The majority of observational studies (Furner et al, 1989
; Jacobs et al, 1994
; Folsom et al, 1995
; Newcomb and Storer, 1995
; Persson et al, 1996
; Kampman et al, 1997
; Fernandez et al, 1998
; Grodstein et al, 1998
; Nanda et al, 1999
; Prihartono et al, 2000
) and the results from the WHI (Writing Group for the Women's Health Initiative Investigators, 2002
) randomised controlled trial suggest that oral conjugated oestrogens reduce the risk of colorectal cancer by 20–30%. The effect on colorectal cancer risk of oestrogen replacement delivered by the transdermal mode has not been examined previously. We observed a protective effect for colorectal cancer that is as great, and possibly greater, in magnitude than that reported for OE replacement therapy.
Although our results for transdermal and OE were not adjusted for known important risk factors for colorectal cancer including diet and physical activity, we were able to investigate the impact of some potential covariates obtained from prospectively recorded healthcare databases. The impact of the past use of prescription NSAIDs and vitamins, the use of oral contraceptives, and the frequency of physician visits prior to index dates had a negligible effect on ORs. The minimal impact of covariates including diet and physical activity on the measures of association between HRT and colorectal cancer has been reported by other investigators (Kampman et al, 1997
; Fernandez et al, 1998
; Grodstein et al, 1998
; Prihartono et al, 2000
We also report results among women who had not had a sigmoidoscopy during the 3–5-year period prior to index dates. All ORs remained strongly protective for TDE and some associations became even more protective. We assumed that a sigmoidoscopy during this time period was a ‘screening’ sigmoidoscopy rather than a ‘diagnostic’ procedure. The impact of having had a sigmoidoscopy during this time could conceivably have a protective effect, if premalignant polyps had been removed (Winawer et al, 1993
). We did not, however, have information on whether or not a polypectomy was performed, nor did we have information to indicate the reason for which a woman had had a sigmoidoscopy.
The protective effect of having had a polypectomy during a sigmoidoscopy could falsely be attributed to HRT if a woman who had used HRT had been more likely to have had this procedure performed. The association between HRT and colorectal cancer was not examined among women who had had a sigmoidoscopy 3–5 years prior to index dates, because too few women had had the procedure. However, in the majority of women who had not had a sigmoidoscopy at that time, the results for HRT remained protective. We can therefore rule out the impact of having had a polypectomy playing a role in the observed protective effect.
Women choosing to use HRT are known to be different from women choosing not to use HRT, and it had been suggested prior to the release of WHI results that some of these differences may be responsible for many of the protective effects against chronic diseases attributed to HRT (Barrett-Connor, 1991
; Matthews et al, 1996
; Brett and Madans, 1997
). However, in many studies, even in well-designed studies where all known important covariates are measured prospectively (Grodstein et al, 1998
), crude relative risks have not been very different from multivariate relative risks. It is of relevance to our results that many of the healthy lifestyle characteristics associated with HRT use are also likely to be associated with TDE use. Whether health-related differences exist between transdermal and OE users has not to our knowledge been explored. We examined the use of prescribed vitamins, NSAIDs, having had a sigmoidoscopy 3–5 years prior to the index date and frequency of physician visits among women who had been prescribed OE and TDE. The histories were very similar in these two groups. Transdermal oestrogen users were somewhat younger (mean 63.8 (s.d. 8.6) years) than OE users (mean 68.6 (s.d. 9.0) years), but all analyses were age-adjusted either by matching (conditional logistic regression) or in analyses where the matching was broken, by including age as a covariate in the model. There is no evidence to suggest that differences, if they exist, would be as extreme as the differences between HRT users and nonusers. In the analysis where women who had used OE comprised the reference group, we eliminated the important potential for bias that may exist with a comparison group of non-HRT users. The protective effect of TDE remained.
Our study is limited by the fact that we only have a small number of women with colorectal cancer who are exposed only to TDE. We do, however, have a sufficient number of exposed controls and, if TDE is truly protective, we may always have fewer exposed cases than controls. Despite a small number of exposed cases, some of our findings were significant. Nonetheless, in order to study the effect of the duration of TDE use adequately, a larger population with a longer history of TDE use must be studied.
An important strength of this study is the fact that we have highly accurate oestrogen exposure and diagnostic data, prospectively documented for all of our subjects. This avoids the potential problem of violations of temporal order, misclassification of exposure data, and selection bias due to incomplete diagnostic information, which can plague case–control studies. We also have accurate data from health service databases for all of our subjects, with regard to the dispensing of several other prescription drugs, and history of having had a sigmoidoscopy. Since 91% of the Saskatchewan population is eligible to receive provincial prescription drug benefits, our results are generalisable to the general population, with the exception of members registered with First Nations, who receive benefits under a federal program. It must therefore be cautioned that our results are not generalisable to this group.
Clinical studies have revealed metabolic differences between transdermal and OE. The induction of hepatic protein synthesis that occurs with OE administration (sex hormone-binding globulin, corticosteroid-binding globulin, thyroxine-binding globulin, transferrin, ceruloplasmin, apolipoprotein A1, rennin substrate, and various coagulation and fibrinolysis factors) is avoided with TDE (Chetkowski et al, 1986
; Crook, 1997
). The estradiol/estrone ratio produced with TDE administration is closer to premenopausal levels, than is the ratio produced with OE administration (Chetkowski et al, 1986
; Jewelewicz, 1997
; IARC Working Group on Hormonal Contraception and Post-menopausal Hormonal Therapy, 1999
). In addition, with the TDE patch, the release of oestrogen is constant and the peaks and troughs characteristic of OE are avoided (Jewelewicz, 1997
). The clinical relevance of these different biological properties is an area that requires study.
In summary, the reduction in colorectal cancer risk with TDE use has not previously been reported and warrants further investigation. Knowing more about the differences in health outcomes between TDE and OE could help women and their physicians individualise therapy in order to minimise the risks and maximise the benefits of treating distressing menopausal symptoms.