Main

We genotyped a large set of both in situ and invasive squamous-cell cervical cancer cases and controls. Our material was derived from a strictly population-based epidemiological study of women with a diagnosis of cervical cancerin situ (n=488), as well as age-matched controls (n=626). We also analysed samples from 63 cases of invasive cancer from the same population. All in situ cases were collected in the county of Uppsala, Sweden, and only women born in Sweden were included2.

We found no statistically significant differences in the distribution of p53 genotypes between the control women and patients with either in situ or invasive cervical cancer (Table 1). Homozygosity for arginine at residue 72 was not associated with an increased risk for in situ(odds ratio, 1.06; P=0.648) or invasive cancer (odds ratio, 1.12; P=0.684). We also compared the frequencies of p53 codon-72 genotypes, using only cases and controls with a positive human papilloma virus (HPV)-16 history, as determined by amplification of DNA from archival smears using the polymerase chain reaction (PCR), and found no significant difference between them (Table 1). Thus, there is no indication that the p53 genotype increases the risk of developing cervical cancer in HPV-16-exposed women

Table 1 Distribution of codon-72 p53 genotypes in patients and controls

There are at least four possible explanations for the discrepancy between our results and those of Storey et al.1. First, their results could be due to chance. Our study was based on a sample size more than ten times larger than theirs and so gives a more accurate estimate of the association between the p53 polymorphism and cervical carcinoma. This allows us to detect a true odds ratio of two for association with the arginine-homozygote genotype with more than 95% certainty.

Second, selection bias may be introduced when a convenience sample of blood donors is used as a reference group. By contrast, our study was population based, so the controls were representative of the population in which the cases arose.

Third, the p53 polymorphism might be relevant only for women from certain populations. This would require that variants of the E6 protein differ in their ability to degrade p53, and that geographical differences exist in their distribution. But, given the similarity of British and Swedish populations, ethnic differences are unlikely to explain the discrepancy.

Finally, the DNA source and the techniques used for genotyping might have affected the results. Poor-quality DNA, such as that derived from formalin-fixed tissue, can inhibit PCR amplification, with failure being correlated with the length of the fragment3. Storey et al.1 used a PCR product for the proline allele that is 25% longer than the product for the arginine allele. This length difference may cause a bias in the PCR in favour of the arginine allele and an overestimate of the number of homozygotes for arginine 72. Such a bias would have less influence on large pieces of DNA, such as that isolated from the blood of the controls. It is dangerous to use an allele-specific assay in a study when the quality of the DNA differs between patients and controls.