An association between codon-72 p53 polymorphism and risk of human papillomavirus (HPV)-induced cervical cancer has been found recently, but it has been difficult to replicate. In this study, we assess the impact of inter-laboratory variation in p53 genotyping on the validity of the proposed association. DNA specimens were randomly selected from 54 invasive, squamous cell carcinoma cases, 52 HPV-negative, and 39 HPV-positive controls from a previous case-control study in Brazil. Codon-72 polymorphism was blindly analyzed in three different laboratories. We calculated age- and race-adjusted odds ratios (OR) and 95% confidence intervals (CI) using logistic regression for gauging the association between p53 polymorphism and cervical cancer risk. The proportions of the Arg/Arg, Arg/Pro, and Pro/Pro genotypes varied substantially among laboratories with Kappa coefficients in the 0.49-0.63 range. When disagreement between labs was allowed, the OR for the Arg/Arg genotype, compared to other forms, was as low as 1.5 (95% CI: 0.5-3. 9). In contrast, the OR increased to 8.0 (95% CI: 2.3-28.5) after exclusion of discordant genotypes. Restricting the comparison to HPV-positive controls increased the magnitude of the relation appreciably. After exclusion of all discordant diagnoses, the OR was 21.5 (95% CI: 3.4-137.8), whereas with disagreed genotypes the association was not significant (OR = 2.9, 95% CI: 0.7-11.9). Homozygous codon-72 p53-Arg apparently confers a higher susceptibility to HPV-associated cervical tumorigenesis. However, exposure misclassification consequent to inter-laboratory variation in protocols may affect the ability to detect the association.
Copyright 2000 Wiley-Liss, Inc.