We evaluated 13,619 oncological patients affiliated with Oncosalud-AUNA, treated between 2000 and 2019 for 20 different sex-unrelated types of cancer. The type of cancers, ICD codes, and number of patients are shown in Supplementary Table S1 .

OS was estimated from the date of cancer diagnosis until the date of death or last follow-up. The follow-up for OS was conducted up to January 2020, using the National Civil Registry (RENIEC). To evaluate the influence of sex in OS, HR for death of female vs. male patients in each specific tumor was estimated and the proportional hazard assumption verified ( Supplementary Figure S1 ). The meta-analysis using random effects was conducted through the software RevMan 5.4 (19).

Paraffin-embedded tumor tissues were collected from advanced or metastatic MEL (n = 104), CRC (n = 208), and NSCLC (n = 291) patients diagnosed at Oncosalud-AUNA. Tumor cell content was assessed and ranged from 15% to 90% through examination of hematoxylin and eosin-stained slides by a pathologist. A commercial reference standard, Horizon Quantitative Multiplex Reference Standard HD200, was tested to validate the performance of next-generation sequencing (NGS) for detection of somatic mutations.

Eight formalin-fixed paraffin-embedded (FFPE) 8-μm-thick tissue sections were cut from FFPE tumor samples. DNA was extracted using the ReliaPrep FFPE gDNA Miniprep System (Promega, Madison, USA) following the manufacturer’s protocol. The DNA concentration was determined by fluorometric quantitation using Qubit 4.0 Fluorimeter with Qubit dsDNA HS Assay Kit (Invitrogen, USA).

The chosen panel targeted single-nucleotide variants (SNVs) and insertion/deletions (indels) in the following genes: BRAF, EGFR, KRAS, NRAS, PIK3CA, and TP53. Libraries were prepared using the AmpliSeq Focus Panel and AmpliSeq Library Plus (Illumina, San Diego, USA) following the manufacturer’s protocol without modifications using a total of 10 ng input DNA per sample. Multiplex polymerase chain reaction (PCR) was performed in 20 cycles. Sequencing adapters with unique indexes (AmpliSeq CD Indexes Set A for Illumina) were ligated to the amplification products and purified using Agencourt AMPure XP beads (Beckman Coulter, CA, USA) according to the manufacturer’s instructions. Libraries with 2-nM molarities were subjected to clustering using a standard flow cell and were sequenced on the MiSeq platform (Illumina) using the MiSeq Reagent Micro Kit v2 (300 cycles).

Raw data were processed automatically on the BaseSpace Sequence Hub (Illumina) and aligned to the hg19 reference genome. An average of 93.45% (87.2–99.7%) on-target reads, 95.2% (91.3–99.1%) read uniformity, and 500× average coverage were obtained per sample. The default limit of detection (LOD) was set at 5% allelic frequency (VAF).

BaseSpace Variant Interpreter (Illumina) was used to annotate and interpret genetic variants. Genetic variants were annotated in accordance with the nomenclature of the Human Genome Variation Society (HGVS). The interpretation was done using the Single Nucleotide Polymorphism Database (dbSNP, http://www.ncbi.nlm.nih.gov/projects/SNP/), ClinVar database, InSIGHT/LOVD database, and COSMIC. Integrative Genomics Viewer was applied to visualize the variants. All identified variants were checked with VarSome Clinical (Saphetor, Suiza). The interpretation about the pathogenicity of the variants followed the latest recommendations of the American College of Medical Genetics (ACMG), European Society for Medical Oncology (ESMO), Scale for Clinical Actionability of molecular Target (ESCAT), and OncoKB.

Eighteen datasets of eight types of cancers including CC, gastric (GC), HNC, LC, MEL, PC, CCRCC, and TC were retrieved from a public repository (https://www.ncbi.nlm.nih.gov/gds). Datasets with less than seven patients or datasets containing metastatic tumors were excluded. In the included datasets, secondary tumors, metastatic tumors, normal tissue, xenografts, and cases without sex information were excluded ( Table 1 ).

Gene expression values were log2-transformed and median centered before analysis. In GC datasets, the “null” value was replaced by 0.000001.

The GSEA Analysis was done using the java GSEA Jar application (20), and 4,436 biological processes included in the Gene ontology version 6.2 (c5.bp.v6.2.symbols.gmt) were analyzed. Both are available at http://software.broadinstitute.org/gsea/downloads.jsp.

Patterns of immune gene sets were compared between tumors from female vs. male patients. We used gene set permutation without additional normalization and default parameters. The analysis was conducted individually in each dataset. A gene set was considered enriched when it was present in at least three cancer types with a p-value < 0.05.

The significantly overexpressed processes were grouped according to the Gene Ontology Browser (http://www.informatics.jax.org/vocab/gene_ontology/). The biological processes included in each group are detailed in Supplementary Data S1 .

The study was approved by the Ethical Review Board of the Universidad Privada San Juan Bautista (084-2021-CIEI-UPSJB) and conducted in compliance with all relevant ethical guidelines.

The meta-analysis of 13,619 patients (54.4% women; 45.6% men) that pooled HR from the analysis of 20 different types of cancer determined a decreased risk of death in female patients (HR = 0.73, CI95: 0.58–0.92). Significant differences in the risk of death were observed in HNC (HR = 0.75, CI95: 0.58–0.97), leukemia (HR = 0.33, CI95: 0.26–0.43), LC (HR = 0.74, CI95: 0.64–0.85), lymphoma (HR = 0.67, CI95: 0.55–0.83), MEL (HR = 0.46, CI95: 0.31–0.69), MM (HR = 0.58, CI95: 0.43–0.78), and non-melanoma skin cancer (HR = 0.44, CI95: 0.43–0.44) ( Figure 1 ).

In total, 301 LC patients were evaluated, of which 160 were women (53.2%). A significant association between sex and mutational status was found in LC with regard to EGFR mutations, namely, 22.7% in men vs. 34.4% in women (p = 0.035) ( Figure 2 ). On the contrary, the mutational status of BRAF in MEL (p = 0.422), KRAS (p = 0.341), BRAF (p = 0.895), and PIK3CA (p = 0.704) in CC patients was not related to sex ( Table 2 and Supplementary Data S2 ).

The number of immune processes overexpressed in women was higher than in men in GAC (218 vs. 48), HNC (170 vs. 51), LC (155 vs. 0), CC (151 vs. 3), MEL (89 vs. 0), and PC (85 vs. 0) ( Supplementary Data S1 ). Overall, 22 categories of gene ontology for immune processes were enriched in tumors from women ( Figure 3 ).

According to the Gene Ontology groups, the processes that were significantly enriched in women in all evaluated malignancies were related to response to stimuli, multiorganism process, localization of immune cells, immune system development, immune response, cellular processes, and activation of an immune response.

In response to stimulus pathways, the most frequent gene sets were those associated with interferon gamma (IFN-gamma), inflammatory response, positive regulation of defense response, and response to tumor necrosis factor (TNF). On the other hand, in the multiorganism process, gene sets related to innate immune response and inflammatory response were overexpressed.

In addition, processes related to the localization of immune cells included enriched expression of genes involved in leukocyte migration, lymphocyte migration, and T cell activation related to immune response.

The immune system development process was characterized by the overexpression of datasets related to lymphocyte, dendritic cell, and leukocyte differentiation; innate immune response; activating cell surface receptor signaling pathway; lymphocyte activation; positive regulation of immune effector processes; and regulation of antigen-receptor-mediated signaling pathways.

With regard to the immune response process, positive regulation of cytokine production, innate immune response, and immune cell activation as well as negative regulation of cytokine production were the most frequent gene sets.

Furthermore, in cellular processes, responses to IFN-gamma and IFN-gamma-mediated signaling pathway; cytokine-mediated signaling pathway; cell surface receptor signaling pathways; regulation of leukocyte, lymphocyte, and natural killer cell activation; Fc-gamma receptor signaling pathway; and Fc-receptor signaling pathway, among others, were overexpressed.

The most frequent gene sets in the activation of immune response processes were antigen-receptor-mediated signaling pathway, activation of immune response, and T cell-receptor signaling pathway.

In total, 18 immune processes were enriched in tumors from male patients. In contrast to other cancers, in TC there was a slightly more frequent expression of IP in men compared to women (82 vs. 61), whereas in CRCC, the enrichment in IP was significantly increased in men (184 vs. 23).

The most frequent gene ontology groups were cellular processes, multiorganism processes, immune response, and immune system development ( Figure 4 ). Cellular processes included positive regulation of cell activation; cellular response to IFN-gamma-, chemokine-, and cytokine-mediated signaling pathways; leukocyte and lymphocyte chemotaxis; and cellular response to Interleukin 1 (IL-1). In the multiorganism process, inflammatory response, positive regulation of defense response, response to IFN-gamma, and response to bacteria were the most frequently enriched gene sets.

Regarding immune response processes, the most frequently overexpressed gene sets were regulation of cytokine secretion, positive regulation of immune response, activation of immune response, regulation of adaptive immune response, B cell-mediated immunity, and humoral immune response. In addition, in the immune system development process, positive regulation of cytokine production, regulation of T cell proliferation, and leukocyte and lymphocyte differentiation, among others, were overexpressed.