Background and methodological approach to investigating sex differences in cancer incidence

A portion of this doctoral work was published. Read the paper

1. Understanding the problem

Cancer is the second leading cause of death globally and its prevention is one of the most significant challenges in the 21st century.

Sex differences in cancer incidence are well documented across populations, with men generally at higher risk than women for cancers that are common to both sexes.

Source: IARC Global Cancer Observatory, Cancer Over Time. Age-standardised incidence rates for men and women in England, 1971 to 2017.
  1. Epidemiological studies reported men generally have 3x higher risk compared to women for cancers of the bladder, oesophagus, oral cavity and liver.
  2. Contrastingly, women have a higher risk of cancers of the thyroid, anus and lung adenocarcinoma, and breast.
  3. 40% of cancer incidence in the UK is attributable to known modifiable risk factors.
  4. Over the past 25 years in the UK, cancer incidence has declined and then stabilised in men, while rising modestly in women. Here on this graph, these trends partly reflect changes in major risk factors. Smoking, which accounts for an estimated 15.1% of UK cancer cases, has fallen, particularly among men. By contrast, overweight and obesity, linked to 6.3% of cases, have increased in both sexes.
  5. Temporal trends in cancer incidence differ by sex in ways that broadly mirror sex-specific patterns of risk factor exposure, and it is hypothesized that differential risk factors distributions contribute substantially to observed sex differences in cancer incidence.
  6. Health-related factors, specifically smoking, overweight and obesity, and alcohol consumption, are important contributors for risk of many cancers and are hypothesized to contribute to the sex differences observed for many cancer sites given their predominantly higher prevalence in men compared to women.

Two considerations for this doctoral work have been to

Carefully evaluate contributions from three risk factors (smoking, alcohol, and overweight and obesity) in examining sex differences in cancer incidence, especially for cancer sites for which they are well-established risk factors.

Aim to comprehensively assess whether and to what extent adjustment for risk factors contribute to observed sex differences in cancer incidence.

2. Conceptually mapping it

Which risk factors may conceptually contribute to sex differences?

Extrinsic risk factors

Environmental factors

  • Infections
  • Occupational exposures
  • Sociodemographic characteristics

Health-related factors

  • Smoking
  • Obesity & overweight
  • Alcohol
  • Physical activity
  • Diet
  • Healthcare-utilisation and screening behaviours

Intrinsic risk factors

Biological factors

  • Body size (BMI)
  • Height
  • Molecular biomarkers (immune system, metabolism)
  • Underlying medical conditions and diseases

Sex-related biological factors

  • Sex chromosomes
  • Sex hormones
Classification of extrinsic and intrinsic risk factors hypothesized to contribute to sex differences in cancer incidence. Extrinsic factors are shown in blue and intrinsic factors in pink, matching the supplied chart.

Here, extrinsic factors include environmental and health-related exposures that differ by sex and are potentially modifiable, whereas intrinsic factors reflect biological and sex-related characteristics that may influence cancer susceptibility.

This classification is necessarily simplified and partially overlapping but is intended to guide as a structured framework for organising the range of risk factors evaluated throughout the research.

Note: For example, body size, as measured by body mass index (BMI), is classified as an intrinsic biological factor reflecting differences in overall body size and composition between men and women. This is conceptually distinct from overweight and obesity, which are treated as health-related extrinsic factors, as they represent physiological states arising from the cumulative influence of health-related and environmental exposures. In contrast, underlying medical conditions and diseases are categorised as intrinsic risk factors, as they proxy intrinsic chronic biological or pathological states, thus influencing cancer susceptibility. For example, inflammatory bowel disease is an immune-mediated condition that lies on the causal pathway for colorectal cancer risk but also arises partly from extrinsic exposures. Accordingly, this classification is intended to be analytic rather than absolute.

3. Taking a data-driven approach to assess

Prospective Analyses in the UK Biobank

  1. 1

    The central aim of my doctoral work was to conduct a comprehensive investigation of sex differences in cancer incidence and to assess the extent to which observed disparities persisted after accounting for established and measured risk factors.

  2. 2

    Prospective analyses in the UK Biobank cohort on 28 cancer endpoints (including 15 main cancer sites and 13 subtypes) were performed, and collectively comprised the pan-cancer scope. An outcome-wide analytical approach was adopted, allowing standardised analyses across multiple cancer sites within a cohort. Using a consistent modelling framework enabled direct comparison of sex differences and their attenuation across cancers.

  3. 3

    Cancer site-specific risk factors were selected based on three considerations: evidence as established risk factors, epidemiological evidence supporting their contribution to observed sex differences, and availability of relevant data within the UK Biobank cohort.

  4. 4

    The following risk factors were considered and assessed for all cancer endpoints:

    • Educational attainment
    • Ethnicity
    • Height
    • BMI
    • Smoking status
    • Alcohol intake

    Additionally site-specific risk factors were modelled.

Research findings

A note on terminology

Throughout this work, the terms “men” and “women” were used to describe participants and are intended to correspond to biological sex as male and female. This terminology reflects a preference that acknowledges subjects as people rather than categories.

Acknowledgements

While the research described here formed part of my doctoral work, the development of its questions and methods was deeply collaborative. At Oxford, the work was guided throughout by my supervisors, Professor Ruth Travis, Professor Tim Key, and Dr Joshua Atkins. Its direction and methodology were also shaped by discussions and collaborations with Kirstin Pirie, Dr Keren Papier, Dr Karl Smith-Byrne, Professor Tammy Tong, Professor Gill Reeves, Professor Isobel Barnes, Professor Ben Lacey, Professor Derrick Bennett, and others across the Cancer Epidemiology Unit and Oxford Population Health.

The work also built on my earlier exploration of related questions in cardiovascular disease at Cambridge, where I worked with and learnt from Professor Antonis Antoniou, Dr Nasim Mavaddat, Dr Stephen Burgess, and Professor John Perry. Together, these relationships brought perspectives from people approaching analogous questions in different domains and helped sharpen both the research questions and the methods used to investigate them.