Meta-analysis in cancer epidemiology

Gandini, Sara (2004). Meta-analysis in cancer epidemiology. University of Birmingham. Ph.D.

Preview

Gandini04PhD_A1b.pdf
PDF
Download (1MB)

Abstract

A published meta-analysis on breast cancer and vegetables and fruit consumption was described to present a methodology used on meta-analysis in Epidemiology. Meta-analysis confirmed the association between intake of vegetables (RR=0.74; 95%CI 0.65-0.84) and, to a lesser extent, fruit and breast cancer risk (RR=0.93; 95%CI 0.79-1.09). Using this methodology, present in a peer-reviewed journal, a systematic meta-analysis on melanoma was conducted extracting RRs from published studies. Fully adjusted estimates were obtained from those studies, when available; RRs adjusted for confounders not related to sun exposure, such as naevi, were considered for sun exposure and sunburns pooled estimates. Pooled estimates were obtained for all main risk factors for melanoma: sun exposure (total, intermittent and chronic), sunburns (in childhood and in adulthood), indicators of actinic damage, family history of melanoma and phenotype characteristics. Investigation of biases and inconsistencies among studies was one of the key phases of the meta-analysis to look for patterns among studies that might explain discrepant findings. The analyses on pigmented lesions and sun exposure showed that the choice of sources of cases and controls influenced significantly the estimate. An indication of a protective effect of chronic sun exposure came from studies that did not include subjects with dermatological problems (significantly different from the other studies: p=0.01). Publication year was an important factor for total sun exposure (p=0.005). Latitude of the study seemed to be an important factor for sunburns (p=0.002) and for high density of freckles (p=0.04). Estimates for hair colour and eye colour adjusted for phenotype and/or photo-type were significantly lower than unadjusted ones (p=0.06 and p=0.06, respectively). This study highlighted how several features of study design, type of analysis, categorization of exposures, study location and populations significantly explained between-study heterogeneity.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):