CONTRACEPTION
ORAL CONTRACEPTIVE USE DOES NOT PROTECT AGAINST LARGE BOWEL CANCER
Gabriel A. Kune, MD, Professor of Surgery Susan Kune, PhD, Epidemiologist Lyndsey F. Watson, BSc, Biostatistician University of Melbourne, Department of Surgery 61 Erin Street, Richmond, Victoria 3121, Australia
ABSTRACT The association between oral contraceptive (OC) use and colorectal cancer was examined in 190 female colorectal cancer cases and 200 age-matched female controls in data derived from a population-based study of large bowel cancer, "The Melbourne Colorectal Cancer Study" conducted in Melbourne, Australia. There were 47 cases (24 colon cancer, 23 rectal cancer cases) and 39 controls, who were past OC users. After adjustment was made fqr the confounding factors of age, number of children and age at birth of first child, a statistically significant risk was found among rectal cancer OC users, but not among colon cancer OC users (RR rectal cancer = 2.04, 95% CI = 1.00-4.14, p = 0.04; RR colon cancer = 1.17, 95% CI = 0.59-2.29, p = 0.60). These risks were not affected by adjustment for socioeconomic level, country of birth, religion, previous diet and family history of colorectal cancer. Rectal cancer risk was higher among those OC users who were also beer drinkers ( RR = 6.96, 95% CI 2.09-23.1, p = 0.001). INTRODUCTION It has been suggested recently that previous oral contraceptive use protects women from the development of colorectal cancer [1,2]. This communication examines the association between OC use and colorectal cancer risk in females, in data drawn from a large, population-based, clinicopathological and epidemiological study of large bowel cancer, "The Melbourne Colorectal Cancer Study" [3]. This study has three principal substudy arms, named the incidence study, the case-control study and the survival study, and the data in this paper are from the case-control arm of the study, which examined all the major current hypotheses of colorectal cancer etiology and risk.
Powlation
Studies
The study was population-based, the target population being Melbourne, specifically the Melbourne Statistical Division, which Submitted for publication November 28, 1988 Accepted for publication September 26, 1989 JANUARY
1990 VOL. 41 NO. 1
19
CONTRACEPTION
at the time of the study had a population of 2.81 million [3]. cases of the case-control study represented all histologically confirmed new cases of adenocarcinoma of the colon and rectum first diagnosed in the 12-month period between 21 April 1980 and 20 April 1981, who were residents of the Melbourne Statistical Division. The controls were also from the Melbourne Statistical Division, selected according to a random cluster sampling plan devised by the Australian Bureau of Statistics, the government agency responsible for the Australian Population Census and the collection of other Australian vital statistics. The controls were frequency matched according to sex and fiveyear age groupings. The methods of selecting both cases and controls, the criteria for exclusions and the characteristics of cases and controls not included in the study when compared to those who were included in the study, have been described in detail in several previous publications [3,4,5,6,7] and for brevity will not be described here. Based on the analysis of the characteristics of those excluded versus those included in this study, the authors concluded that selection bias and exclusion bias were not significant biases in the study [6,7]. The participation rate for cases was 62% and for controls 71% [6,7].
The
In the case-control study there were 715 cases and 727 community controls, of which there were 327 female cases and 329 female controls. Keeping in mind the year of introduction of OC into Australia (1961), for this paper, those females 70 years of age or over, who could not have used OC, were excluded. Also excluded were four colon cancer cases, four rectal cancer cases and one control, whose age at birth of first child was not known. This last factor had previously been found to be associated with the risk of colon and rectal cancer [6]. Thus the analysis which follows relates to 108 colon cancer cases, 82 rectal cancer cases and 200 controls. Data
Collection
Two questionnaires were administered by personal interview, each on a separate occasion. The interviewers knew the status of the respondents, that is, whether they were a case or a control, but were not told of any of the hypotheses being tested. The first questionnaire included data on age, sex, country of birth and religion, current and past illnesses, operations, medications (including oral contraceptive use), bowel habit, biopsychosocial factors, number of children, age at birth of first child and family history data [3]. The second interview was the dietary questionnaire which included alcohol intake and tobacco use [3,5]. Cases were interviewed shortly after diagnosis. The two interviews followed each other within days. The part of the interview which dealt with previous illnesses, operations, and medications, including previous OC use, was uniformly introduced by the interviewer as follows: "Now, I'd like to talk to you about your general health. I am going to read through a list of operations, illnesses and medications. Would you tell me if you have had any of these operations or illnesses or taken these medications, and if so,
JANUARY
1990VOL. 41 NO. 1
CONTRACEPTION
when and for how long". Thus the respondents were unaware of the purposes of the study and unaware of the hypotheses being tested. Responses were recorded as "Yes", VIN~ll, "Don't know". An OC user was defined as a respondent who has had at least one course of one month of OC pills. The year of commencement of the first OC course was recorded, and also the year of the last course. No data were obtained on the formulation of the various OC pills used. The data given were not verified by any other means, such as by checking physician or hospital records. Data Analvsis Data analysis was done using the GLIM statistical package Relative risk estimates (RR) and their 95% confidence intervals (CI) were obtained from unconditional logistic regression analysis [9]. Age, a design constraint, was included in all logistic regression models. [81.
There were 108 colon cancer cases (mean age 58.1 years, SD 8.0), 82 rectal cancer cases (mean age 57.9 years, SD 8.0) and 200 controls (mean age 58.0 years, SD 7.9). There were 24 colon cancer and 23 rectal cancer cases and 39 controls who were OC users. When the cases were further divided into subsites of "right colon" (which included caecum, ascending colon and transverse colon), "left colon" (which included descending colon and sigmoid colon) and rectum, there were nine OC users in 49 right colon cancers cases (18%), 15 OC users in 59 left colon cancer cases (25%) and 23 OC users in 82 rectal cancer cases (28%). Of the 200 controls, 39 (20%) were OC users.TheTable shows the number of past OC users, their mean age, mean age at the start of OC use, mean age at the completion of OC use, mean time between first and last course (calculated in months, but quoted in years). The mean time between the first and last OC course is the mean duration of OC use. In the Melbourne Colorectal Cancer Study, increasing number of children had previously been found to be protective for males and females in both colon cancer and rectal cancer [6]. Also, increasing age at birth of first child was found to be a risk factor in colon cancer and rectal cancer, statistically significant in colon cancer only [6]. These factors were considered to be likely confounding variables for OC use and therefore required to be adjusted for in modelling OC use and colon and rectal cancer risk. After adjustment for number of children and age at birth of first child, no statistically significant association between previous OC use and colon cancer was found (RR = 1.17, 95% CI = 0.059-2.29, p = 0.60) but there was a statistically significant risk between previous OC use and rectal cancer (RR = 2.04, 95% CI = 1.00-4.14, p = 0.04).
JANUARY
1990VOL.
41 NO. 1
21
2
5
8
2
g 9
k
(%)
%I0
age in years
5
P value
first
to difference
between
cases
OC course(W)
and last OC
of stopping
OC course(SD)
of all respondents.
refers
ll Percentage
Mean time between course (SD)
Mean
of starting
(%)
3-9 years
years
(%)
2 years
ORAL CONTRACEPTIVE USE DOES NOT PROTECT AGAINST LARGE BOWEL CANCER
Gabriel A. Kune, MD, Professor of Surgery Susan Kune, PhD, Epidemiologist Lyndsey F. Watson, BSc, Biostatistician University of Melbourne, Department of Surgery 61 Erin Street, Richmond, Victoria 3121, Australia
ABSTRACT The association between oral contraceptive (OC) use and colorectal cancer was examined in 190 female colorectal cancer cases and 200 age-matched female controls in data derived from a population-based study of large bowel cancer, "The Melbourne Colorectal Cancer Study" conducted in Melbourne, Australia. There were 47 cases (24 colon cancer, 23 rectal cancer cases) and 39 controls, who were past OC users. After adjustment was made fqr the confounding factors of age, number of children and age at birth of first child, a statistically significant risk was found among rectal cancer OC users, but not among colon cancer OC users (RR rectal cancer = 2.04, 95% CI = 1.00-4.14, p = 0.04; RR colon cancer = 1.17, 95% CI = 0.59-2.29, p = 0.60). These risks were not affected by adjustment for socioeconomic level, country of birth, religion, previous diet and family history of colorectal cancer. Rectal cancer risk was higher among those OC users who were also beer drinkers ( RR = 6.96, 95% CI 2.09-23.1, p = 0.001). INTRODUCTION It has been suggested recently that previous oral contraceptive use protects women from the development of colorectal cancer [1,2]. This communication examines the association between OC use and colorectal cancer risk in females, in data drawn from a large, population-based, clinicopathological and epidemiological study of large bowel cancer, "The Melbourne Colorectal Cancer Study" [3]. This study has three principal substudy arms, named the incidence study, the case-control study and the survival study, and the data in this paper are from the case-control arm of the study, which examined all the major current hypotheses of colorectal cancer etiology and risk.
Powlation
Studies
The study was population-based, the target population being Melbourne, specifically the Melbourne Statistical Division, which Submitted for publication November 28, 1988 Accepted for publication September 26, 1989 JANUARY
1990 VOL. 41 NO. 1
19
CONTRACEPTION
at the time of the study had a population of 2.81 million [3]. cases of the case-control study represented all histologically confirmed new cases of adenocarcinoma of the colon and rectum first diagnosed in the 12-month period between 21 April 1980 and 20 April 1981, who were residents of the Melbourne Statistical Division. The controls were also from the Melbourne Statistical Division, selected according to a random cluster sampling plan devised by the Australian Bureau of Statistics, the government agency responsible for the Australian Population Census and the collection of other Australian vital statistics. The controls were frequency matched according to sex and fiveyear age groupings. The methods of selecting both cases and controls, the criteria for exclusions and the characteristics of cases and controls not included in the study when compared to those who were included in the study, have been described in detail in several previous publications [3,4,5,6,7] and for brevity will not be described here. Based on the analysis of the characteristics of those excluded versus those included in this study, the authors concluded that selection bias and exclusion bias were not significant biases in the study [6,7]. The participation rate for cases was 62% and for controls 71% [6,7].
The
In the case-control study there were 715 cases and 727 community controls, of which there were 327 female cases and 329 female controls. Keeping in mind the year of introduction of OC into Australia (1961), for this paper, those females 70 years of age or over, who could not have used OC, were excluded. Also excluded were four colon cancer cases, four rectal cancer cases and one control, whose age at birth of first child was not known. This last factor had previously been found to be associated with the risk of colon and rectal cancer [6]. Thus the analysis which follows relates to 108 colon cancer cases, 82 rectal cancer cases and 200 controls. Data
Collection
Two questionnaires were administered by personal interview, each on a separate occasion. The interviewers knew the status of the respondents, that is, whether they were a case or a control, but were not told of any of the hypotheses being tested. The first questionnaire included data on age, sex, country of birth and religion, current and past illnesses, operations, medications (including oral contraceptive use), bowel habit, biopsychosocial factors, number of children, age at birth of first child and family history data [3]. The second interview was the dietary questionnaire which included alcohol intake and tobacco use [3,5]. Cases were interviewed shortly after diagnosis. The two interviews followed each other within days. The part of the interview which dealt with previous illnesses, operations, and medications, including previous OC use, was uniformly introduced by the interviewer as follows: "Now, I'd like to talk to you about your general health. I am going to read through a list of operations, illnesses and medications. Would you tell me if you have had any of these operations or illnesses or taken these medications, and if so,
JANUARY
1990VOL. 41 NO. 1
CONTRACEPTION
when and for how long". Thus the respondents were unaware of the purposes of the study and unaware of the hypotheses being tested. Responses were recorded as "Yes", VIN~ll, "Don't know". An OC user was defined as a respondent who has had at least one course of one month of OC pills. The year of commencement of the first OC course was recorded, and also the year of the last course. No data were obtained on the formulation of the various OC pills used. The data given were not verified by any other means, such as by checking physician or hospital records. Data Analvsis Data analysis was done using the GLIM statistical package Relative risk estimates (RR) and their 95% confidence intervals (CI) were obtained from unconditional logistic regression analysis [9]. Age, a design constraint, was included in all logistic regression models. [81.
There were 108 colon cancer cases (mean age 58.1 years, SD 8.0), 82 rectal cancer cases (mean age 57.9 years, SD 8.0) and 200 controls (mean age 58.0 years, SD 7.9). There were 24 colon cancer and 23 rectal cancer cases and 39 controls who were OC users. When the cases were further divided into subsites of "right colon" (which included caecum, ascending colon and transverse colon), "left colon" (which included descending colon and sigmoid colon) and rectum, there were nine OC users in 49 right colon cancers cases (18%), 15 OC users in 59 left colon cancer cases (25%) and 23 OC users in 82 rectal cancer cases (28%). Of the 200 controls, 39 (20%) were OC users.TheTable shows the number of past OC users, their mean age, mean age at the start of OC use, mean age at the completion of OC use, mean time between first and last course (calculated in months, but quoted in years). The mean time between the first and last OC course is the mean duration of OC use. In the Melbourne Colorectal Cancer Study, increasing number of children had previously been found to be protective for males and females in both colon cancer and rectal cancer [6]. Also, increasing age at birth of first child was found to be a risk factor in colon cancer and rectal cancer, statistically significant in colon cancer only [6]. These factors were considered to be likely confounding variables for OC use and therefore required to be adjusted for in modelling OC use and colon and rectal cancer risk. After adjustment for number of children and age at birth of first child, no statistically significant association between previous OC use and colon cancer was found (RR = 1.17, 95% CI = 0.059-2.29, p = 0.60) but there was a statistically significant risk between previous OC use and rectal cancer (RR = 2.04, 95% CI = 1.00-4.14, p = 0.04).
JANUARY
1990VOL.
41 NO. 1
21
2
5
8
2
g 9
k
(%)
%I0
age in years
5
P value
first
to difference
between
cases
OC course(W)
and last OC
of stopping
OC course(SD)
of all respondents.
refers
ll Percentage
Mean time between course (SD)
Mean
of starting
(%)
3-9 years
years
(%)
2 years
