|Module 3: Interpreting Data
7.1 Case Study: Vaginal cancer and synthetic oestrogen, DES
In 1971 a group of researchers published a report in which they examined eight cases of vaginal cancer in women aged between 15 and 22 (Herbst, Ulfelder et.al 1971). Vaginal cancer is a rare disease normally diagnosed in women over 70. The researchers were interested in examining the possibility that the mothers of these women had taken diethylstilbestrol (DES), a synthetic oestrogen, when they were pregnant and this had caused vaginal cancer in their daughters. In the US between 1940 and 1971, DES had been prescribed to prevent miscarriage and to ensure a healthy pregnancy.
Their study compared the eight cases of vaginal cancer with 32 referents (the referents [4 for each case] were women who had been born within 5 days of one of the women with the cancer, and under similar conditions) to see if the use of DES was more common among mothers of the women who had vaginal cancer. They found that in 7 of the 8 cancer cases the mothers had taken DES during pregnancy. Among the referents, none of the mothers had received DES. This study indicated a very strong association between a mother's use of DES and vaginal cancer in her female children.
This study used available data. This type of study can be called a retrospective observational study because the researchers are trying to establish a cause or explanatory variable for an observed effect or response variable. In this case, it obviously was not ethical or practicable to conduct an experiment to see if the taking of DES by a mother causes vaginal cancer in her female children. In this case, the results were so compelling that they led to a much wider examination of DES and its withdrawal from the market. It is now described as a known carcinogen.
The use of referents in this study is an example of one way of setting up 'matched' controls in observational studies. Be aware however, that in retrospective studies, such as this one, the researcher is relying on the memory of the participants to be accurate. Care needs to be taken when using the past as a source of data for observational studies. The following question, drawing on research by Utts (1996), highlights this issue.