The PD-1 receptor on T cells interacts with its ligands PD-L1 (CD274) and PD-L2 (CD273) on tumor cells or antigen-presenting cells, delivering inhibitory signals that induce T-cell exhaustion, energy, and apoptosis mechanisms that tumors exploit for immune evasion (14). PD-1/PD-L1 blockade with monoclonal antibodies (e.g., nivolumab, pembrolizumab targeting PD-1; atezolizumab, durvalumab targeting PD-L1) disrupts this axis, restoring effector T-cell function, cytokine production (IFN-γ), and cytolytic activity (15).

In hematologic malignancies, Reed-Sternberg (RS) cells in cHL and malignant B cells in PMBCL frequently harbor copy number alterations at chromosome 9p24.1 encompassing PD-L1/PD-L2 and JAK2, leading to JAK/STAT-mediated overexpression of PD-L1/PD-L2 and profound dependence on the pathway for survival (9). This genetic addiction renders these tumors exquisitely sensitive to PD-1 blockade, resulting in one of the highest response rates observed across oncology (ORR 69–87%) (5, 16). Additional mechanisms include EBV-driven PD-L1 expression in subsets of lymphomas and leukemias, and chronic antigenic stimulation in the bone marrow niche promoting exhaustion (17).

In contrast to solid tumors, hematologic cancers often exhibit lower somatic mutational burden and fewer neoantigens, yet the amplified PD-L1 expression in specific subtypes overrides this limitation, explaining the outlier success in cHL/PMBCL (18). Preclinical models further demonstrate that PD-1 blockade enhances NK cell activity and reverses myeloid-derived suppressor cell suppression in the marrow microenvironment (19).

Classical Hodgkin lymphoma, PMBCL represents the flagship success of PD-1 blockade in hematologic oncology. Pivotal phase II trials established nivolumab (CheckMate 205) and pembrolizumab (KEYNOTE-087) as standards for r/r disease post-autologous transplant and brentuximab vedotin, with ORR 69–73%, complete response (CR) rates 16–29%, and median duration of response exceeding 16–24 months (16, 20). Long-term follow-up of CheckMate 205 (6–7 years) reported 5-year progression-free survival (PFS) with 72% in transplant-naïve patients and durable remissions beyond 5 years in ∼60% of responders (21). Pembrolizumab demonstrated 5-year OS ∼85% in KEYNOTE-087 updated analyses (22). These agents received FDA accelerated approval in 2016–2017 and full approval thereafter, transforming third-line management and now incorporated earlier (e.g., with AVD chemotherapy in the frontline) with 3-year PFS > 90% in phase III trials (23, 24).

Primary mediastinal large B-cell lymphoma shares the 9p24.1 alteration profile with cHL. Pembrolizumab in KEYNOTE-170 (r/r PMBCL) achieved ORR 45% (CR 13%) with durable responses (median DOR not reached at 3 + years), leading to FDA approval in 2018 (25). Nivolumab has shown similar activity (ORR ∼40–50%) in smaller series (26). PD-1 blockade is now guideline-preferred in r/r PMBCL.

PD-L1 expression by IHC (22C3 or 28-8 assays) is the most established biomarker in cHL/PMBCL, where 9p24.1 amplification correlates with near-universal expression and superior ORR/CR rates (> 80–90%) (9, 43). Tumor mutational burden is generally low (< 5 mut/Mb) in hematologic malignancies compared to MSI-high solid tumors, limiting its predictive value outside rare hypermutated cases (44). EBV status (LMP1-driven PD-L1) predicts response in NK/T-cell and PTCL (45).

These biomarkers fail in many hematological settings for several reasons: (i) low neoantigen load reduces baseline T-cell priming; (ii) dominant immunosuppressive marrow microenvironment [myeloid-derived suppressor cells (MDSCs), M2 macrophages, TGF-β] overrides PD-1 blockade; (iii) lack of standardized PD-L1 scoring in liquid tumors and assay discordance; (iv) confounding inflammation elevating PD-L1 without functional significance; (v) rapid resistance via alternative checkpoints (TIM-3, LAG-3, TIGIT) or MHC loss (46, 47). Composite scores incorporating soluble PD-L1, IFN-γ signature, or circulating tumor DNA are under investigation but not yet clinically implemented (48).

Immune cells express receptors for virtually all hormones, adipokines, and nutrient-sensing pathways (mTOR, AMPK, IGF-1R, VDR, AhR), allowing systemic metabolic cues to orchestrate leukocyte development, trafficking, and effector differentiation (54, 55). Conversely, cytokines (IFN-γ, IL-6, TNF-α) modulate hypothalamic–pituitary axes, adipocyte function, and nutrient transporter expression, creating feedback loops that can either amplify or suppress antitumor responses (50, 56). In the tumor microenvironment (TME), this crosstalk is hijacked: tumors induce chronic low-grade inflammation that elevates cortisol, leptin, and insulin while depleting micronutrients, driving T-cell dysfunction and resistance to PD-1/PD-L1 blockade (57). Preclinical models and clinical cohorts show that correcting these imbalances (e.g., vitamin D repletion, metformin-mediated insulin sensitization) restores CD8⁺ T-cell metabolism and synergizes with checkpoint inhibition (58, 59). The concept has particular relevance in hematologic malignancies, where marrow infiltration, cachexia, and prior therapies frequently induce profound endocrine–nutritional dysregulation (60). A mechanistic overview of how endocrine and nutritional pathways converge on T-cell immunity is shown in Figure 1.

Chronic activation of the HPA axis and resultant hypercortisolism exerts profound systemic immunosuppression by directly inducing T-cell exhaustion and apoptosis while expanding immunosuppressive myeloid populations (105). Glucocorticoids (GCs) upregulate PD-1, CTLA-4, and TIM-3 on CD8⁺ T-cells via glucocorticoid receptor-mediated transactivation and transrepression of pro-inflammatory transcription factors (NF-κB, AP-1), simultaneously inhibiting IL-2 and IFN-γ production (61, 62).

Observational findings consistently link baseline GC use (≥ 10 mg prednisone equivalent daily, often for symptom control or comorbidities) with markedly inferior outcomes on PD-1/PD-L1 therapy. A landmark study in NSCLC showed that baseline corticosteroids were associated with reduced ORR (ORR: 8.5% vs. 32.3%), PFS (HR = 1.93), and OS (HR = 2.34) independent of performance status or brain metastases (106). Meta-analyses confirm this detrimental effect across tumor types, with dose- and duration-dependent impairment (107, 108). Endogenous hypercortisolism (e.g., Cushing syndrome or chronic stress-elevated morning cortisol > 500–600 nmol/L) similarly predicts resistance, as shown in preclinical models where tumor-derived GC reactivation via HSD11B1 limits IFN-γ signaling and CD8⁺ T -cell infiltration (66).

In hematologic malignancies, data are sparser but supportive: real-world cHL cohorts treated with nivolumab/pembrolizumab demonstrate shorter response duration in patients requiring GCs for symptom control or immune-related adverse events (irAEs) (5). Hypophysitis-induced secondary adrenal insufficiency (common irAE, incidence 5–15% with combination CTLA-4/PD-1 blockade) paradoxically does not worsen prognosis when promptly replaced with physiologic hydrocortisone, suggesting that supra-physiologic GC doses, not cortisol deficiency per se, drive immunosuppression (109, 110). The redistribution of circulating memory T cells under hypercortisolism and CXCR4 signaling is illustrated in Figure 3.

Thyroid hormones (T3/T4) are essential for T-cell development, proliferation, and Th1 polarization via thyroid receptor expression on lymphocytes and dendritic cells (67). PD-1/PD-L1 inhibitors frequently induce thyroid irAEs (irAE-thyroiditis, incidence 10–40%, highest with combination therapy), manifesting as transient thyrotoxicosis followed by hypothyroidism, reflecting destructive autoimmunity against thyroperoxidase/glutamate decarboxylase antibodies (69).

Multiple large cohorts and meta-analyses (> 10,000 patients) establish that development of thyroid dysfunction is a robust positive predictor of ICI benefit that compared to patients without thyroid irAEs (70, 111, 112). This association holds across melanoma, NSCLC, and renal cancer and appears independent of other irAEs, likely reflecting systemic immune activation spilling into thyroid autoimmunity (113). Mechanistically, ICI-induced thyroiditis correlates with increased CD8⁺ T-cell reinvigoration, reduced Treg frequency, and lower PD-1 expression on peripheral T cells (114).

Conversely, pre-existing hypothyroidism (often autoimmune Hashimoto’s) is associated with poorer outcomes, possibly via baseline Treg expansion and impaired T-cell priming (68). Subclinical hyperthyroidism or low TSH at baseline may also predict resistance in some series (115). In hematologic cancers, thyroid irAEs occur in 15–25% of cHL patients on PD-1 blockade and similarly predict prolonged responses, while untreated baseline hypothyroidism correlates with early progression in small DLBCL/MM cohorts (22, 31). The cell-specific effects of thyroid hormones on immune activation and differentiation are shown in Figure 4.

Sex hormones profoundly shape antitumor immunity, with estrogen enhancing CD8⁺ T-cell effector function, dendritic cell cross-presentation, and IFN-γ signaling via ERα/β, while testosterone suppresses Th1 responses and promotes Treg/MDSC accumulation through androgen receptor signaling (72, 75).

A seminal 2018 meta-analysis of > 20 randomized trials (n > 11,000) found that males derive greater benefit from PD-1/PD-L1 inhibitors than females, a difference most pronounced in monotherapy versus chemotherapy controls (116). Subsequent analyses confirmed male advantage in melanoma, NSCLC, and head and neck cancer, with hazard ratios favoring males by 20–40% (76, 117). Mechanistically, androgen deprivation in prostate cancer models dramatically boosts CD8⁺ T-cell infiltration and synergizes with PD-1 blockade, while estrogen in females may upregulate alternative checkpoints (LAG-3, TIM-3) (118).

Hormonal aging exacerbates immune senescence: post-menopausal estrogen decline and age-related hypogonadism both correlate with reduced ICI efficacy (73). In hematologic malignancies, male sex is an adverse prognostic factor in some cHL real-world series treated with nivolumab (shorter PFS), consistent with testosterone-driven immunosuppression (119).

Insulin resistance, type 2 diabetes, and metabolic syndrome drive chronic low-grade inflammation (elevated IL-6, TNF-α) that promotes T-cell exhaustion, MDSC expansion, and PD-L1 upregulation on tumor cells (78). Hyperinsulinemia activates PI3K/AKT/mTORC1 in T cells, inducing energy while impairing memory formation (79).

Retrospective studies and meta-analyses show that diabetes at baseline confers a 30–70% increased risk of progression or death on PD-1/PD-L1 therapy (HR 1.3–1.7), with hyperglycemia (> 200 mg/dL) during treatment independently predicting resistance (81, 120). Metabolic syndrome components (obesity + hypertension + dyslipidemia) compound this effect (121). Metformin, by ameliorating hyperinsulinemia and activating AMPK, has shown synergistic effects with ICIs in preclinical models and observational cohorts (improved ORR/PFS in diabetic patients) (122). Data in hematologic cancers are emerging: insulin resistance correlates with inferior responses in myeloma and DLBCL treated with PD-1 combinations (123).

Vitamin D (via VDR on immune cells) promotes CD8⁺ T-cell activation, dendritic cell maturation, chemokine production (CXCL10), and gut microbiome diversity while inhibiting Treg and PD-1 expression (12, 124).

Multiple cohorts demonstrate that vitamin D deficiency (< 20 ng/mL) at baseline is associated with significantly worse ORR, PFS, and OS across ICI-treated cancers (125–127). A 2023 prospective study showed that systematic vitamin D supplementation (≥ 2,000 IU/day) increased ORR from 36 to 56% and median PFS from 5.8 to 11.3 months, with reduced severe irAEs (127). Genetic polymorphisms in VDR and CYP27B1 also predict outcomes (128). In hematologic cohorts, low vitamin D is prevalent (> 60% in lymphoma/MM) and correlates with poorer survival on PD-1-based regimens (129).

Adipose tissue secretes adipokines that bidirectionally regulate immunity: leptin promotes T-cell exhaustion via mTOR activation and PD-1 upregulation while expanding Tregs, whereas adiponectin exerts anti-inflammatory, CD8⁺ T-enhancing effects (82, 83).

Leptin levels are elevated in obesity and directly induce PD-1 on CD8⁺ T-cells in mouse models and human tumors, accelerating exhaustion; leptin signaling blockade restores ICI efficacy (82, 84). Yet a robust “metabolic obesity paradox” emerges: overweight/obese patients (BMI ≥ 25–30) consistently show superior ORR, PFS, and OS on PD-1/PD-L1 therapy across melanoma, NSCLC, and renal cancer in meta-analyses (> 30,000 patients) (13, 130). Proposed mechanisms include leptin-driven tonic signaling preventing terminal exhaustion, increased CD8⁺ T-cell infiltration, and altered pharmacokinetics (131). This paradox extends to hematologic malignancies: higher BMI predicts longer response duration in cHL treated with nivolumab/pembrolizumab in real-world series, though sarcopenic obesity negates the benefit (132, 133).

Protein-energy malnutrition (PEM) is highly prevalent in hematologic malignancies, affecting 30–60% of patients with aggressive lymphomas and MM, and is characterized by involuntary weight loss, hypoalbuminemia (<35 g/L), low prealbumin (<20 mg/dL), and reduced lean body mass. Multiple large retrospective cohorts have demonstrated that low baseline serum albumin and low Prognostic Nutritional Index (PNI) (PNI = 10 × albumin [g/dL] + 0.005 × lymphocyte count [/mm³]) are independent predictors of inferior objective response rate (ORR, PFS), and OS in patients receiving anti-PD-1/PD-L1 therapy (13, 82, 130, 136).

Mechanistically, hypoalbuminemia reflects chronic systemic inflammation (elevated IL-6, TNF-α) that drives T-cell exhaustion through persistent PD-1 upregulation and reduced mTOR pathway suppression, limiting the amino acid availability required for effector T-cell expansion and memory formation (137). In a multicenter Italian study of > 1 000 patients treated with anti-PD-1/PD-L1 agents, albumin < 35 g/L conferred a hazard ratio of 1.92 for death (82). Similar findings have been reported with the PNI, where PNI < 45 predicted significantly lower ORR (≈15–20% vs. 40–50% in high PNI) and shorter median OS in NSCLC and other cancers (136, 138).

In hematologic malignancies, low albumin is incorporated into established prognostic scores (e.g., International Prognostic Score for Hodgkin lymphoma, R-IPI for DLBCL) and retains prognostic significance in the immunotherapy era. Real-world cohorts of relapsed/refractory cHL treated with nivolumab or pembrolizumab have shown that albumin < 35 g/L is associated with markedly shorter duration of response and increased risk of early progression (139, 140). Prealbumin may be even more sensitive, reflecting acute changes in visceral protein status and predicting non-response in a small series of myeloma patients receiving PD-1-based combinations (141). These data strongly suggest that routine pre-treatment nutritional screening with albumin or PNI could identify patients requiring aggressive supportive care to optimize immunotherapy benefit.

Anemia affects up to 70% of patients with lymphoma or myeloma at diagnosis and is exacerbated by disease progression or prior therapies. Baseline anemia (Hb < 10–11 g/dL) consistently predicts poorer outcomes with PD-1/PD-L1 blockade across tumor types, with meta-analyses showing hazard ratios of 1.5–2.0 for death (99, 100). High serum ferritin (> 300–500 ng/mL, context-dependent) as an acute-phase reactant reflects systemic inflammation and correlates with resistance to checkpoint inhibition, increased risk of hyperprogressive disease (142, 143).

Mechanistically, elevated hepcidin in inflammatory states sequesters iron in macrophages, depriving T cells of the iron required for proliferation and effector differentiation while simultaneously promoting M2-polarized immunosuppressive macrophages and myeloid-derived suppressor cells (144). In NSCLC cohorts, ferritin > 400 ng/mL was independently associated with reduced OS (HR 1.6–2.1) and lower ORR (145). In hematologic malignancies, hyperferritinemia is a well-known adverse factor in cHL and DLBCL and likely contributes to the immunosuppressive marrow microenvironment that limits PD-1 blockade efficacy in myeloma or leukemia (131). Careful correction of anemia (preferably with transfusion or erythropoietin-sparing approaches) and avoidance of iron overload may represent low-risk interventions to improve immunotherapy outcomes.

Zinc and selenium deficiencies are frequent in hematologic malignancies due to poor intake, malabsorption, and increased utilization. Zinc is essential for thymic function, ZAP-70 signaling, and cytotoxic T-cell activity; low serum zinc levels (< 70 μg/dL) have been associated with significantly lower response rates and shorter PFS/OS in NSCLC patients receiving PD-1 inhibitors (84, 93). Selenium, incorporated into selenoproteins (e.g., GPx-4), protects T cells from oxidative stress during chronic antigen stimulation; low selenium status correlates with higher Treg frequency and reduced effector T-cell function in DLBCL and other cancers (146).

Supplementation studies, though limited, suggest that restoring zinc or selenium levels can enhance CD8⁺ T-cell cytotoxicity and synergize with PD-1 blockade in preclinical models (92). In DLBCL, low selenium was associated with increased PD-1⁺ Treg populations, suggesting a direct mechanism of immune evasion that could be reversed by supplementation (93). Routine micronutrient screening and correction may therefore represent a simple, low-cost strategy to augment checkpoint inhibitor efficacy in hematologic patients.

Dietary patterns profoundly shape the gut microbiome, which in turn systemically modulates response to PD-1/PD-L1 blockade. High-fiber, plant-rich diets are associated with enrichment of responder-associated taxa (Akkermansia muciniphila, Faecalibacterium prausnitzii, Bifidobacterium spp.) and significantly higher ORR and PFS in melanoma and NSCLC patients treated with anti-PD-1 therapy (147, 148). Conversely, Western-style high-fat/processed food diets correlate with unfavorable microbiome composition and reduced benefit (147). Metabolomic profiling has identified pre-treatment elevations in short-chain fatty acids (butyrate, propionate) and certain amino acid metabolites as predictors of response, reflecting microbial fermentation products that enhance dendritic cell function and T-cell infiltration (132).

Although direct evidence in hematologic malignancies is limited, dysbiosis is common in lymphoma and myeloma patients (often exacerbated by antibiotics or chemotherapy), and the gut microbiome’s influence on systemic immunity and GVHD after transplant suggests similar relevance for checkpoint inhibitor efficacy (149). Dietary interventions promoting favorable microbiome composition represent a promising adjunctive strategy.

An “obesity paradox” has been repeatedly demonstrated in immunotherapy, with overweight/obese patients (BMI ≥ 25 or ≥ 30 kg/m²) exhibiting superior ORR (up to 2-fold higher), PFS, and OS compared to normal-weight patients in large multicenter cohorts and meta-analyses. The effect is particularly striking in melanoma, NSCLC, and renal cell carcinoma, with hazard ratios for death of 0.6–0.7 in obese patients (150, 151). Proposed mechanisms include increased leptin signaling enhancing CD8⁺ T-cell function, greater energy reserves supporting prolonged immune activation, and altered adipokine profiles favoring Th1 polarization (152). Preliminary data in classical Hodgkin lymphoma treated with PD-1 inhibitors also suggest a similar trend, with higher BMI associated with improved response duration (153). Representative clinical studies examining endocrine and nutritional predictors of ICI outcomes are summarized in Table 3.

Individual endocrine or nutritional markers demonstrate inconsistent predictive performance for ICI efficacy due to high inter-patient variability, overlapping confounders (comorbidities, steroids, cachexia), and inability to capture synergistic interactions across immuno-metabolic pathways (155, 156). For example, while hypoalbuminemia robustly predicts inferior outcomes, its specificity is limited by non-nutritional causes (hepatic dysfunction, nephrotic syndrome) and failure to account for compensatory mechanisms like adipokine signaling in obesity (85). Similarly, vitamin D deficiency correlates with resistance but shows heterogeneous effect sizes across cohorts, modulated by baseline inflammation or microbiome status (157). BMI alone yields paradoxical results (obesity benefit in many settings) yet ignores muscle quality, where sarcopenia independently drives exhaustion (82). Meta-analyses confirm modest hazard ratios (1.3–1.8) for single markers versus > 2.5–4.0 for composites, underscoring the need for integrated approaches that better reflect the complex “whole-body immunologic tone” governing T-cell fitness (136, 137). Table 5 synthesizes the current clinical evidence linking endocrine and nutritional parameters with PFS and OS in cancer patients receiving ICIs. Collectively, the studies summarized in this table highlight that host metabolic and hormonal status is not merely a background variable but a biologically active determinant of immunotherapy efficacy. Across multiple tumor types and treatment settings, endocrine alterations such as dysregulated thyroid function, cortisol excess, insulin resistance, and sex hormone imbalance emerge as modulators of antitumor immune responses, influencing T-cell activation, exhaustion dynamics, and immune-related adverse event profiles. In parallel, nutritional indices including BMI, sarcopenia, cachexia, serum albumin, PNI, and inflammatory-nutritional composites (e.g., CONUT, GPS, NLR-based scores) consistently correlate with survival outcomes, underscoring the role of systemic energy availability and muscle–immune crosstalk in sustaining effective immune surveillance. Notably, several studies report paradoxical findings such as improved outcomes in overweight patients supporting the concept of an “immunometabolic reserve” that may buffer immune cells against ICI-induced metabolic stress.

The systemic immunologic readiness profile conceptualizes pre-treatment host status as a composite continuum from “fit” (optimal nutrient reserves, balanced hormones, low inflammation supporting robust CD8⁺ T cell reinvigoration) to “unfit” (malnutrition, endocrine dysregulation, chronic inflammation enforcing exhaustion and Treg/MDSC dominance) (51, 164). This paradigm shifts focus from tumor-intrinsic features (PD-L1, TMB) to modifiable host factors that determine whether PD-1/PD-L1 blockade can restore effective immunity. Preclinical models demonstrate that combined insults (e.g., cortisol elevation + zinc deficiency + arginine depletion) synergistically impair mTOR signaling and IFN-γ production beyond any single factor (57). Clinically, patients with “fit profiles exhibit higher ORR (50–80% vs. < 20%), longer PFS/OS, and fewer severe irAEs, likely via enhanced T-cell metabolism and reduced alternative checkpoint upregulation (165). In hematologic malignancies, where systemic disease and prior therapies amplify host dysregulation, this profile may explain outlier successes (cHL) versus failures (MM monotherapy) (166).

Hematologic malignancies exhibit unique metabolic vulnerabilities: aggressive lymphomas and myeloma frequently induce protein-energy wasting (30–60% prevalence), hypercatabolism, and marrow adipose tissue remodeling that amplifies MDSC and impair hematopoiesis (19, 176). In cHL, the obesity paradox is pronounced higher BMI/leptin associates with superior duration of response to nivolumab/pembrolizumab, possibly via sustained CD8⁺ T-cell metabolism overriding 9p24.1-driven evasion (119). DLBCL shows similar trends, with low PNI/GPS predicting early relapse post-ICI combinations (177). MM’s refractory nature to PD-1 monotherapy may stem from profound cachexia, vitamin D deficiency (bone disease), and IGF-1 dysregulation in the marrow niche, fostering T-cell energy (178).

The marrow microenvironment rich in adipocytes, cytokines (IL-6), and hepcidin—amplifies endocrine-nutritional effects: leptin from marrow fat promotes Treg, while iron dysregulation starves T cells (179). Real-world cHL series incorporating PNI report low scores in 40–50% of relapsed patients, correlating with reduced CR rates (139). Emerging data in CAR-T (analogous immune activation) validate CONUT/PNI as predictors in MM, suggesting relevance for ICI combinations (180).

Composite panels integrating 3–6 markers (e.g., PNI/GPS + vitamin D + sarcopenia index + leptin/adiponectin ratio ± cortisol) could achieve > 80% accuracy in stratifying ICI responders, enabling precision supportive care (nutrition repletion, exercise, metformin, vitamin D) to convert “unfit” to “fit” hosts (180, 181). Machine learning-derived scores (e.g., combining SII + CONUT + BMI) outperform traditional tools in solid tumors and are being validated in lymphoma trials. In hematologic settings, proposed panels might add marrow-specific factors (e.g., ferritin, free light chains) for personalized prediction, potentially augmenting low baseline efficacy in MM or leukemias (182). Prospective trials are essential to standardize cut-offs, timing (circadian variability in cortisol), and interventions.

T-cell exhaustion, a hypofunctional state characterized by hierarchical loss of effector cytokines, high co-inhibitory receptor expression (PD-1, TIM-3, LAG-3), and epigenetic remodeling, is metabolically regulated (185). Glucocorticoids, via GR-mediated transactivation, directly upregulate PD-1 and TOX/TOX2 transcription factors while inhibiting mTORC1 and glucose uptake, locking CD8⁺ T-cells in exhaustion and preventing proliferation/memory formation; blockade of GC signaling restores ICI responsiveness in preclinical models (61, 66).

Leptin, elevated in obesity, activates STAT3/mTORC1 in T cells, leptin-neutralization reinvigorates exhausted T cells and synergizes with PD-1 blockade (82, 84). Hyperinsulinemia/IGF-1 similarly hyperactivates PI3K/AKT/mTORC1, inducing exhaustion while impairing CD8⁺ memory T precursors; metformin-mediated AMPK activation reverses this and boosts ICI efficacy (186).

Vitamin D/VDR signaling inhibits exhaustion genes (PDCD1, HAVCR2) via direct binding to super-enhancers and promotes stem-like TCF1 + PD-1int progenitors capable of reinvigoration upon PD-1 blockade (12, 124). Zinc deficiency impairs ZAP-70/LCK signaling and IL-2 production, accelerating exhaustion; zinc supplementation restores cytotoxicity and reduces PD-1 expression (187). Arginine and tryptophan availability are critical: tumor/MDSC-derived ARG1/IDO1 deplete these amino acids, activating GCN2/ATF4 stress pathways that upregulate PD-1 and inhibit memory formation; arginine supplementation or IDO inhibitors prevent exhaustion and enhance PD-1 blockade (95, 96). Figure 5 summarizes how endocrine and nutritional factors shape T-cell states and influence PD-1/PD-L1 therapy outcomes.

Endocrine–nutritional cues reprogram APCs and MDSCs, determining co-stimulatory capacity and immunosuppressive polarization. Vitamin D is a master regulator of DC maturation: 1,25(OH)2D induces tolerogenic phenotype (↑CD14, ↓CD80/86, ↑IL-10) in steady state but enhances cross-presentation and CXCL10 production in inflammatory contexts, promoting CD8⁺ priming and ICI efficacy (188, 189). Zinc and selenium support lysosomal function and ROS detoxification in DCs, preserving MHC-II expression and co-stimulatory molecule expression; deficiency impairs antigen presentation and favors MDSC expansion (94).

Cortisol and leptin drive M2/macrophage polarization via PPARγ/STAT6 and suppress MHC-II/CD86 on APCs, limiting T-cell priming (65). In obesity, leptin-rich marrow adipocytes secrete IL-6/TGF-β that expand PMN-MDSCs expressing PD-L1 and ARG1, directly inhibiting CD8⁺ T-cell function (19). Iron dysregulation (high ferritin/hepcidin) promotes MDSC accumulation and M2 polarization while starving T cells; ferroptosis induction in MDSCs enhances ICI responses (179). These myeloid shifts create a “cold” pre-treatment state resistant to PD-1/PD-L1 blockade.

The bone marrow TME in lymphomas and myeloma is uniquely sensitive to endocrine–nutritional inputs due to high adipocyte content, chronic IL-6 signaling, and nutrient competition for glucose/amino acids (190). Marrow adipose tissue (MAT) expands dramatically in MM and aggressive lymphomas, secreting leptin, adiponectin (often reduced), and free fatty acids that reprogram resident immune cells (191). Leptin from MAT drives PD-L1 expression on myeloma cells via JAK2/STAT3 and promotes MDSC/Treg infiltration, while adiponectin deficiency removes CD8⁺-supportive signals (192).

Vitamin D deficiency, prevalent in MM due to bone disease, exacerbates marrow immunosuppression by reducing CXCL10-mediated T-cell trafficking and enhancing RANKL-driven osteoclastogenesis that releases TGF-β (193). Insulin/IGF-1 hyperactivation in obese myeloma patients upregulates PD-L1 on plasma cells and fosters “exhausted-like” marrow-resident T cells with high TOX expression (194). Zinc/selenium depletion—common from poor intake/malabsorption—impairs NK-cell degranulation against RS cells in cHL, contributing to the immunosuppressive niche (93). Thus, unfavorable endocrine–nutritional profiles reinforce marrow as an immune-privileged sanctuary, explaining limited PD-1 monotherapy success in MM vs. cHL (where 9p24.1 amplification overrides some constraints).

Endocrine–nutritional status orchestrates cytokine networks that determine whether PD-1 blockade triggers productive IFN-γ-driven tumor control or futile inflammation. IFN-γ induces PD-L1 on tumor/APCs to enable adaptive resistance; however, chronic IL-6 (driven by obesity, stress, malnutrition) trans-signals STAT3 to upregulate alternative checkpoints (TIM-3, LAG-3) and anti-apoptotic genes, blunting IFN-γ efficacy (97). Cortisol suppresses IFN-γ/IL-2 while elevating IL-10/TGF-β, shifting from Th1 to Treg-polarizing milieu (105).

Vitamin D and short-chain fatty acids (from fiber-rich diets enhance IFN-γ/CXCL9/10/11 production by DCs, promoting CD8⁺ T-cell trafficking and PD-1 blockade sensitivity (100). Leptin amplifies IL-6/TNF-α loops that sustain exhaustion, whereas adiponectin dampens them (83). Arginine depletion inhibits iNOS-derived NO needed for chemokine receptor expression, impairing T-cell migration; restoration reinvigorates cytokine networks (95).

Paradoxically, endocrine–nutritional factors that impair antitumor efficacy often protect against severe irAEs by limiting systemic immune activation. Thyroid irAEs (most common endocrine toxicity, 10–40%) reflect breakthrough autoimmunity: patients developing thyroiditis exhibit higher baseline T-cell reactivity and IFN-γ signatures, translating to both superior tumor control and autoimmunity risk (69, 195). Vitamin D sufficiency correlates with increased irAE incidence (particularly colitis, pneumonitis) but improved survival, likely via enhanced T-cell reinvigoration (125).

Obesity/leptin excess is associated with higher rates of severe irAEs, possibly via tonic mTOR activation, lowering activation threshold (196). Conversely, malnutrition (low albumin, PNI) predicts fewer irAEs but worse oncologic outcomes, reflecting globally suppressed immunity (197). Baseline GC use or endogenous hypercortisolism dramatically reduces irAE risk but abolishes ICI benefit (106). Thus, the same mechanisms driving resistance (exhaustion promotion) often confer irAE protection, while “fit” profiles yield both better tumor responses and manageable toxicity.

Routine, low-cost blood tests (albumin, prealbumin, CRP, vitamin D, ferritin, zinc, selenium, morning cortisol, TSH/free T4, fasting glucose/insulin or C-peptide, leptin/adiponectin ratio) combined with CT-based body composition (skeletal muscle index, subcutaneous/visceral fat) provide a feasible “immuno-nutritional panel” implementable in standard oncology practice (183, 198). PNI, CONUT score, and GPS require only CBC and chemistry, achieving high stratification power in ICI cohorts (136). Point-of-care or multiplex assays for adipokines and micronutrients are emerging, with turnaround < 48 h. In hematologic settings, adding serum-free light chains or hepcidin may refine marrow-specific risk (199).

Prospective trials increasingly support pre-emptive correction of deficiencies to augment ICI outcomes.

Risk-stratified supportive algorithms, intensive intervention for high-risk profiles (low PNI + vitamin D deficiency + sarcopenia) could increase population-level ICI benefit by 15–30% at minimal cost (13). Ongoing trials combine metformin + high-protein diet + vitamin D in “unfit” patients starting PD-1 therapy.

Composite scores merging host (PNI/GPS + vitamin D) with tumor factors (PD-L1 TPS, TMB, 9p24.1 status in cHL) outperform either alone, achieving > 85% accuracy in predicting durable response in retrospective validation (206, 207). Machine learning models incorporating sarcopenia, leptin, and IFN-γ signature are in development for hematologic ICI combinations (208).

This pragmatic framework, adaptable to resource settings, warrants prospective validation in phase III adjuvant/neo-adjuvant ICI trials in lymphoma and myeloma. Figure 6 outlines the proposed clinical workflow for endocrine–nutritional optimization before initiating PD-1/PD-L1 therapy. To operationalize the proposed assessment model, we provide a structured framework summarizing endocrine and nutritional biomarkers that can be incorporated into standard pre-treatment evaluation for ICI candidates (Table 6). These markers are chosen based on availability, clinical relevance, and potential to inform risk stratification and supportive interventions.

Table 6 presents a structured pre-treatment endocrine and nutritional assessment framework designed to optimize patient selection and risk stratification prior to initiation of ICI therapy. This framework integrates routinely accessible hormonal, metabolic, and nutritional parameters to capture baseline host immunometabolic competence, which is increasingly recognized as a determinant of ICI responsiveness and toxicity. Pre-existing endocrine abnormalities such as subclinical thyroid dysfunction, adrenal axis dysregulation, glucose intolerance, and sex hormone imbalance—may predispose patients to altered immune activation thresholds, exaggerated immune-related adverse events, or premature immune exhaustion, thereby influencing both therapeutic efficacy and tolerability. Concurrently, nutritional and body composition metrics including BMI, sarcopenia indices, serum albumin, prealbumin, PNI, and inflammation-linked nutritional scores serve as surrogates for systemic energy reserve, muscle-derived immunomodulatory signaling, and chronic inflammatory burden. By synthesizing these dimensions, the framework outlined in Table 6 moves beyond single biomarker evaluation toward a holistic, host-centered stratification model, enabling identification of patients who may benefit from prehabilitation strategies such as nutritional supplementation, endocrine correction, or metabolic optimization before ICI exposure. Importantly, this approach supports a shift from reactive management of immune-related complications to proactive personalization of immunotherapy, positioning endocrine–nutritional profiling as a pragmatic and scalable tool for improving clinical outcomes.

Although endocrine–nutritional signatures show strong biological plausibility and prognostic relevance, their clinical utility depends on feasibility, standardization, and cost-effectiveness. Importantly, many of the proposed biomarkers are already routinely measured in standard oncology practice, facilitating near-term translation without the need for specialized assays (210, 211).

Advancing endocrine–nutritional signatures for predicting PD-1/PD-L1 blockade efficacy in hematologic malignancies requires robust prospective cohort designs to overcome retrospective limitations, such as selection bias and incomplete data. Prospective studies enable standardized baseline assessments of hormones (e.g., cortisol, thyroid), nutrients (e.g., vitamin D, zinc), and composites like the PNI, alongside longitudinal monitoring of ORR, PFS, and irAEs. The PROVIDENCE study (2023), a prospective observational trial in advanced cancer patients on ICIs, demonstrated that systematic vitamin D supplementation improved ORR from 36 to 56% and median PFS from 5.8 to 11.3 months, while reducing thyroid irAEs. In hematologic contexts, adapting this approach could involve cohorts of relapsed/refractory cHL or MM patients, stratifying by marrow infiltration and prior therapies, which often induce cachexia and dysregulate the immuno-endocrine axis. Challenges include heterogeneity in disease subtypes and high attrition, but powering for endpoints like 12-month PFS could yield level I evidence, informing guidelines for pre-treatment interventions (125, 216–218).

Integrative biomarker panels enhance predictive accuracy by combining endocrine (e.g., insulin/IGF-1), nutritional (e.g., albumin, ferritin), inflammatory (e.g., CRP/neutrophil-lymphocyte ratio), and body composition metrics (e.g., sarcopenia via CT). Panels like the Systemic Immune-Inflammation Index predict poorer OS/PFS in ICI-treated cancers, reflecting T-cell exhaustion and myeloid suppression. In hematologic malignancies, the Systemic Immunologic Readiness Score integrating GPS, PNI, and microbiome data stratifies “fit” vs. “unfit” patients, with high scores correlating to superior ORR in cHL. Preclinical models validate synergy: hypercortisolism plus hypoalbuminemia triples resistance via mTOR/IFN-γ inhibition. Hematology-tailored panels could incorporate hepcidin for iron dysregulation in MM, enabling targeted therapies like metformin or nutritional support to boost efficacy (219–223). AI-based prediction models revolutionize personalization by integrating multi-omics (genomics, metabolomics) with endocrine–nutritional data. Machine learning algorithms, such as random forests, predict ICI response with > 85% accuracy using routine blood parameters in pan-cancer cohorts. In 2025 updates, AI models for PD-1/PD-L1 efficacy in hematologic malignancies incorporate PD-L1 expression estimation from H&E slides, forecasting PFS in NSCLC analogs adaptable to lymphomas. Federated learning across registries (e.g., LYSA) handles real-time inputs like cortisol rhythms, identifying non-responders for adjunctives. Ethical issues like bias require validation in trials (e.g., NCT05352750), but these models could optimize r/r settings, addressing gaps in tumor-centric biomarkers (224–228).

Future studies should address ethnic and geographic variability in pre-treatment endocrine–nutritional signatures to enhance the global applicability of these predictive biomarkers for PD-1/PD-L1 blockade. Baseline levels of key components such as vitamin D, iron stores, thyroid hormones, cortisol, and nutritional indices vary substantially across populations due to differences in latitude, sunlight exposure, dietary patterns, genetic polymorphisms, socioeconomic factors, and healthcare access. For example, vitamin D deficiency is highly prevalent in populations residing at higher latitudes and among individuals with increased skin pigmentation, cultural sun-avoidance practices, or limited dietary fortification, potentially influencing immune competence and immunotherapy responsiveness. Similarly, regional differences in iron deficiency or overload, micronutrient availability (e.g., zinc and selenium), and sarcopenia prevalence may differentially shape systemic immunologic readiness across ethnic groups. Prospective, multi-ethnic cohorts are therefore needed to define population-specific reference ranges and thresholds for these signatures, as well as to determine whether their predictive value is consistent across diverse hematologic malignancies and treatment settings. Incorporating geographic, lifestyle, and genetic modifiers into predictive models may enable more accurate patient stratification and support the development of tailored nutritional or endocrine interventions to optimize immunotherapy outcomes worldwide.

Endocrine markers display marked circadian variability, with cortisol peaks at 06:00–08:00 and nadirs at midnight, with flattened rhythms prognostic of poorer survival in multiple cancers necessitating standardized morning sampling that is rarely enforced in retrospective cohorts (231, 232). The evidence base remains predominantly observational, with few prospective interventional trials: vitamin D repletion studies show promise but are limited by small size and lack of randomization, while nutritional prehabilitation lacks phase III data in ICI settings (233, 234). Finally, the absence of consensus cut-offs hampers comparability. Albumin thresholds range 30–35 g/L, PNI 40–50, vitamin D 20–30 ng/mL across studies reflecting population-specific inflammation burdens and assay differences (235). These gaps underscore the need for large, prospective, biomarker-driven trials incorporating serial sampling, confounder adjustment, and harmonized thresholds to establish causative links and enable routine clinical use (236, 237).

The mechanistic frameworks we describe draw from studies with diverse designs, patient populations, and endpoints. Biomarker associations identified in retrospective cohorts may not fully capture causal relationships and may reflect underlying confounders. Moreover, different studies employ varying assay platforms, cut-offs, and composite index definitions, which complicates direct comparisons and quantitative synthesis.

Several mechanistic insights and prognostic associations derive from small or single-institution cohorts. These studies may be underpowered to detect modest effects, and findings may not generalize across broader demographic or disease subgroups. The limited sample sizes also restrict our ability to evaluate interactions among biomarkers, treatment regimens, and clinical outcomes.

Although many endocrine–nutritional pathways have plausible roles in modulating immune function and treatment response, some proposed mechanisms remain speculative due to incomplete experimental validation. For example, the interplay between adipokines, systemic metabolism, and antitumor immunity has strong preclinical support but requires further confirmation in longitudinal clinical studies with standardized sampling.

As discussed above, biomarker measurements may vary with circadian rhythms, acute illness, and assay methodology. Most published studies do not report standardized timing or repeated measures, which could influence observed associations. This limitation is particularly relevant for hormones with diurnal variation or for markers influenced by concurrent medications (e.g., corticosteroids).

The manuscript’s reliance on retrospective cohorts introduces significant biases, including selection bias from non-randomized patient inclusion, confounding by unmeasured variables (e.g., comorbidities or prior therapies), and incomplete data on endocrine-nutritional profiles, potentially overestimating associations with PD-1/PD-L1 outcomes. For instance, real-world studies in hematologic malignancies often lack standardized biomarker assessments, leading to inconsistent findings on predictors like hypoalbuminemia or vitamin D deficiency. Small sample sizes further compound issues, limiting statistical power and increasing the risk of spurious results or type II errors, as seen in early-phase ICI trials where cohorts under 100 patients fail to detect subgroup effects in diverse hematologic subtypes. Overrepresentation of Asian populations in cited studies (e.g., higher EGFR mutation rates influencing immunotherapy responses) may reduce generalizability to Western cohorts, where genetic and environmental factors differ, potentially skewing efficacy estimates for endocrine signatures (238–242). Future research should prioritize prospective, multicenter trials with larger, diverse cohorts to mitigate biases and validate biomarkers like PNI or GPS. Strategies include integrating multi-omics (e.g., metabolomics with nutritional data), machine learning for predictive models, and global collaborations to ensure ethnic representation, enhancing translational applicability in hematologic settings (226, 243–245).

Prospective, multicenter studies with standardized protocols are needed to validate promising signatures and to clarify causal mechanisms. Harmonized reporting standards and collaborative consortia will facilitate meta-analyses that overcome individual cohort limitations. Integration of mechanistic studies with clinical outcomes will strengthen the biological rationale and translational potential of endocrine–nutritional markers. By acknowledging these constraints, our review aims to present a balanced interpretation of current evidence while highlighting opportunities for future research.