Childhood exposure to infectious diseases and cancer risk – a systematic review and meta-analysis

By triggering the immune system, childhood infections are assumed to be protective in cancer development, particularly regarding non-solid tumors. However, findings of observational studies on this association are inconsistent and have not been summarized so far.

Using PubMed and Web of Science, we systematically searched the literature for studies conducted in humans on the association between childhood infections and cancer. We only included studies that were able to assign the infection to childhood and where the infection occurred prior to cancer diagnosis. Common effect estimates were calculated using random effects models for specific research questions if at least three studies were available for a given hypothesis.

The systematic literature search identified 12,901 studies. Overall, 31 studies fulfilled inclusion criteria for the review (including 61 effect estimators for an increased risk, 60 for a decreased risk and 366 for a non-significant association) and allowed meta-analyses for six different potential hypotheses. Regarding Hodgkin lymphoma, pooled odds ratios (OR) showed a risk reduction after measles infection [OR = 0.60 (95% CI 0.37, 0.97), number of pooled studies (N)= 4], an increased risk after infectious mononucleosis [OR = 1.91 (1.06, 3.44), N = 3], and a non-significantly reduced OR with chickenpox infection [OR = 0.66 (0.31, 1.38), N = 3] compared to children without the specific infection. The risk of non-Hodgkin lymphoma was increased for high [OR = 3.90 (1.14, 13.35), N = 4] and for medium [OR = 2.80 (1.39, 5.67), N = 4] versus low antibody levels of Epstein-Barr virus. The OR for leukemia after mumps infection was 0.80 [(0.49, 1.31), N = 3] compared to children without mumps.

We observed both positive and inverse associations, depending on the infectious pathogens and the cancer type. Different biological mechanisms of cancer development may explain these findings, but chance, exposure misclassification and uncontrolled confounding cannot be excluded.