【关键词】 case-control
1 school of public health and family medicine, university of cape town, south africa.
2 slone epidemiology center, boston university, usa.
3 slone epidemiology unit, boston university, usa.
in her commentary1 on our study of pap smear screening and cervical cancer in south africa,2 dr raffle makes three major criticisms.
dr raffle cites the importance of validating screening histories from records.3 given the lack of complete and available records on pap screening and therefore the need to rely on self-report, the nurse interviewers in our study were specially trained to conduct the interviews in a highly standardized manner designed for the most complete ascertainment of information. dr raffle states that the ‘unreliability of the data is betrayed by the implausibly high 73% participation (in pap screening) amongst the controls’. the rate is not implausible: a recent population-based survey by fonn et al. reported a prevalence of 70.6% in the western cape.4 in addition, as mentioned in our article, until recently opportunistic screening was performed when women attended family planning and antenatal clinics in the public health sector. attendance at these clinics is high in the western cape: 73.7% of women of child-bearing age currently use contraception and 91.7% attend antenatal clinics.5
dr raffle states that ‘there is no description of how the date of diagnosis was defined, of whether cases that became advanced during the 6 months they were captured [sic], and of whether fatal cases were included.’ we stated that incident cases were enrolled no more than 6 months after diagnosis. in fact, almost all of the cases were enrolled at the time of diagnosis. there was no opportunity for the cases to progress between the time of diagnosis and the time of enrolment. as we indicated in our paper, we conducted an interview-based case-control study; thus, fatal cases were not included.
dr raffle states that ‘healthy screenee bias’ cannot be ruled out. we agree as indicated by the statement in our paper that ‘there may have been some distortion from uncontrolled residual confounding’. dr raffle’s statement that we attempted ‘to remove potential bias by matching for decade of age, urban/ rural residence, race, education parity, age at first sexual activity, use of contraceptives, and cigarette smoking’ is incorrect. we used unconditional logistic regression to adjust for potential confounding by those factors. dr raffle may be confused by the fact that when we selected controls, we used series matching (not individual matching) in order to obtain similar distributions in the comparison groups according to age, race, and area of residence; these three factors were, in addition, allowed for in the multivariate analysis.
finally, a hypothesized explanation for why limited screening may be effective against cervical cancer in south africa is currently in press.6
references
1 raffle ae. commentary: case-control studies of screening should carry a health warning. int j epidemiol 2003;32:57778.
2 hoffman m, cooper d, carrara h et al. limited pap screening associated with reduced risk of cervical cancer in south africa. int j epidemiol 2003;32:57377.
3 weiss ns. application of the case control method in the evaluation of screening. epidemiol rev 1994;16:10208.
4 fonn s, bloch b, mabina m et al. prevalence of precancerous lesions and cervical cancer in south africa-a multi-centre study. s afr med j 2002;92:14856.
5 south africa demographic and health survey 1998: burden of disease research unit―reports. medical research council of south africa, 2001.
6 shapiro s, carrara h, allan b et al. the act of taking a papanicolaou smear reduces the prevalence of human papilloma virus infection: a potential impact on the risk of cervical cancer. cancer causes control 2003 (in press).
