte chemoattractant protein-1; MIG, monokine induced by gamma interferon; MIP, macrophage inflammatory protein; MPIF-1, myeloid progenitor inhibitory factor-1; RANTES, regulated on activation, normal T cell expressed, and secreted; sCD40L, soluble CD40 ligand; Th, T helper cell; TNF, tumor necrosis factor; Treg, regulatory T cell; VEGF-A, vascular endothelial growth factor A.

Of the total cohort, 52 patients had baseline and early-on-treatment CyTOF available (online supplemental table 6). In a similar early fold-change analysis, cytotoxic central memory CD8+ T cells (TcCM_I) were significantly expanded in the morning treatment group as compared with decreased in the afternoon group (p=0.013) (figure 4C). Cytotoxic effector memory (TcEM_I) and cytotoxic effector CD8+GZMB+ T cells were also expanded in the morning group as compared with decreased in the afternoon group (p=0.062 and p=0.063, respectively), although this difference did not reach statistical significance. All immune cell cluster log₂-fold change values are shown in online supplemental table 7.

Given that cytokines and immune cells have been shown to fluctuate in concentration and abundance throughout the day, it is possible that fold changes of peripheral blood immune correlates are due to blood-draw timing discrepancies between baseline and on-treatment draws, rather than infusion timing. To address this, we harmonized collection windows to include only patients who had baseline and on-treatment draws within 3 hours of each other. Despite a limited sample of 36 patients, we find reproduced results in our cytokine and CyTOF data. IL-21 remains significantly elevated in the afternoon group compared with the morning (p=0.004) group (online supplemental figure 4A). IL-6 levels remain expanded in the afternoon group and decreased in the morning group, although this result only trends toward significance in this abbreviated cohort (p=0.106) (online supplemental figure 4B). We also reproduce TcCM_I’s significant expansion in the morning group as compared with the afternoon group (p=0.033) (online supplemental figure 4C).

Discussion

In this retrospective cohort of patients with advanced/unresectable HCC, we find that receipt of the first ICI infusion in the morning is associated with superior PFS and higher odds of objective response without a significant increase in irAE incidence. These findings extend prior retrospective observations made across multiple solid tumors, including preliminary data in HCC.²⁵ Importantly, this study presents the first data to identify systemic immunological correlates that support a mechanistic basis for these improved outcomes.

While previous studies classified patients primarily by the proportion of early and late infusions, we demonstrate that the timing of the initial infusion alone is associated with superior clinical outcomes, whereas subsequent dosing times have no effect. This distinction offers an explanation for a long-standing puzzle: given the long half-lives of ICIs, with later doses largely serving to maintain drug exposure, it has been unclear why time-of-day should matter. Our findings suggest that the first infusion imprints an early immunological program that shapes downstream responses. The optimal immune priming hypothesis is supported by preclinical findings where checkpoint blockade during CD8+ T cell priming and activation is associated with ICI efficacy and that suboptimal priming can result in resistance.^{26 27} Similarly, preclinical models have also discovered that specific T-cell responses to dendritic cell vaccination are stronger when immunized at specific times of the day.²⁸ Finally, an important preclinical study demonstrated that time-of-day of anti-PD-1 infusion significantly impacted tumor growth, an effect that was abrogated in circadian gene, Bmal1, knockout mice.¹⁵ In total, this study suggests that the improved efficacy from C1D1 morning infusions is due to an optimal circadian window that provides the foundation for downstream antitumor immune profiles.

Importantly, this is the first study to integrate key correlates from serial blood samples to understand how immunological mechanisms differ between morning and afternoon infusion groups. Our cytokine data show that IL-6 is the most significantly expanded cytokine in those who receive baseline treatment in the afternoon as compared with the morning; notably, IL-6 also decreases in the morning group. These observations are congruent with existing literature that IL-6 is associated with pro-tumorigenic properties and inferior responses to ICI therapy in multiple tumor types, including in HCC.^{29 30} Furthermore, IL-6 is thought to contribute to diurnal, circadian regulation, with IL-6 levels displaying a clear trough in the morning hours.³¹ We therefore posit that afternoon dosing, coinciding with increased IL-6 levels, may amplify a pro-inflammatory milieu that counteracts ICI efficacy, whereas morning infusion aligns with the physiological nadir and may be more favorable. Interestingly, morning CAR-T administration in large B-cell lymphomas also demonstrated lower IL-6 concentrations on treatment as compared with afternoon administration.¹² We also found that cytotoxic central memory CD8+ T cells are significantly expanded in the morning group compared with the afternoon group. This corroborates the hypothesis that CD8+ T cells demonstrate their own circadian fluctuations in both level and activity and are better primed for expansion at certain times of the day.^{19 28 32} These data are also strengthened by the observation that central memory T cells are lowest in abundance in the early afternoon.²⁰ Collectively, these immunological correlates provide important biologic plausibility to survival outcomes in HCC based on ICI infusion time.

With regard to toxicity, previous studies provide mixed evidence on the incidence of irAEs in early infusion groups, with some suggesting increased incidence of grade 1–3 toxicities.⁴ Although we observed a non-significant increase in the incidence of grade 1 toxicities in the morning treatment group, we found no difference in clinically significant (ie, grade 2+) irAE rates between the morning and afternoon groups. Our data suggest that the increased efficacy from morning ICI initiation can be achieved without safety trade-offs in HCC.

A limitation of this study is that the timing of the first ICI infusion was not randomized, although in our own practical experience, the timing of first infusion is often dictated by infusion chair availability or timing of oncologist follow-up rather than patient-specific factors. Nevertheless, it is not possible to fully exclude the possibility that patients treated in the morning differed in unmeasured ways from those treated in the afternoon, despite adjustment for observable baseline characteristics. In addition, improved outcomes with morning dosing may not reflect circadian regulation of immune populations per se but instead could represent a surrogate marker for other factors correlated with morning treatment, such as a more fasted metabolic state in patients receiving therapy prior to meals. Furthermore, while we attempted to control for the contribution of circadian influences on systemic immune cell and cytokine profiles, it is possible that there remains residual confounding related to sample collection timing that could not be fully accounted for. It is therefore important that these results are interpreted cautiously.

It is notable that the composition and spatial distribution of the gut microbiome fluctuates in a circadian manner^{33 34}; whether the circadian microbiome influences these findings remains to be studied. It also remains unclear what specific time window results in best outcomes, and our data suggest that the most optimal timing occurs in the late morning, although more robust analyses are needed for confirmation. Given the recent concerns raised regarding the only prospective randomized controlled trial showing clinical benefit,¹³ future prospective trials in multiple tumor types, including HCC, are necessary to validate the clinical impact of early day ICI infusion.

Our data link initial (C1D1) morning initiation of ICI therapy to improved clinical benefit in HCC and provide serial immunological correlates—both reduced IL-6 and expansion of cytotoxic central memory CD8+ T cells—that align with existing hypotheses on circadian immunology. These findings require validation in prospective, randomized trials across different patient cohorts.

The study sponsors were not involved in the study design, data analysis, data interpretation, or initial drafting of the manuscript. The sponsors were involved only in providing feedback on and revisions to the manuscript.

Data availability statement

Data are available on reasonable request.