Objective: Studies of the association between circulating IGF-I and cancer risk have shown conflicting results. We have previously observed a U-shaped association between IGF-I and cancer mortality. This study test the hypotheses of a U-shaped association between IGF-I and incident cancer.
Design: Elderly men (2368), randomly recruited from the general community.
Methods: IGF-I was measured in a cohort of elderly men. Complete data for incident cancer were obtained from the Swedish Cancer Registry. Statistical analyses included Cox proportional hazards regressions with or without a spline approach.
Results: Three hundred and sixty-nine participants had incident cancer after baseline. Prostate cancer was most frequent (n = 140). There was no association between serum IGF-I and all cancer or prostate cancer incidence. However, there was a nonlinear association between IGF-I and nonprostate cancer incidence (P = <0·05). Exploratory analyses were performed for low and high serum IGF-I (quintiles 1 and 5) vs intermediate (quintiles 2-4, referent). There was a tendency of increased nonprostate cancer risk in men with high IGF-I (HR = 1·26, 95% confidence interval (CI): 0·92-1·71, P = 0·15). After excluding participants with follow-up of less than 2·6 years (half median follow-up time), to control for potential diagnostic delay, the association was statistically significant (HR = 1·55, CI: 1·03-2·35).
Conclusion: There was a significant nonlinear association between IGF-I and nonprostate cancer. No association between IGF-I and prostate cancer was observed. Future studies are warranted to further investigate this nonlinear association, including whether IGF-I concentration is a reproducible, and useful, risk marker of nonprostate cancer.
© 2015 John Wiley & Sons Ltd.