Background Insulin-like growth factor-I (IGF-I) is suggested to support cancer cell growth and proliferation. Pre-diagnostic circulating IGF-I concentrations have been shown to be positively associated with breast cancer, prostate cancer and colorectal cancer but evidence for less common cancer sites is limited. The aim of this study was to investigate the associations between serum IGF-I concentrations and the incidence of rarer cancers using an outcome-wide approach to study cancers at 26 sites in UK Biobank, in which serum concentrations of IGF-I were measured for ~ 467,000 participants (93%).

Methods We analysed data from 394,406 cancer-free participants (52% women). IGF-I was measured in serum collected at baseline and in a subsample of 14,149 participants again in repeat samples collected during follow-up. Cancer diagnosis and death due to cancer during follow-up were determined using data-linkage with cancer and death registries. Multivariable-adjusted Cox proportional hazards models were used to determine associations between baseline serum IGF-I concentrations and cancer incidence, using the repeated measurements to correct estimates for regression dilution.

Results After a mean follow-up of 6.9 years, 23,496 participants were diagnosed with a malignant cancer. Higher IGF-I concentration was associated with an increased risk of colorectal cancer (hazard ratio per 5 nmol/l 1.10, 95%-CI 1.05–1.15), colon cancer (1.11, 1.05–1.17), malignant melanoma (1.08, 1.01–1.15), breast cancer in women (1.11, 1.07–1.15), prostate cancer (1.08, 1.04–1.11), thyroid cancer (1.23, 1.05–1.43) and multiple myeloma (1.13, 1.01–1.27), and a reduced risk of oral (0.86, 0.77–0.97), liver (0.37, 0.30–0.45), endometrial (0.90, 0.82–1.00) and ovarian cancer (0.88, 0.78–0.99).

Conclusion Higher IGF-I concentrations were associated with higher risks of cancer at the established sites (breast, colorectal and prostate cancer) and malignant melanoma, thyroid cancer and multiple myeloma. Higher IGF-I concentrations were associated with lower risks of oral, liver, endometrial and ovarian cancer; longer follow-up is needed to investigate the possible role of reverse causality.