May 5, 1996
In a recent paper, A N Phillips and George Davey
demonstrated mathematically that smoking could be falsely blamed for cervical cancer by
failure to detect all exposure to a causative pathogen with a high risk.
Subsequently, Bosch et al re-analyzed their
cervical cancer study data to assess whether this had happened. They found that "when
HPV [human papillomavirus] positive women are examined, thus removing those with potential
undetected HPV, the odds ratios os cervical cancer in five data sets 3 on invasive cancer
and 2 of CIN III lesions) were consistently and statistically not different from 1"
for cigarette smoking. They concluded, "Real data are thus in agreement with
the model proposed by Phillips and Smith".
Since the studies claimed by anti-smokers to show
that smoking causes cervical cancer invariably fail to test their results in this way,
they are therefore making a false and spurious claim based on their own methodological
malpractice.
This also explains why these studies routinely
show cervical cancer risks associated with low education that are equal to or greater than
those claimed for smoking, which they never feel compelled to invent any specious theories
of biological plausibility to explain.
And, these results indicate that studies blaming
secondhand smoke for cervical cancer in passive smokers are actually just exploiting the
socio-economic similarity between smokers and passive smokers.
THE LETTER FROM ANDREW PHILLIPS AND GEORGE DAVEY SMITH to BOSCH et al.
Sir- Whidden suggests that studies underestimate the effect of smoking by comparing
smokers with a baseline group who are themselves exposed, albeit passively, to cigarette
smoke. Two meta-analyses have estimated the summary relative odds for the 'independent'
effect of current smoking compared with non-smoking to be 1.46 and 1.69. [1,2] Even
if one believe the effect to be causal, given the modest size of the difference in risk of
cervical cancer in women who actively smoke versus those who do not, it seems unlikely
that there is substantial difference in risk between non-smokers in a smoking environment
and non-smokers in a non-smoking environment. Therefore, even if thew comparison group
were non-smokers in a non-smoking environment the relative odds estimates are unlikely to
be much larger.
Even so, we illustrated that relative risks of as high as two and above could
easily arise merely as a result of the inability to fully adjust for exposure to the
sexually transmitted pathogen involved [3] . This was based on plausible
estimates of the association between the presence of the aetiological pathogen and both
smoking and risk of cervical cancer. The conclusion holds equally whether one is comparing
active smokers with all non-smokers or active smokers with non-smokers in a non-smoking
environment.
Further, in the study which Whidden says 'claims to have proven that second-hand
smoking is definitely a risk factor for this cancer [4] the adjusted relative odds
associated with passive smoking (which was 2.96) was almost the same as that associated
with active cigarette smoking (3.42). As pointed out by Layde [5] since active
cigarette smokers are also passive smokers because of the side-stream smoke from their own
cigarettes they would be expected -- if passive smoking truly contributes to the
development of cervical cancer -- to have a greater risk than women whose only exposure
was passive. This, and the fact that adjustment for confounders -- including number of
sexual partners -- reduced the relative odds associated with passive smoking from 14.4 to
2.96, are consistent with the hypothesis that the passive smoking effect was due to
residual confounding resulting from the inability to measure the presence of the sexually
transmitted pathogen involved.
_____________
University Department of Public Health, Royal Free Hospital School of
Medicine. Rowland Hill Street, London NW3 2PF, UK
_____________
REFERENCES
1 Sood A.K. Cigarette smoking and cervical cancer:
meta-analysis and critical review of recent studies, Am J Prev Med 1991:7:208-13.
2 Licciardone JC Brownson RC, Chang JC, Wilkins JR. Uterine
cervical cancer risk in cigarette smokers: a meta-analytic study, Am J Prev Med
1990:6:274-81.
3 Phillips AN, Davey Smith G. Cigarette smoking as a potential
cause of cervical cancer: has confounding been controlled? Int. J Epidemiol
1994:23:42-49.
4 Slattery ML, Robinson LM, Schuman KL et al. Cigarette
smoking and exposure to passive smoking are risk factors for cervical cancer. JAMA
1989:261: i593-98
5 Layde PM. Smoking and cervical cancer: cause or coincidence?
JAMA 1989: 261 : 1631-33
THE ANSWER FROM F X BOSCH, S DE SANJOSÉ AND N MUÑOZ
Sir -- In the article by Phillips and Smith 'Cigarette smoking as a potential cause of
cervical cancer: Has confounding been controlled?' [1] it is argued that the
increase in risk of cervical cancer associated with smoking observed in some studies could
well be explained by an insufficient adjustment by some measurements of exposure to the
aetiological pathogen.
Recent developments have shown that the major risk factor of cervical cancer is
human papillomavirus (HPV). We have completed five case-controlled studies on cervical
cancer where the prevalence of HPV-DNA had been assessed using polymerase chain reaction
(PCR). [2-4]
Our results show that the association between HPV infection and cervical cancer is very
strong with odds ratios ranging from 15 to 100. Many of the 'traditional' risks factors
strongly associated with cervical neoplasia in either pre-invasive or invasive forms, such
as number of sexual partners or age at first sexual intercourse were not associated with
cervical cancer among women who were HPV DNA negative., while the association persisted
among the HPV DNA negative. In our data the association between smoking status and
cervical cancer was weak with odds ratios ranging from 1.4 to 2.0 after adjustment for
major confounders including HPV status. However, when HPV positive women were examined,
thus removing those with potential under-detected HPV, the odds rations of cervical cancer
in five data sets (3 on invasive cancer and 2 in CIN III lesions) were consistently and
statistically not different from 1.
Real data are thus in agreement with the model proposed by Phillips and Smith.
_____________
Servei d'Epidemiologia Registre de Cancer. Insitut Oncologic Duran i
Reynals. Ciudad Sanitaria i Universitaria de Bellvirge. Autovia de Castelldefels Km. 2.7
08907 l'Hospitalet de Llibregat, Barcelona, Spain
_____________
REFERENCES
1 Phillips AN, Davey Smith G. Cigarette smoking as a potential
cause of cervical cancer: has confounding been controlled? Int. J Epidemiol
1994:23:42-49.
2 Bosch F X Muñoz N, de Sanjosé S et al. Risks factors for
cervical cancer in Colombia and Spain. Int J Cancer 1992: 52: 750-58
3 Muñoz N. Bosch F X de Sanjosé S et al. Risks factors for
CIN III in Spain and Colombia. Cancer epidemiology, Biomarkers & Prevention 1993:
2: 423-31
4 Eluf-Neto J. Booth M. Bosch F X. Meijer C J L M, Walboomers
J M M. Human papillomavirus and invasive cervical cancer in Brazil. Br. J Cancer 1994,
69: 114-19.
_____________
Courtesy of Carol Thompson
08/23/93
Smokers' Rights Action Group
P.O. Box 259575
Madison, WI 53725-9575
Phone: 608-249-4568 |
