As demonstrated earlier, males generally have higher incidence and mortality rates than females for bladder, colorectal, liver, lung, oesophagus, and stomach cancers (29). These sex disparities cannot solely be explained by the biological sex; lifestyle and environmental exposures are indispensable as well. In the UK, excluding sex-specific cancer types, modifiable risk factors account for 36.4% of male cancer cases and 25.6% of female cases. Tobacco smoking alone contributed to 15% of preventable cancer cases in the UK in 2015 and represents the highest proportion of preventable cancer cases in the US and Australia (30). Male and female smokers are 23 and 13 times more likely to develop lung cancer compared to non-smokers, respectively (31). Chronic alcohol consumption is also strongly linked to various cancers, with dose-response relationships seen in multiple epidemiological studies for liver, colorectal and upper aerodigestive tract cancers (32–34). Subgroup analyses in people with alcohol use disorders have shown that females have a higher risk of developing cancers compared to men (OR=1.767) (35). Additionally, consuming the same amount of alcohol leads to a greater increase in absolute lifetime cancer risk for women (1.4%) compared to men (1%), although these higher cancer rates in women may be attributed to breast cancer (36).

Obesity represents a major public health challenge, with approximately 55% of cancers in females and 24% in males in the USA considered obesity related. Importantly, 42% of new cases of overweight and obesity related cancers are gynaecological and breast cancers. This implies a stronger correlation between high body-mass index (BMI) and female cancers, highlighting the role of aromatase and oestrogen in gynaecological and breast cancer development (37). Non-sex specific cancers have a higher incidence in males, particularly oesophageal (male to female ratio of adenocarcinoma 4.4, squamous cell carcinoma 2.7) and colorectal cancers (38, 39). Obesity also plays a role in the tumorigenesis of these cancers, possibly involving chronic inflammation and systemic insulin and adipokine dysregulation (40, 41) that raise the incidence of metabolic syndrome (including metabolic dysfunction-associated steatotic liver disease, MASLD) particularly in males (42).

Globally, oncogenic viruses contribute to approximately 10% of all malignancies, although this varies between higher and lower income countries (43). The population attributable fractions are higher in females than males, primarily due to the inclusion of sex-specific cancers. Causative agents include human papillomaviruses (HPV), hepatitis B/C viruses (HBV/HCV), Epstein-Barr virus (EBV), and human immunodeficiency virus (HIV) (30). HPV⁺ head and neck squamous cell carcinoma (HNSCC) (44), EBV-driven NPC (24), as well as HBV/HCV-driven hepatocellular carcinoma (HCC) (45) all show a strong male predominance.

Sex chromosome differences may also contribute to variations in cancer incidence. Females have XX and males have XY sex chromosome combinations, while intersex individuals such as those with Turner’s or Klinefelter’s syndrome have chromosomal patterns deviating from the typical configurations. In XX individuals, some pseudoautosomal genes can escape X-chromosome inactivation (XCI) providing a “buffering” effect against allele mutations. Incomplete XCI occurs in 23% of X chromosome genes (46). Thus, a single allele mutation leads to complete alteration of gene function in males, as opposed to a heterozygous alteration in females. This serves as a safeguard, preserving tissue function in the presence of mutations. Many of these genes, including ATRX, KDM5C, KDM6A, and MAGEC3, have tumour suppressor functions. Additionally, mutated alleles on the inactive X chromosome are typically expressed at lower levels or not expressed at all, mitigating their impact on cellular function (47). In females, the selective proliferation of specific mosaic subpopulations exhibiting preferential expression of one X chromosome can lead to skewed XCI (48). This can confer advantageous immunomodulation against cancer – a protective mechanism not available to males who obligatorily express the same mutated maternal X-linked gene.

X-linked genes, including HUWE1, FLNA and MED12, can directly modulate TP53 expression. This association may render males at a higher risk of p53 dysfunction. Females exhibit a higher incidence of non-expressed mutations among p53-associated X-linked genes. Bioinformatic analyses in 12 non-reproductive cancers have shown that in females, less than half of these exome mutations were transcribed into mRNA, whereas the majority underwent mRNA transcription in males (49). These findings suggest tumour suppressor effects of the X chromosome.

Loss of Y chromosome (LOY) has been implicated in the pathogenesis of lung cancer, renal tumours and up to 40% of bladder cancer (50–52). In muscle invasive bladder cancer, patients exhibiting low Y chromosome gene expression of KDM5D, KDM6C, TBL1Y and ZFY demonstrate worse prognosis (52). Mosaic LOY in peripheral leukocytes is also associated with solid tumour incidence. Extreme downregulation of Y is linked to increased cancer risk and resistance against EGFR tyrosine kinase inhibitors (53), which may also impact immunotherapy response downstream. Loss of the entire X chromosome(s) has been documented in early-stage astrocytoma, neuroblastoma and medulloblastoma (54–56).

The link between oestrogen or ER and non-reproductive cancers is unclear. At the molecular level, oestrogen and ER affect PD-1 signalling, Wnt/β-catenin pathways and the Ras/MAPK pathway, amongst many other aspects of cancer biology (57–62).

Circulating E1 (oestrone) and E2 (oestradiol) levels were found to have no statistically significant relationship with colon cancer in a cohort of 1000 postmenopausal women (62). However, a 2015 meta-analysis revealed a reduced ratio of ERβ expression in CRC compared to the normal mucosa (OR=0.216), associated with poorer overall and disease-free survival (63). Conversely, exogenous oestrogen reduces the risk of CRC by 37% as demonstrated by the landmark Women’s Health Initiative study (64). In vitro, ERβ was shown to modify the hypoxic response by downregulating HIF-1α, VEGFA and PDGF (65).

Oestrogen plays a complex role in the liver. It has been implicated in various liver pathologies like fibrosis and fatty liver disease, but its role in HCC remains unclear. In a cohort of 275 men, higher total E2 is associated with increased HCC risk (OR=1.58) (66). A recent cohort study shows a survival advantage for female HCC patients over males in perimenopausal and early-menopausal ages but not in postmenopausal women, possibly due to declining endogenous oestrogen production (67). However, female patients in phase III trials for immune checkpoint inhibitors (ICI) for HCC are found to have worse overall survival (OS) than males (68). Whether this discrepancy can be attributed to oestrogen is unclear. Studies exploring the use of tamoxifen in HCC have yielded mixed results, with some showing prolonged survival but larger studies finding no significant association (69–71).

In lung cancer, oestrogen appears to have a protective effect. A meta-analysis of female lung cancer cases demonstrates that higher levels of sex steroid hormone exposure, both endogenous and exogenous, reduce lung cancer risk by 10% (72), yet the role of ERα or ERβ is unclear. Some studies suggest that ERα is associated with worse prognosis in non-small cell lung cancer (NSCLC) (73), while others find no significant effect. Some meta-analyses indicate an association between ERβ and better prognosis in NSCLC (73, 74), while others consider it an unreliable prognostic marker (75, 76) depending on the methods employed, such as uni- vs multivariate analysis, bioinformatics, or immunohistochemistry (IHC) analysis. Finally, female reproductive factors like breastfeeding are associated with a decreased risk of oesophageal and gastric adenocarcinoma, though parity, menstruation, and the use of hormone replacement therapy have no association (77). Interestingly, the use of tamoxifen, a selective oestrogen receptor modulator (SERM), is associated with an increased risk of gastric adenocarcinoma (78) as well as endometrial cancer. The tissue-specific agonist/antagonist role of SERMs like tamoxifen reflects the complex role of the oestrogen-ER signalling axis in tumorigenesis.

Androgens include testosterone, dihydrotestosterone (DHT), and dehydroepiandrosterone (DHEA), among others. Testosterone produced by the testes plays a pivotal role in initiating the development of masculine traits, hence exists in higher levels in males and lower in females. Androgen deficiencies in males can result in the development of feminine traits (79), while increased androgen production in females can lead to a shift from feminine to masculine traits and also be associated with polycystic ovarian syndrome (PCOS) (80). Their biological functions are executed by binding with AR and activating intracellular AR signalling downstream. Besides prostate cancer, the role of androgens in tumorigenesis is less studied compared to oestrogen. Higher concentrations of testosterone are associated with increased risk of liver cancer, particularly in men, while higher levels of DHEA, the adrenal precursor, are associated with a 53% decrease in risk (66, 81). Higher circulating testosterone is associated with a decreased risk of CRC in men, but this is not shown in women (81). The association between testosterone and oesophageal cancer is unclear, with varying degrees of significance across studies (81, 82). Gastric, pancreatic and bladder cancers are also shown to have no significant association with testosterone levels (81). Interestingly, androgen deprivation therapy (ADT) using finasteride has shown improved survival in patients with non-muscle invasive bladder cancer, suggesting a potential strategy to reduce bladder cancer incidence and recurrence (83).

AR is a member of the nuclear receptor superfamily acting as a ligand-dependent transcription factor (84). Consisting of eight exons, the AR gene is located on the X chromosome. It comprises a ligand-binding domain (LBD), a DNA-binding domain (DBD), and an N-terminal domain (NTD). In the unbound state, AR forms a complex with co-chaperones, heat shock proteins, and cytoskeletal proteins in the cytoplasm. Ligand binding induces conformational changes, receptor dimerization, and translocation to the cell nucleus. The NTD influences transcriptional activity, while the DBD allows binding to and recognition of androgen response elements (AREs) on target genes where it serves to induce or repress gene expression through binding to chromatin at cis AREs (85). AR can also modulate post-translational modifications by phosphorylation, methylation, or ubiquitination (86, 87) (Figure 3). While AR exerts effects mostly in sex hormone-dependent tissues, such as the prostate, testes, ovaries, and endometrium (88, 89), it is also widely expressed in kidneys, liver, urinary bladder, as well as the cardiovascular, immune, musculoskeletal and nervous systems (88, 90–94). It is also noted that membrane androgen receptors (mARs), such as ZIP9 and GPRC6A, are a group of G protein-coupled receptors that directly alter cellular signalling upon androgen stimulation, also known as the non-genomic pathway (95, 96) (Figure 3). While studies have demonstrated the implications of mARs on prostate cancer, they are beyond the scope of this review.

A report of teenagers developing hepatocellular carcinoma due to excess androgen intake have spurred interest in the effect of androgen and AR on cancer (97, 98). In 1980, an article published in The Lancet highlighted the association between elevated levels of free testosterone in males and an increased risk of melanoma (99). While multiple observations support the hypothesis that excess androgens may be tumorigenic (100), a definitive mechanistic explanation is still lacking, which necessitates our summary of current knowledge below.

AR signalling is the primary driver of castration resistant prostate cancer (CRPC) (101). Enzalutamide, an AR antagonist, competes with androgens to bind to AR and blocks nuclear ARE binding, thereby inhibiting downstream transcriptional activity (102) and enabling antitumor effect (103). AR and ER exhibit similarities as nuclear receptors, allowing substantial signalling crosstalk (Figure 3) (104). In ER+ breast cancer, AR competes with ER for oestrogen response elements (EREs) and inhibits ER activity, playing a tumour-suppressive role especially in premenopausal patients (105). However, AR may promote cancer progression in certain ER– breast cancers. A study indicated that the luminal AR (LAR) subtype accounts for 15% of triple-negative breast cancer and AR is an attractive therapeutic target (106). Higher AR expression and corresponding aggressive phenotypes are observed predominantly in tissue samples from African American women, with a strong interaction between AR and JAK-STAT signalling (107). Another study shows that PIK3CA is highly mutated in the LAR subtype, where PI3K inhibitors can reduce LAR cell proliferation (108). Salivary duct carcinoma (SDC), a male-dominant cancer, is a rare, aggressive malignancy also characterised by high AR expression, ranging from 70% to 97.8% (109–111). Recent studies have found that the occurrence of SDC is closely related to the AR signalling pathway (112), sharing similar molecular profiles with high-grade breast ductal carcinoma and apocrine breast cancer (113). AR-V7, an AR splicing variant, accounts for over 50% of AR in SDC and plays a crucial role in the resistance and progression in CRPC (114). Other studies have reported that FOXA1 mutations are present in 10% of SDC cases, resulting in drug resistance and tumour progression also via the AR pathway (113).

With its homolog crucial for primary sex determination in C. elegans (115), FOXA1 is a key transcription factor necessary for AR and ER activities in prostate and breast cancers (116). AR driven transcription in molecular apocrine breast cancer is mediated by FOXA1 (117). In prostate cancer, FOXA1 exhibits a high mutation rate, thereby affecting AR transcription (118). Elevated levels of FOXA1 have been associated with poor prognosis in prostate cancer. FOXA1 function in AR signalling and its impact on prostate cancer differs markedly from its role in ER signalling and breast cancer progression (119). A study published in 2012 highlights the significance of FOXA1 and FOXA2 in sexual dimorphism in liver cancer, noting that modulation of these factors can reverse the observed gender differences (120). Other FOX family genes are also crucial in regulating the PI3K-AKT-mTOR pathway. FOXO3a, a PI3K/AKT downstream substrate, can induce AR expression as a positive regulator (121). FOXO1, a downstream effector of AR, can also lead to AR hyperactivation in prostate cancer with PTEN loss, independent of androgen binding (122).

The crosstalk between AR and other signalling pathways has also been reported (123, 124). With a strong association between nuclear AR expression and Wnt/β-catenin signalling in bladder cancer, ADT has shown great therapeutic potential (124). Moreover, TCF1 and AR have overlapping binding sites on β-catenin (125). β-catenin translocates into the nucleus and interacts with TCF1 and lymphoid enhancer factor, activating the transcription of target genes. TCF1 is required for the self-renewal of stem-like CD8+ T cells in response to viral or tumour antigens, preserving heightened responses to checkpoint blockade immunotherapy (126). This implies not only a causal relationship between AR signalling and tumour progression via β-catenin pathways, but also a connection between AR and antitumor immune responses (more in Section 5.2). In addition, androgens can also influence the effectiveness of BRAF-targeted therapy in melanoma. AR expression is elevated in BRAF-resistant melanoma. Inhibition of both the AR and BRAF/MEK pathways counteracts resistance and hence improves cytotoxicity (127). Intriguingly, blocking AR not only inhibits the proliferation of BRAF-resistant cells, but also enhances the infiltration of CD8+ T cells and promotes cancer cell apoptosis (128). This prompts further investigation on how AR affects immune responses, and targeting AR may offer new combination therapies for cancer treatment.

There are several meta-analyses evaluating the comparative efficacy of immuno-oncology (I/O) on various cancers across genders (Table 1). A 2018 meta-analysis summarises 20 clinical trials involving ICIs across various cancer types, with a total of 11,351 participants (129). These trials predominantly focus on melanoma (32%) and NSCLC (31%). The meta-analysis reveals significant sex differences in clinical outcomes, where females experience lower response rates than males. However, the significant heterogeneity calls for analysis specific to individual cancer types and treatments. In 2019, the same team conducted another meta-analysis of chemotherapy and I/O for advanced lung cancer; this time with opposite conclusions compared to a year ago (130). Women with advanced lung cancer seem to derive a larger benefit from the addition of chemotherapy to anti-PD-1/PD-L1 compared with men. Another meta-analysis on NSCLC patients receiving combination chemo-immunotherapy first-line also concludes that females show a more significant improvement in OS and progression-free survival (PFS) (132). These findings highlight the potential impact of gender on the effectiveness of both targeted and combination of chemo-immunotherapies in NSCLC.

In 2020, a meta-analysis on NSCLC includes 13 studies with monotherapy and 5 with combination regimens (KEYNOTE 010/024 with pembrolizumab versus chemotherapy and CHECKMATE 017/026/057 with nivolumab versus chemotherapy), a total of 1028 female and 1435 male patients (131). The result confirms that EGFR wild-type patients could benefit from immunotherapy monotherapy (HR=0.77; p<0.001) while those of mutant types experienced no survival benefit (HR=1.11; p=0.54). While EGFR mutations are more likely to occur in females (134), there is no apparent efficacy-sex association overall (131). Therefore, to explore the effect of AR in I/O efficacy, confounding factors such as mutations will need to be properly controlled and stratified.

There are no sex differences in the superior OS benefits from first-line ICI-based combination therapies in metastatic renal cell carcinoma (RCC) or metastatic urothelial carcinoma (UC) (133). In locally advanced RCC, however, adjuvant I/O monotherapy reduces recurrence risk in female patients (HR=0.71, 95% CI 0.55–0.93) but not in male patients (133). On the other hand, males with muscle-invasive bladder cancer have better DFS on adjuvant I/O compared to females (133). A meta-analysis on HCC shows single-agent I/O exhibits less OS benefit in females than males. On the other hand, combination atezolizumab-bevacizumab (Atezo/Bev) – a first-line standard of care for advanced HCC – yields comparable efficacy between males and females in a real-world cohort (68). Nevertheless, these studies did not conduct stratified analysis based on AR expression, which may have overlooked the importance of AR in sex disparity.

Other phase III trials in advanced urothelial, hepato-pancreato-biliary and upper or lower gastrointestinal tract cancers have also been individually screened for outcome differences in patients treated with ICI based on sex or AR levels (135–145). However, none of these trials made explicit analysis on how sex or AR affects the outcomes in these cancers. Importantly, though, one study on 23,296 patients enrolled in SWOG trials shows a 49% increased risk of adverse events (AE) in females receiving I/O, especially of haematological AEs (146, 147). It is hoped that more prospective studies on the relationship between sex, AR expression and I/O efficacy and AEs can be carried out to further explore the role of AR signalling in cancer immunology and immunotherapy.

While previous sections have attempted to address how genetic, environmental, and hormonal effects lead to distinct tumorigenesis and disease progression patterns between males and females, studies over the past decade have emerged to explain how AR within the tumour microenvironment (TME) conspire to this process and alter patient response to treatments such as ICIs. This section summarises the effect of AR signalling in different TME cell types, laying the foundation for subsequent discussion on another dimension of I/O resistance (Figure 4).

Although AR signalling plays a key role in tumour immunosuppression, it is important to note the caveat when interpreting preclinical studies involving AR and biological sex. Cells harvested from male subjects have long been exposed to androgens and AR signalling. Hence, when manipulating AR-related pathways, male cells may behave very differently compared with female cells. Consequently, when designing and analysing clinical trials for I/O and antiandrogen combination, it is indispensable to include stratification based on biological sex, along with other variables such as circulating androgens, AR mutation/amplification/IHC status, and PD-L1 scoring.

Patient scRNA-seq and murine models have suggested that an increased AR signalling may predict I/O resistance, resulting from downregulation of IFNγ and upregulation of CD8⁺ T cell exhaustion programmes. Indeed, as previous sections have shown, a recurrent in vivo finding is that castration or T cell-specific AR knockout can improve I/O response in male mice, while antiandrogens rescue I/O response and tumour control in androgen-exposed females. While it is natural to test I/O-antiandrogen combinations in the clinical setting, I/O nevertheless fails to synergise with AR antagonists in metastatic CRPC after all, as evident in the IMbassador250 trial (194). Why is this?

One explanation, as discussed earlier, is that male CD8⁺ T cells have experienced long-term androgen exposure, predisposing them towards exhausted phenotypes during tumour progression, irrespective of subsequent AR signalling manipulation. In preclinical studies, castration or cell-specific AR knockout is almost always performed before tumour inoculation and I/O treatment. The dynamics of interaction between malignant cells and the TME may well be different from research involving antiandrogens. It reminds us that the sequence of I/O versus AR signalling manipulation is crucial to an optimised patient response.

Another hypothesis is that AR antagonists suppress anti-tumour immunity independently of AR. One study has shown that AR antagonists inhibit initial T cell priming via an off-target effect on GABA-A (195). Even if T cell exhaustion may be reduced with antiandrogen treatment, the initial neoantigen presentation and infiltration into the tumours can also be compromised, cancelling out the beneficial effect of AR antagonist on checkpoint inhibition. Indeed, another study also shows increased monocytic MDSC infiltration, decreased CD8+ TIL number and increased PD-L1 expression in enzalutamide-treated murine Myc-CaP tumours (168). When these tumour cells acquire enzalutamide resistance, they upregulate PD-L1 expression and possess an increased capacity to skew myeloid cells towards MDSCs and M2 macrophages (168, 195), further suppressing T cell function. Strikingly, another study shows a signalling crosstalk between AR and the glucocorticoid receptor (GR) (196). AR inhibition upregulates GR while high-dose steroids confer enzalutamide resistance to a prostate cancer model (196). This finding necessitates a more thorough understanding of the escape mechanisms of tumour cells when treated with combined I/O and antiandrogen (101).

Also importantly, as evident in previous sections, AR signalling exhibits heterogeneous effects on different TME cell types, resulting in equivocal efficacy when combining I/O with systemic antiandrogen administration. While AR on lymphocytes (T, B, NK) negatively regulates their cytotoxic functions in general, AR on macrophages and neutrophils regulate their functions in a sequence-dependent manner. Specifically, AR promotes the proliferation, maturation and infiltration of macrophages and neutrophils into the tissues. However, it subsequently renders these cells anti-inflammatory in the TME. AR inhibition also enhances the immunosuppressive functions of DCs and MDSCs. Furthermore, increasing evidence has shown that AR prevents fibroblast differentiation towards CAFs and regulates endothelial cell proliferation. Therefore, there is much unknown as to how systemic AR inhibition on a heterogenous TME affect immunotherapy efficacy. Interestingly, a recent analysis of NSCLC exosome and transcriptome datasets show significant enrichment of DCs and T cells as well as a T cell dysfunction phenotype in the TME of female patients, while the male patients generally possess a T cell excluded TME (197). These findings are highly consistent with the effects of AR signalling on TME cell types as described earlier, demonstrating the key role of AR in regulating tumour immunology and I/O response. Future combinatory I/O with AR modulation will require delicate consideration into the individual tumour characteristics.

Specifically in prostate cancer, preclinical studies have shown, as discussed above, how blocking AR signalling can in fact compromise T cell priming or activation (195), upregulate CCL2/STAT3-mediated macrophage recruitment (163), reduce neutrophil maturation or expansion (174), promote CAF accumulation (184), and increasing tumour cell expression of GR (196), all of which may negate the benefits of checkpoint blockade in these patients (198). Future combinatory trials in advanced prostate cancer will need to select patients early in their disease progression, and give careful thoughts on both the checkpoint (PD-1, PD-L1, CTLA4, TIGIT, etc.) to be targeted, as well as the timing of I/O relative to AR inhibition (198, 199).