Agreement between women's histories of oral contraceptive use and physician records

Cannabis smoking, tobacco cigarette smoking, and adenomyosis risk

OBJECTIVE

To investigate cannabis smoking and tobacco cigarette smoking in relation to adenomyosis risk.

DESIGN

We used data from a case-control study of adenomyosis conducted among enrollees ages 18-59 years of an integrated health care system in Washington State. The case-control study used 2 control groups given the challenge of selecting noncases when cases are diagnosed by hysterectomy.

SUBJECTS

Cases (n = 386) were enrollees with incident, pathology-confirmed adenomyosis diagnosed between April 1, 2001, and March 31, 2006. The 2 control groups comprised hysterectomy controls (n = 233) with pathology-confirmed absence of adenomyosis and population controls (n = 323) with an intact uterus selected randomly from the health care system population and frequency matched to cases on age.

EXPOSURE

Detailed data on cannabis and tobacco cigarette smoking history were ascertained through in-person structured interviews, allowing estimation of joint-years of cannabis smoking and pack-years of tobacco cigarette smoking.

MAIN OUTCOME MEASURES

Odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between cannabis smoking, tobacco cigarette smoking, and adenomyosis were estimated using multivariable unconditional logistic regression. Analyses were adjusted for age, reference year, menarche age, education, and pack-years of cigarette smoking (or joint-years of cannabis smoking).

RESULTS

No association was observed between cannabis smoking history and adenomyosis risk. However, we did observe the suggestion of an association between ever tobacco cigarette smoking and adenomyosis risk, comparing cases to hysterectomy controls (OR, 1.3; 95% CI, 0.9-1.9) and population controls (OR, 1.2; 95% CI, 0.8-1.8). Our data suggested a 50% increased odds of adenomyosis with >15 pack-years of smoking (vs. never smoking), comparing cases to hysterectomy controls (OR, 1.5; 95% CI, 0.9-2.6; Ptrend=.135). The suggestion of a 40% increased adenomyosis odds was observed with smoking >5-15 pack-years (vs. never smoking), comparing cases to population controls (OR, 1.4; 95% CI, 0.8-2.4; Ptrend=0.136).

CONCLUSION

In the first study of cannabis smoking and adenomyosis risk, no association was observed. However, our data suggested an increased odds of adenomyosis with history of tobacco cigarette smoking. Further research is warranted to replicate our results given the substantial morbidity with adenomyosis and frequency of cigarette smoking and recreational and medical cannabis use.

Respondents' Exposure to Community-based Services and Reported Fertility-Regulation Behavior: A Decade of Data from the Navrongo Community Health and Family Planning Project

This article examines a decade of demographic surveillance data from rural northern Ghana linked with prospective panel survey data recording respondents' reported fertility-regulation behavior. We assess the impact of access to community-based contraceptive services, reported fertility-regulation behavior, and their interaction on the risk of a conception that results in a birth. The effects of service exposure differ by marital status. Reported use of any method to delay or avoid pregnancy appears to be more effective in reducing the risk of conception among the unmarried in areas offering community-based contraceptive services, relative to those in areas where services are facility based. Among both married and unmarried survey respondents who state that they are not using contraception, the risk of conception is lower among women in areas with community-based services than among women in communities without these services. The lower risk of conception among women who are receiving community-based services and who report that they are regulating their fertility may be due to increased efficacy and duration of fertility regulation. Among women who report that they are not regulating their fertility, under-reporting of contraceptive use in experimental areas is likely to play a role in explaining these findings.

A case-control analysis of oral contraceptive use and breast cancer subtypes in the African American Breast Cancer Epidemiology and Risk Consortium

Introduction

Recent oral contraceptive (OC) use has been consistently associated with increased risk of breast cancer, but evidence on specific breast cancer subtypes is sparse.

Methods

We investigated recency and duration of OC use in relation to molecular subtypes of breast cancer in a pooled analysis of data from the African American Breast Cancer Epidemiology and Risk Consortium. The study included 1,848 women with estrogen receptor-positive (ER+) breast cancer, 1,043 with ER-negative (ER-) breast cancer (including 494 triple negative (TN) tumors, which do not have receptors for estrogen, progesterone, and human epidermal growth factor 2), and 10,044 controls. Multivariable polytomous logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for exposure categories relative to never use, controlling for potential confounding variables.

Results

OC use within the previous 5 years was associated with increased risk of ER+ (OR 1.46, 95% CI 1.18 to 1.81), ER- (OR 1.57, 95% CI 1.22 to 1.43), and TN (OR 1.78, 95% CI 1.25 to 2.53) breast cancer. The risk declined after cessation of use but was apparent for ER+ cancer for 15 to 19 years after cessation and for ER- breast cancer for an even longer interval after cessation. Long duration of use was also associated with increased risk of each subtype, particularly ER-.

Conclusions

Our results suggest that OC use, particularly recent use of long duration, is associated with an increased risk of ER+, ER-, and TN breast cancer in African American women. Research into mechanisms that explain these findings, especially the association with ER- breast cancer, is needed.

A qualitative exploration of the use of calendar landmarking instruments in cancer symptom research

Background

Late diagnosis is considered to be a major factor contributing to poorer cancer survival rates in the UK. Interventions have focussed on the promotion of earlier diagnosis in patients with potential cancer symptoms. However, to assess the effectiveness of these interventions, the time from symptom onset to presentation needs to be reliably and accurately measured. This qualitative study explored the use of calendar landmarking instruments in cancer symptom research.

Methods

We performed a secondary analysis of transcripts of interviews using the calendar landmarking instrument, undertaken with patients who had either been diagnosed with cancer (n = 40, IRCO study, Western Australia), or who had symptoms suggestive of cancer (n = 38, SYMPTOM study, North East and Eastern England). We used constant comparison methods to identify use of the calendar landmarking instruments and the impact of their application.

Results

The calendar landmarking instrument appeared to help many patients, either by acting as a prompt or helping to refine recall of events. A combination of personal (e.g. birthday) and national (e.g. Christmas) landmarks seemed to be the most effective. Calendar landmarking instruments appeared more useful where the time period between onset of symptoms and date of first consultation was less than three months. The interviewee’s age, gender and cancer type did not appear to influence whether or not the instrument facilitated recall, and there were no instances where the use of the instrument resulted in the disclosure of a new first symptom. Symptoms of similar chronic conditions could create difficulties when applying the instrument; it was difficult for these participants to characterise and disentangle their symptoms which prompted their decisions to seek help. Some participants tended to prefer to use their own, already personalised, diaries to assist in their recall of events.

Conclusions

This study is the first to describe the potential role of calendar landmarking instruments to support research interviews which explore symptoms and events along the cancer diagnostic pathway. The major challenge remains as to whether they actually improve accuracy of recall.

A comparison of self-reported oral contraceptive use and automated pharmacy data in perimenopausal and early postmenopausal women

PURPOSE

Oral contraceptive (OC) use can occur throughout a woman's reproductive life span with the potential for long-term impacts on health. To assess potential measurement error in prior OC use, this study compared level of agreement between self-reported prior OC use and pharmacy dispensing data in perimenopausal and/or early postmenopausal women.

METHODS

The study's 1399 women (ages, 45-59 years) were participants in a population-based case-control study of the association between OC use and fracture risk. Episodes of lifetime self-reported OC use (in months) were collected, by telephone interview, for January 1, 2008 through November 25, 2012. Pharmacy fills, back to 1980, were collected from automated data. Agreement was measured using the prevalence-adjusted and bias-adjusted kappa index.

RESULTS

The number of women with OC pharmacy fills was 11% to 45% higher than those who reported OC use during each time period. Between-measures agreement was better for more recent use. Prevalence-adjusted and bias-adjusted kappa index values ranged from 0.88 (95% confidence interval, 0.85-0.90) within 5 years from the reference date to 0.65 (95% confidence interval, 0.59-0.71) within 15 to 20 years.

CONCLUSIONS

For studies designed to assess the long-term effects of OC use, the current results are reassuring in noting moderate agreement between self-reported OC use and pharmacy data for up to 15 to 20 years before the interview.

Comparison of prospective daily diaries and retrospective recall to measure oral contraceptive adherence

BACKGROUND

The purpose of this study was to determine if retrospective recall of oral contraceptive (OC) adherence provides data that are similar to data collected via daily diaries over the same time period. Factors associated with inconsistent agreement between prospective and retrospective measurements of adherence also were explored.

STUDY DESIGN

A total of 185 women participated in a longitudinal, prospective cohort of OC users, and 113 of these women provided complete information on OC adherence during follow-ups. Concordance beyond chance was assessed using weighted kappa statistics, and logistic regression was used to identify factors associated with inconsistent reporting of adherence.

RESULTS

There was substantial agreement between prospective and retrospective adherence information (weighted kappa=0.64; 95% CI: 0.52-0.77). Participants with a high school education or less and those who had problems with feeling sad while using OCs had increased odds of inconsistent reporting of adherence (OR=4.38, 95% CI: 1.41-13.61 and OR=3.52, 95% CI: 1.25-9.94, respectively).

CONCLUSION

While prospective data collection via diaries may improve accuracy, the added expense and burden on study participants may not be necessary. However, the use of retrospective recall may not be appropriate for all study populations.

Self-report of current and prior antiretroviral drug use in comparison to the medical record among HIV-infected patients receiving primary HIV care

OBJECTIVE

Patient antiretroviral (ARV) therapy knowledge is essential for regimen adherence, successful therapeutic response, and minimization of resistance evolution. Moreover, a complete and accurate patient ARV history is needed to construct efficacious and tolerable future regimens. In this study we assessed the ability of HIV-infected patients receiving care in a university infectious diseases clinic to accurately recall current and past ARVs.

METHODS

A convenience sample (n = 205) of UNC HIV Clinical Cohort participants (n = 1840) completed a comprehensive in-person interview. Patients were asked about current and ever ARV use and were provided proprietary and generic ARV names and photographs. Self-reported sensitivity for current and ever ARV use (proportion that correctly identified all recorded ARVs), was calculated using the medical record as the gold standard.

RESULTS

One hundred and eighty-five patients had received ARVs at some point after enrollment in the cohort study (ever users). For current ARV use (n = 138), self-reported sensitivity was 63% (95% CI: 54-71). For ever use (n = 185), sensitivity was 18% (95% CI: 13-24).

CONCLUSION

Self-reported cumulative ARV use is not accurate. Since HIV-infected patients are prescribed a number of medications over their treatment course, it is necessary to develop new medication reconciliation techniques that are not dependent on patient memory or knowledge in order to improve patient outcomes.

Reliability and accuracy of interview data in non-smoking female lung cancer case-control study

Background

Valid interview data is critical to the final results of the study. The purpose of this study was to investigate the reliability of epidemiological data obtained in non-smoking female lung cancer case-control study in China.

Methods

Fifty-six pairs of cases and controls, 10% percent of all the collected subjects were re-interviewed by three interviewers who underwent identical standardized training. A limited number of questions included in the original survey were asked again, the responses from the re-interview were compared with the original interview. Kappa was calculated by negative rates of agreement, positive rates of agreement and total rates of agreement to the accordance degree between the two interviews.

Results

The Kappa values were all more than 0.5 in all the studied indexes. The Kappa values descended from 0.92 in family history of cancer to 0.56 in oral contraception use. Errors in collecting and classifying data did occur, and were especially common for complicated clinical events, such as a drug exposure occurring many years before.

Conclusion

We identified four sources of this variability, three in collecting the data, and one in coding. As a result of these findings, strategies are proposed for improving the quality of interview data obtained in epidemiological research. Before finding a good solution, the strategy of data collecting and coding should be simple and easy to inspect.

Applications of calendar instruments in social surveys: a review

Retrospective reports in survey interviews and questionnaires are subject to many types of recall error, which affect completeness, consistency, and dating accuracy. Concerns about this problem have led to the development of so-called calendar instruments, or timeline techniques. These aided recall procedures have been designed to help respondents gain better access to long-term memory by providing a graphical time frame in which life history information can be represented. In order to obtain more insights into the potential benefits of calendar methodology, this paper presents a review of the application of calendar instruments, their design characteristics and effects on data quality. Calendar techniques are currently used in a variety of fields, including life course research, epidemiology and family planning studies. Despite the growing interest in these new methods, their application often lacks sufficient theoretical foundation and little attention has been paid to their effectiveness. Several recent studies however, have demonstrated that in comparison to more traditional survey methods, calendar techniques can improve some aspects of data quality. While calendar instruments have been shown to be potentially beneficial to retrospective data quality, there is an apparent need for methodological research that generates more systematic knowledge about their application in social surveys.

Hormone replacement therapy and breast cancer in former users of oral contraceptives—The Norwegian Women and Cancer study

Combined estrogen-progestin menopausal therapy (HRT) and combined estrogen-progestin contraceptives (OC) both increase breast cancer risk during current use and a few years after. We investigated risk of breast cancer in women who were users of HRT dependant on former history of OC use in a large, national population-based cohort study, the Norwegian Women and Cancer study (NOWAC). Exposure information was collected through postal questionnaires. Based on follow-up of 30,118 postmenopausal women by linkage to national registers of cancer, deaths, and emigration we revealed 540 incident breast cancer cases between 1996 and 2004. Compared to never users of either drugs current use of HRT gave a significant (p = 0.002) higher risk of breast cancer in former OC users, RR = 2.45 (95% CI 1.92-3.12), than among never users of OCs, RR = 1.67 (1.32-2.12). Relative risk of current use of HRT was similar for estrogen only and combinations with progestin added in ever users of OCs. The increased risk of breast cancer in current HRT users with a history of former OC use could have potential great impact on postmenopausal breast cancer risk as the proportion of postmenopausal women with former OC use will continue to increase.

Audio computer-assisted self-interviewing in reproductive health research: reliability assessment among women in Harare, Zimbabwe

PURPOSE

Research on the measurement of HIV risk demonstrates that interview mode can affect reporting; however, few studies have applied these findings to assessments of hormonal contraceptive use. This paper examines how audio computer-assisted self-interviewing (ACASI) influenced reports of hormonal contraceptive use and pregnancy among Zimbabwean women.

METHODS

Using a prospective, randomized, cross-over design, we compared self-reports obtained with ACASI and face-to-face (FTF) interview among 655 women enrolled in a prospective study on hormonal contraceptive use and HIV acquisition. In addition, self-report data were compared to those collected during clinical exams.

RESULTS

Compared to FTF interviews, reports of hormonal contraceptive use were lower in ACASI [odds ratio (OR)=0.6; 95% confidence interval (95% CI)=0.5-0.6], and reports of pregnancy were higher (OR=1.5; 95% CI=1.1-1.9). Both modes of self-report differed from records on contraceptive method disbursement.

CONCLUSION

Although ACASI yielded higher reports of several reproductive health behaviors, discrepancies between self-reports and clinical data on method disbursement highlight persistent measurement challenges.

The Role of Body Weight in Oral Contraceptive Failure: Results from the 1995 National Survey of Family Growth

PURPOSE

Many unintended pregnancies occur in women who use contraception. We conducted this study to determine if increasing body weight is associated with oral contraceptive (OC) failure.

METHODS

This retrospective cohort study consists of the 1916 women who reported using OCs in January 1993 and provided complete covariate information on the 1993 National Health Interview Survey and 1995 National Survey of Family Growth. Body weight and body mass index (BMI) were self-reported in 1993. The outcome was defined to be any conception occurring in women reporting OC use during the month of conception. Cox proportional hazards models were used to model the body weight/BMI-OC failure association.

RESULTS

Women with a BMI >/= 30 had a statistically significant increased risk of having an OC failure as compared to women with BMIs of 20 to 24.9 (HR=1.80, 95% CI, 1.01, 3.20). However, after adjustment for age, marital status, education, poverty, race/ethnicity, parity, and dual method use, this increased risk was attenuated and no longer statistically significant (HR=1.51, 95% CI, 0.81, 2.82). Increasing body weight was not associated with an increased risk of OC failure in the unadjusted or adjusted models.

CONCLUSIONS

We did not find a strong or statistically significant association between increasing body weight/BMI and OC failure among this population of women. Prospective studies specifically designed to examine this association are needed to determine if heavier women should be advised to use a contraceptive method other than OCs to prevent pregnancy.

An interview strategy was critical for obtaining valid information on the use of hormone replacement therapy

OBJECTIVE

We compared telephone reports of hormone replacement therapy (HRT) use to claims for drugs dispensed.

STUDY DESIGN AND SETTING

The study subjects included 106 women who were dispensed HRT and 107 who were not dispensed HRT.

RESULTS

Recall of drug use overall was relatively good (65/79=82.3%, 95% confidence interval [CI] 73.9-90.7). Agreement between recall of drug name and the claims for dispensed drugs was lower (61/79=77.2%, 95% CI 68.0-86.5). Of 65 women reporting use of HRT in response to the indication prompt, nine (13.8%) could not identify the drug name using the drug list prompt, but all 65 women identified a drug using the photo prompt. Recall of start date of drug use was poor (29.2% agreement on month/year; 45.8% agreement within 1 month), and recall of end date of drug use was poorer yet (7.7% agreement on month/year; 21.5% agreement within 1 month).

CONCLUSION

Recall of drug use and drug names is far better than recall of dates of use. Recall can be enhanced with lists of drug names and color photos, but even with memory prompts, recall remains imperfect. If drug use is the primary exposure of interest in a study, considerable effort is needed to collect it correctly. If not, then perhaps drug histories should be omitted.

Oral contraceptive use and risk of breast cancer by histologic type

We examined the association between oral contraceptive use and risk of specific breast cancer histopathologies in a large, multi-center, population-based, case-control study. Women younger than age 75 with a new diagnosis of invasive breast cancer were identified from 4 statewide tumor registries. We compared women with lobular (n = 493) and ductal carcinoma (n = 5,510) to randomly selected controls (n = 9,311). Odds ratios (OR) and 95% confidence intervals (CI) for each histologic type were estimated using polytomous logistic regression, adjusted for other breast cancer risk factors. Current oral contraceptive use was associated with increased risk of lobular carcinoma (OR = 2.6, 95%CI = 1.0-7.1) and there was a significant trend (p = 0.017) of increased risk with more recent use. Oral contraceptive use was not clearly associated with ductal carcinoma (OR = 1.2, 95%CI = 0.8-1.9). These results suggest that the association between oral contraceptive use and risk of breast cancer may vary by histologic type.

Oral contraceptives and venous thromboembolism: a five-year national case-control study

The objective of this study was to assess the influence of oral contraceptives (OCs) on the risk of venous thromboembolism (VTE) in young women. A 5-year case-control study including all Danish hospitals was conducted. All women 15-44 years old, suffering a first ever deep venous thrombosis or a first pulmonary embolism (PE) during the period January 1, 1994, to December 30, 1998, were included. Controls were selected annually, 600 per year in 1994-1995 and 1200 per year 1996-1998. Response rates for cases and controls were 87.2% and 89.7%, respectively. After exclusion of nonvalid diagnoses, pregnant women, and women with previous thrombotic disease, 987 cases and 4054 controls were available for analysis. A multivariate, matched analysis was performed. Controls were matched to cases within 1-year age bands. Adjustment was made for confounding influence (if any) from the following variables: age, year, body mass index, length of OC use, family history of VTE, cerebral thrombosis or myocardial infarction, coagulopathies, diabetes, years of schooling, and previous birth. The risk of VTE among current users of OCs was primarily influenced by duration of use, with significantly decreasing odds ratios (OR) over time: <1 year, 7.0 (5.1-9.6); 1-5 years, 3.6 (2.7-4.8); and >5 years, 3.1 (2.5-3.8), all compared with nonusers of OCs. After adjustment for confounders, current use of OCs with second- (levonorgestrel or norgestimate) and third- (desogestrel or gestodene) generation progestins when compared with nonuse resulted in ORs for VTE of 2.9 (2.2-3.8) and 4.0 (3.2-4.9), respectively. After adjusting for progestin types and length of use, the risk decreased significantly with decreasing estrogen dose. With 30-40 microg as reference, 20 and 50 microg products implied ORs of 0.6 (0.4-0.9) and 1.6 (0.9-2.8), respectively (p(trend) = 0.02). After correction for duration of use and differences in estrogen dose, the third/second-generation risk ratio was 1.3 (1.0-1.8; p <0.05). In conclusion, use of OCs was associated significantly to the risk of VTE. The risk among current users was reduced by more than 50% during the first years of use. The risk increased more than 100% with increasing estrogen dose, and the difference in risk between users of third- and second-generation OCs, after correction for length of use and estrogen dose, was 33%.

Contraceptives and cerebral thrombosis: a five-year national case-control study

The object of this study was to assess the influence of oral contraceptives (OCs) on the risk of cerebral thromboembolic attacks (CTA) including thrombotic stroke and transitory cerebral ischemic attacks. A 5-year case-control study including all Danish hospitals was conducted. All women 15-44 years old suffering a first ever CTA during the period January 1, 1994 to December 31, 1998, were included. Controls were selected annually, 600 per year in 1994-1995, 1200 per year 1996-1998. Response rates for cases and controls were 88% and 90%, respectively. After exclusion of nonvalid diagnoses, pregnant women, and women with previous thrombotic diseases, 626 cases and 4054 controls were available for analysis. A multivariate matched analysis was performed. Controls were matched to cases within 1-year age bands. Adjustments were made for the following potential confounders: year, length of OC use, smoking, hypertension, migraine, family CTA, and years of schooling. There were 212 and 1208 current users of OCs among cases and controls, respectively. The risk of CTA among current users of OCs decreased significantly with decreasing estrogen dose (nonusers reference): OCs with 50 microg, 30-40 microg, 20 microg ethinyl estradiol (EE) and progestin-only pills implied adjusted odds ratios (ORs) (95% CI) of 4.5 (2.6-7.7), 1.6 (1.3-2.0), 1.7 (1.0-3.1), and 1.0 (0.3-3.0), respectively. Current users of OCs with second- (levonorgestrel or norgestimate) and third- (desogestrel or gestodene) generation progestins combined with 20-30 microg EE had ORs of CTA of 2.2 (1.6-3.0) and 1.4 (1.0-1.9), respectively. After correction for differences in estrogen dose, the third- to second-generation risk ratio was 0.6 (0.4-0.9; p = 0.01). In conclusion, high dose OCs and OCs with second-generation progestins were associated with the risk of CTA. The risk increased 2.5 times with estrogen dose increasing from 20 to 50 microg EE, and users of low-dose OCs with second-generation progestins had a 61% higher risk-association of CTA than users of OCs with third-generation progestins.

Risk factors for epithelial borderline ovarian tumors: results of a Swedish case-control study

OBJECTIVE

Borderline ovarian tumors have a favorable prognosis. Previous epidemiological studies indicate common risk factors for invasive epithelial ovarian cancers and borderline tumors, but it remains unresolved whether these tumors are precursors of invasive cancers or a separate disease entity. The objective of this population-based case-control study conducted in 1993-1995 was to examine reproductive and other factors in relation to the risk of borderline ovarian tumors.

METHODS

Subjects were 193 histologically verified incident epithelial borderline tumor cases and 3899 randomly selected controls aged 50-74 years, whose data were collected through mailed questionnaires. Risk estimates were calculated by unconditional logistic regression.

RESULTS

Ever parous women were at reduced risk, with odds ratios of 0.44 (95% confidence interval (CI) 0.26-0.75) for serous and 0.63 (95% CI 0.34-1.19) for mucinous tumors. No clear trends emerged for age at first birth, at menarche, and at menopause. Lactation reduced tumor risk. Oral contraceptive ever use conferred no protection, with odds ratios of 1.40 (95% CI 0.87-2.26) for serous and 1.04 (95% CI 0.61-1.79) for mucinous tumors. The odds ratio for serous tumors following unopposed estrogen ever use was 2.07 (95% CI 1.08-3.95), whereas no risk increase appeared with estrogens supplemented by cyclic or continuous progestins. Mucinous tumors were not associated with hormone replacement therapy. The odds ratio for serous tumors in the highest category of body mass index was 6.47 (95% CI 3.09-13.5).

CONCLUSIONS

Increasing parity and lactation reduce the risk of borderline ovarian tumors in women aged 50-74, while no protection follows oral contraceptive use. Hormonal situations such as unopposed estrogen use and obesity, where estrogens are not counteracted by progestins, may increase the risk of serous tumors.

Influence of question structure on the recall of self-reported drug use

Epidemiological studies often rely on self-reported information as a source of drug exposure. Several studies have evaluated the accuracy of self-reported information on drug use. The influence of question structure on the accuracy of recall, however, has not been studied extensively in these studies. In this study we examined the recall accuracy of questionnaire information on drug use in a ongoing public health survey with special attention to the influence of question structure on sensitivity of recall. A sample of 372 hypertensive subjects for whom questionnaire information and pharmacy records were available was examined. Self-reported information on drug use was obtained through questions about medications used for a specific condition and one final open-ended question. This information was compared with the pharmacy medication history. About 71% of all drugs that were currently in use according to the pharmacy records were recalled through the self-administered questionnaire, and 94% of all drugs mentioned in the questionnaire could be traced in the pharmacy records. Recall sensitivity was higher for questions about medications used for a specific indication (88%) than for the open-ended question (41%). The type of drug that was used might have caused part of this difference in recall. We conclude that questionnaire structure might be of influence on the accuracy of recall of self-reported drug use, and more attention should be paid to the structure of questions on drug use.

Design of validation studies for estimating the odds ratio of exposure-disease relationships when exposure is misclassified

We compared several validation study designs for estimating the odds ratio of disease with misclassified exposure. We assumed that the outcome and misclassified binary covariate are available and that the error-free binary covariate is measured in a subsample, the validation sample. We considered designs in which the total size of the validation sample is fixed and the probability of selection into the validation sample may depend on outcome and misclassified covariate values. Design comparisons were conducted for rare and common disease scenarios, where the optimal design is the one that minimizes the variance of the maximum likelihood estimator of the true log odds ratio relating the outcome to the exposure of interest. Misclassification rates were assumed to be independent of the outcome. We used a sensitivity analysis to assess the effect of misspecifying the misclassification rates. Under the scenarios considered, our results suggested that a balanced design, which allocates equal numbers of validation subjects into each of the four outcome/mismeasured covariate categories, is preferable for its simplicity and good performance. A user-friendly Fortran program is available from the second author, which calculates the optimal sampling fractions for all designs considered and the efficiencies of these designs relative to the optimal hybrid design for any scenario of interest.

Agreement between self-reported antihypertensive drug use and pharmacy records in a population-based study in The Netherlands

From 1987 to 1991, over 36,000 men and women aged 20-59 years have been examined in the Monitoring Project on Cardiovascular Disease Risk Factors in The Netherlands. Classification of the treatment status of hypertensives in this population-based study was based on self-administered questionnaires. In order to assess the accuracy of self-reported antihypertensive drug use we compared the questionnaire information with computerized pharmacy records from a sample of 372 hypertensive subjects. Most antihypertensive drugs that were mentioned in the questionnaire were present in the pharmacy medication history (93%). However, this percentage was less (76%) when a comparison was made with the calculated duration of use based on the number of units prescribed and the directions for use in the pharmacy records. About 94% of the hypertensive subjects who were using an antihypertensive drug according to the pharmacy records, also mentioned at least one antihypertensive drug in the questionnaire. Agreement between self-reported antihypertensive drug use and pharmacy records was consistently high for all classes of antihypertensive drugs. Among 321 (86%) subjects, the number and types of self-reported antihypertensive drugs were exactly the same as in the pharmacy records. In conclusion, the agreement between self-reported antihypertensive drug use and pharmacy records was high, and the self-reported questionnaire information on antihypertensive drug use can be reliably used for the classification of treatment status of hypertensive subjects in this population-based study.

Comparison of self-reported and physician-reported antidepressant medication use

PURPOSE

Self-reported medication histories obtained in pharmacoepidemiologic case-control studies are subject to non-differential misclassification and to recall bias. The accuracy of self-reported antidepressant medication use has never been evaluated, but it is important in light of the hypothesis that antidepressant medications may be associated with cancer risk.

METHODS

Within a case-control study of several cancer sites, we compared self-reported antidepressant medication use with antidepressant use recorded in physicians' records. All female cases (n = 147) and controls (n = 119) who reported antidepressant medication use, and a 10% random sample (n = 114) of those who reported no antidepressant use, were asked to provide consent to contact, and the name(s) of their physician(s). These physicians completed a data abstraction form including information on antidepressant prescriptions recorded in patients' medical records.

RESULTS

Substantial agreement was found between subject- and physician-reported antidepressant medication use (kappa = 0.60 [95% confidence interval (CI), 0.47-0.74]; agreement = 80%), and use of specific antidepressant medications (agreement ranged from 82 to 100%), while moderate agreement was observed for duration of use (weighted kappa = 0.56 (95% CI, 0.32-0.79)), and date of first use [weighted kappa = 0.48 (95% CI, 0.23-0.72)]. The level of agreement did not differ markedly between cases and controls, except for duration of use, where agreement was somewhat greater for cases.

CONCLUSIONS

The similar level of agreement among cases and controls suggests that differential misclassification (e.g., recall bias) is unlikely in the reporting of most aspects of antidepressant medication use by women. Furthermore, the overall accurate self-reporting of antidepressant use suggests that there should be minimal non-differential antidepressant exposure misclassification.

The lack of associations between rheumatoid arthritis and both nulliparity and infertility

OBJECTIVES

Rheumatoid arthritis (RA) is an autoimmune disease that occurs more commonly in women and frequently onsets in women of childbearing age. Pregnancy often causes disease remission, with a subsequent flare postpartum. Nulliparity may be a risk factor for RA, but the literature does not consistently report this finding. There may be a production of antibodies in women with RA that could lead to infertility, and subsequent nulliparity, but this has not been proved. We wanted to determine whether there was a relationship between nulliparity, infertility, oral contraceptive use, and adverse pregnancy outcome in women with newly diagnosed RA.

METHODS

Through a case control study, using a mailed questionnaire, we compared the fertility and pregnancy outcome histories of 34 women between the ages of 19 and 44 years with recent-onset RA with 68 healthy controls matched for age and marital status. The response rate was 97%. A review of the literature also was performed to study the associations between RA and infertility and nulliparity, using Medline searching key references.

RESULTS

We found no association between infertility and the onset of RA. Seventy-one percent of women with RA and 68% of controls had been pregnant. There was a trend toward increased nulliparity in these patients, but the result was not statistically significant (odds ratio [OR], 1.4; P<.6). There were no differences in the number of children (2.6 v. 2.7; P<.6) and parity outcomes in the two groups. Age at first pregnancy was younger in the women with RA (22.6 v. 25.5 years; P<.008), but the education level was higher in the controls (P<.0001), which may explain these differences. Oral contraceptive use was lower in the RA women, but more RA women had long-term use (greater than or equal to 5 years), and neither result was statistically significant. Literature review shows that at best, there are weak negative associations between current estrogen use and RA, and no association with nulliparity and infertility.

CONCLUSIONS

It appears that infertility, the number of pregnancies, and pregnancy outcome are not strongly associated with the risk of developing RA in women of childbearing age. However, in this study there may have been selection biases in the women with RA and the controls that differentially could have affected their reproductive outcomes. Thus, a true association could have been missed. Most other published studies find no association between nulliparity and RA.

Nonsteroidal anti-inflammatory drug use and Alzheimer's disease: a case-control study in Rochester, Minnesota, 1980 through 1984

OBJECTIVE

To compare the frequency of use of nonsteroidal anti-inflammatory drugs (NSAIDs) among 302 incident cases of Alzheimer's disease (AD) and age- and sex-matched control subjects.

DESIGN

We undertook a retrospective case-control study, using the resources of the Rochester Epidemiology Project.

MATERIAL AND METHODS

In ongoing studies of dementia in Rochester, Minnesota, we identified all incident cases of AD with onset between 1980 and 1984. From among all Rochester residents who received care at Mayo Clinic Rochester during those years, we selected one age- (within 3 years) and sex-matched control subject. For this study, exposure to a prescription NSAID was defined as prescribed use for 7 or more days during the 2-year window of time encompassing the year of onset and the year before onset among cases and the corresponding index year and the year prior for control subjects.

RESULTS

The odds ratio (OR) for exposure, as described, to a prescription NSAID versus no exposure to any NSAID was 0.79 (95% confidence interval [CI], 0.45 to 1.38); the OR was 1.00 (95% CI, 0.52 to 1.92) for women and 0.40 (95% CI, 0.13 to 1.29) for men. Similarly, the overall OR for aspirin exposure versus no NSAID exposure was 0.90 (95% CI, 0.54 to 1.50).

CONCLUSION

These data are suggestive but not confirmatory of a protective effect of NSAIDs for AD.

Oral contraceptives and venous thromboembolism. A case-control study

To assess the influence of oral contraceptives (OC) on the risk for venous thromboembolism (VTE) in young women, a 5-year case-control study including all women 15-44 years old suffering a first deep venous thrombosis or a first pulmonary embolism from all Danish hospitals, along with 1200 control subjects during the period 1994-1995, was conducted. Of 586 patient and 1200 control subject questionnaires sent out, 523 patient (89.2%) and 1074 control (89.5%) questionnaires were returned with an agreement to participate. After exclusion of women with nonvalid diagnoses, women who were pregnant, and women with previous VTE or acute myocardial infarction (AMI), 375 patients and 1041 control subjects were available for analysis. Potential tested confounders included: body mass index, length of OC use, family history of VTE, AMI, or stroke, smoking habits, coagulopathies, diabetes, years of schooling, certainty of diagnosis, previous births, and treated hypertension during any pregnancy. A multivariate analysis was performed. Estrogen dose had no influence on the risk for VTE. The risk for VTE among current users of OC was primarily influenced by duration of use, with significantly decreasing odds ratios (OR) over time: < 1 year; 5.1 (3.1-8.5); 1-5 years; 2.5 (1.6-4.1); and > 5 years; 2.1 (1.5-3.1), all compared with those for nonusers of OC. This trend was still significant after adjustment for progestin types. Without adjustment for duration of use, current users of OC with second generation (levonorgestrel or norgestimate) and third generation (desogestrel or gestodene) progestins had OR of 1.8 (1.1-2.9) and 3.2 (2.3-4.4), respectively. After correction for duration of use, however, no significant differences were found between users of OC with different types of progestins. In conclusion, OC increase the risk for VTE significantly. The risk among current users of OC is primarily influenced by duration of use. No difference in risk was found according to estrogen dose, and the difference in risk between different types of progestins was not statistically significant after adjustment for duration of use.

Cerebral thrombosis and oral contraceptives. A case-control study

To assess the influence of oral contraceptives (OC) on the risk of cerebral thrombosis and transient cerebral ischemic attacks, a 5-year case-control study including all Danish hospitals was conducted. All women 15-44 years old who suffered a cerebral thromboembolic attack (CTA) during the period 1994-95 and 1200 age matched control subjects were included. Of 309 patients and 1200 control subjects questionnaires sent out, 271 patients (87.7%) and 1074 control subjects (89.5%) responded and agreed to participate. After exclusion of women with nonvalid diagnoses, previous thromboembolic diseases, or current pregnancy, 219 patients and 1041 control subjects were available for analysis. After confounder control and with nonusers as reference, current users of first generation OC (50 micrograms of ethinyl estradiol [EE] or estrans) had an odds ratio (OR) of CTA of 1.86 (95% confidence interval [CI] 0.88-3.92); users of second generation OC (levonorgestrel or norgestimate) had an OR of 2.37 (1.35-4.16); and users of third generation OC (desogestrel or gestodene) had an OR of 1.32 (0.78-2.22). Users of OC with 50, 30-40, or 20 micrograms EE had OR of 2.65 (1.11-6.34), 1.60 (1.05-2.43), and 1.59 (0.57-4.58), respectively. Odds ratios for specific progestin types were as follows: estrans 1.37 (0.60-3.13), levonorgestrel 2.43 (1.40-4.21), norgestimate 7.09 (1.87-26.8), desogestrel 1.62 (0.72-3.63), and gestodene 1.24 (0.67-2.30). Duration of use was without significant influence on the risk and the OR were constant across the age bands. Compared with women who had never used OC, former users had an OR of CTA of 0.95 (0.66-1.51). In conclusion, use of OC with 50 micrograms of EE and OC with second generation progestins increased the risk of CTA significantly. OC with third generation progestins did not have any significant influence on the risk of CTA. The risk of CTA among former users of OC was not increased.

Prior oral contraceptive use is associated with higher blood pressure in older women

Studies of young, premenopausal women suggest that changes in coronary risk factors do not persist after discontinuation of oral contraceptive use. However, little is known about longterm effects of past oral contraceptive use in older, postmenopausal women, who typically have a more atherogenic profile. This study examines the relation of past oral contraceptive use (including duration) to heart disease risk factors in postmenopausal women. Subjects were 517 women aged 50-69 from the Rancho Bernardo Study who attended follow-up clinic visits. Past oral contraceptive use was reported by 24.2% and two thirds had used oral contraceptives < or = 5 years. After adjustment for covariates, the only significant difference was that past users had higher diastolic blood pressures than never users (p < 0.01). Among older women, the diastolic blood pressure of past users was 3.9 mm Hg higher than that of never users (p < 0.01). Women who used oral contraceptives for < or = 5 years had diastolic blood pressures 3.6 mm Hg higher than those of never users (p < 0.01). No differences were found among younger women or those who used oral contraceptives > or = 6 years. Past oral contraceptive users may be at increased risk for heart disease because of a small but significant elevation in diastolic blood pressure that persists many years after discontinuing oral contraceptive use.

Reproducibility of oral contraceptive histories and validity of hormone composition reported in a cohort of US women

Histories of oral contraceptive (OC) use were reported by 116,686 women aged 25 to 42 years in the Nurses' Health Study II on a self-administered questionnaire accompanied by a color photo booklet of all OC preparations ever marketed in the US. To evaluate the quality of this information, we compared the responses of a randomly selected sample of 215 participants with their data from a subsequent, detailed telephone interview using a structured life events calendar. Agreement for a history of ever having used OC was high between the two methods (exact agreement 99%). Reported durations of use were equivalent (mean duration 42.7 months by telephone interview and 44.6 months by questionnaire). The Spearman correlation for duration of use calculated from the two methods was 0.94 (p < 0.0001). For a subset of women for whom we were able to obtain OC prescription records, the medical record confirmed the use of an identical or equivalent brand in 75% of intervals of reported use. Acceptably valid OC histories were obtained with a self-administered questionnaire.

Knowledge and use of hormonal emergency contraception in Finland

We studied the knowledge and use of hormonal emergency contraception (EC) in Finland by mailing a questionnaire to a national sample of 3000 women aged 18-44 years (response rate 74%). Ten percent of the women aged under 25 and 4% of all respondents had sometimes used EC. Unmarried women were more likely to report having used hormonal EC than were married women, and nulliparous women reported more use than did parous women. However, no statistically significant difference in EC use among women with or without previous abortion history was observed. Older women were less aware of EC than of other methods; only one-third of the women aged over 35 knew about this method. Current contraceptive practices were otherwise similar among ever-users and never-users of EC, but EC users more commonly reported using condom together with oral contraceptives or IUD. Nobody reported using EC as her only contraceptive method. Our findings suggest that EC is appropriately used in Finland, but more information about use of the method is still needed.

Recent oral contraceptive use and risk of breast cancer (United States)

We examined the association between recent oral contraceptive (OC) use and the risk of breast cancer in data from a large population-based case-control study in the United States. Cases (n = 6,751) were women less than 75 years old who had breast cancer identified from statewide tumor registries in Wisconsin, Massachusetts, Maine, and New Hampshire. Controls (n = 9,311) were selected randomly from lists of licensed drivers (if aged under 65 years) and from lists of Medicare beneficiaries (if aged 65 through 74 years). Information on OC use, reproductive experiences, and family and medical history was obtained by telephone interview. After adjustment for parity, age at first delivery, and other risk factors, women who had ever used OCs were at similar risk of breast cancer as never-users (relative risk [RR] = 1.1, 95 percent confidence interval [CI] = 1.0-1.2). Total duration of use also was not related to risk. There was a suggestion that more recent use was associated with an increased risk of breast cancer; use less than two years ago was associated with an RR of 1.3 (CI = 0.9-1.9). However, only among women aged 35 to 45 years at diagnosis was the increase in risk among recent users statistically significantly elevated (RR = 2.0, CI = 1.1-3.9). Use prior to the first pregnancy or among nulliparous women was not associated with increased risk. Among recent users of OCs, the risk associated with use was greatest among non-obese women, e.g., among women with body mass index (kg/m2) less than 20.4, RR = 1.7, CI = 1.1-2.8. While these results suggest that, in general, breast cancer risk is not increased substantially among women who have used OCs, they also are consistent with a slight increased risk among subgroups of recent users.

A comparison of patient interview data with pharmacy and medical records for patients with acquired immunodeficiency syndrome or human immunodeficiency virus infection

To assess the reliability of information reported by patients with AIDS or HIV infection, 123 homosexual men were recruited in Victoria, Australia, in 1992-1993 and interviewed to collect data retrospectively and prospectively on HIV-related illness and treatment. These data were compared with information in medical and pharmacy records. There was no evidence of bias in the patient report of date of diagnosis of HIV infection or date or result of most recent CD4 count, although there was considerable imprecision. Past and current prescription of AZT were recalled accurately by patients compared with medical records (kappa = 0.89 and 0.86, respectively), but previous HIV-related illnesses and enrollment in clinical studies or trials were generally recalled poorly (kappa < 0.4 in some cases). Data collected prospectively on prescription of drugs and use of health care services were reasonably accurately reported by patients. This analysis supports the need for routine assessment of the validity of patient-reported exposure and outcome data in epidemiological studies in order to minimize the possible biases that can result from inaccurate information.

Sixteen-year-old oral contraceptive users in Finland, 1981-1993

While adolescent fertility and abortion rates continue to decline in Finland, regional variations in teenage pregnancies persist, particularly among 16-17-year-olds. We studied differences in the use of oral contraceptives (OCs) among 16-year-old girls according to sociodemographic variables between 1981 and 1993. Questionnaires were mailed to national samples (N = 462-1594) of 16-year-old girls biannually, giving response rates ranging from 86% to 91% (the Adolescent Health and Lifestyle Survey). The increasing trend in OC use levelled off in 1989 at around 17-19%. OC use was most frequent in the lower socioeconomic groups, least frequent in the rural areas, but did not vary significantly according to region of the country. The pace of adopting OCs did not vary across the socioeconomic groups or regions. The socioeconomic and regional distributions of OC use were consistent with previous studies on teenage sexual activity. The results suggest equal accessibility of family planning services; regional variation in pregnancy rates cannot be explained by unequal distribution of OC use.

Recall bias in case-control studies: an empirical analysis and theoretical framework

Previous work has suggested that recall bias in case-control studies may be more serious when the overall study quality is lower. This paper summarizes a systematic literature search to examine the question. All relevant studies published between 1966 and 1990 were included if they met the following criteria: (1) they represented original work, (2) they used a human population, (3) they used a case-control design, (4) they had a "validated" gold standard applied equally to cases and controls and (5) they reported at least one of crude agreement rates, chance-corrected agreement rates (kappa), sensitivity or specificity. Sixteen such studies were identified. No relationship was found between the absolute differences in agreement between cases and controls and the overall level of agreement, in contradiction to suggestions in previous literature. Comparisons of the data quality for cases and controls using either the crude agreement level, kappa, sensitivity, or specificity gave linear relationships with correlations of 0.81, 0.78, 0.58 and 0.62 respectively. Kappas were generally lower than the corresponding crude agreement levels and specificities were higher than sensitivities. When used together, these types of comparisons can give valuable information regarding (1) the possible existence of differential recall in a particular study and (2) the quality of that study, A theoretical framework is proposed for use in these areas.

General practitioner notes as a source of information for case-control studies in young women. UK National Case-Control Study Group

STUDY OBJECTIVE

The UK National Case-Control Study was carried out to investigate the relationship between oral contraceptive use and breast cancer risk. This study investigates whether general practitioner notes could be used as the sole data source for epidemiological studies of young women and what the effect would be on non-response and recall bias.

DESIGN

Case-control study with data on gynaecological, obstetric, and contraceptive history collected at interview and from general practitioners' notes. Information from these two sources was compared.

SETTING

This was a population-based study.

PARTICIPANTS

Altogether 755 women with breast cancer aged under 36 years at diagnosis, each with an age-matched control, participated in the study. Response rates at interview were 72% and 89% for cases and controls but GP data were available for 90% of the 1049 case and first-selected control pairs.

MAIN RESULTS

There was generally good agreement between the two data sources with respect to obstetric history and gynaecological procedures (hysterectomy, oophorectomy, and tubal ligation). The use of intra-uterine devices, or diaphragm, and partner's vasectomy were not reliably recorded in the GP's notes. The overall results of the UK study would have been qualitatively the same with respect to the relationship between oral contraceptive use and breast cancer risk if GP notes only had been used, in spite of the fact that only about half of all oral contraceptive usage was recorded in the notes. Response rates would have been higher, recall bias eliminated, and the cost of the study halved.

CONCLUSIONS

When planning case-control studies in young women, the possibility of using GP notes as the primary data source should be considered. Lack of data on potential confounding factors is a possible drawback to such use. The practice of destroying GP's notes shortly after the death of patients seriously restricts the possibility of using these notes when studying rapidly fatal conditions.

Completeness of prescription recording in outpatient medical records from a health maintenance organization

Since validity of drug data is often characterized by the agreement between questionnaire and medical record data, medical record completeness for drug therapy was evaluated. Outpatient medical records of 501 randomly selected Group Health Cooperative Health Maintenance Organization (HMO) participants known to have been dispensed a non-steroidal anti-inflammatory drug (NSAID) were reviewed for completeness. Documentation was also evaluated for indication, age, gender, location of visit, and number of visits. The NSAID was recorded in 89% of the charts reviewed. Charts with a vague/missing diagnosis were more likely to lack documentation [RR (95% CI): 20.0 (13.2-30.3)]. The data suggested that poorer documentation was related to older age [1.5 (0.9-2.5)] but not to gender, indication, location of visit, or number of visits. Presence of an indication was positively correlated to the completeness of drug documentation. According to this staff/group HMO, there may be a higher concordance between self-reported medication use and medical record data than found in previous literature.

Childbearing and use of oral contraceptives: impact of educational level. The Nordland Health Study

STUDY OBJECTIVE

The aim was to correlate the level of education to childbearing pattern and previous use of oral contraceptives in middle aged women.

DESIGN AND SETTING

The study was a cross sectional survey of 3608 women aged 40-42 years in a county in northern Norway and involved 72% of all women in the age bracket living in the county. Information about childbearing and use of oral contraceptives was obtained from a self administered questionnaire.

MAIN RESULTS

Level of education was inversely associated with parity and positively related to age at first and last birth, to the use of oral contraceptives at any time, and particularly to the use of oral contraceptives before the first birth.

CONCLUSIONS

Women with high level of education have relatively low parity and postpone childbearing by using oral contraceptives.

Agreement between oral contraceptive users and prescribers: implications for case-control studies

Case-control studies examining the effects of oral contraceptives (OC) are prone to misclassification bias due to errors in assessment of OC use. Concern about inaccurate exposure histories has increased since current studies require women to recall OC use over prolonged periods of time. In preparation for a case-control study of breast cancer and OC use, an investigation was carried out to assess agreement between women's lifetime histories of OC use (covering a period of up to 20 years) and prescribers' records. OC histories were obtained during personal interview with 218 women who had used OC at some point in their lives (127 breast cancer patients, 91 controls). Recall was aided by an album with color photographs of all OC marketed in the Netherlands from 1962 onwards (n = 65), and a calendar that covered the women's life span from date of birth to menopause. The participants were asked for the names of all physicians who prescribed OC for them. The rate of response from the prescribers was high (94%), but only half of the forms provided useful information. Patient-prescriber agreement on brand names (including dosage) was 70%. About half of the women agreed with their prescribers on starting dates to within less than a year's difference. Approximately the same percentage of agreement was found for stopping dates. Multiple linear regression indicated that agreement on brand names and dates of usage was lower for women of low socioeconomic status, for healthy women (as compared to breast cancer patients) and for periods of pill use that had to be recalled from the more distant past. Agreement on total duration of use was high enough to permit testing of a moderately strong duration-response relationship in a case-control study.

Comparison of questionnaire information and pharmacy data on drug use

Information on chronic drug use at any time in the past was collected with a self-administered questionnaire as part of a prospective cohort study on diet, other life-style factors and cancer among subjects aged 55-69 years. The validity of the questionnaire information on drug use was evaluated among 207 subjects by comparing it to pharmacy records of dispensed drugs. The comparison could be made for the 2.5-year period preceding the questionnaire administration. Since the study subjects did not mention prescription drugs that were not dispensed by their pharmacy, indicating no errors of commission, the analyses were focused on the estimation of sensitivity of drug recall and its correlates. Questionnaire recall of drug use amounted overall to 61.2% of drugs prescribed to the subjects for at least 6 months. Drug recall decreased with increasing age and with increasing number of prescribed chronic use of drugs per subject. No difference in recall was observed between men and women. Recall tended to improve with increasing duration of use and varied with type of drug. When only long-term drug use at the time of questionnaire administration was considered, overall recall of drug use was 68.8%.

Ascertainment of risk factors for osteoporosis: comparison of interview data with medical record review

To determine the accuracy of self-reported risk factors for osteoporosis, an age-stratified random sample of Rochester, MN, women was studied. Results from a structured face-to-face interview were compared with information documented in contemporary inpatient and outpatient health care records in the community. Using the kappa statistic to evaluate concordance of these two data sources, we found substantial agreement for a history of proximal femoral and distal forearm fractures, peptic ulcer disease, estrogen replacement therapy and oral contraceptive use, and cigarette and alcohol exposure. Moderate agreement was seen for histories of other age-related fractures, hysterectomy or oophorectomy, thyroidectomy, and use of thyroid supplements. Poor agreement was found for prior thyroid disease, gastrectomy, and corticosteroid or anticonvulsant use. This study demonstrates a need for greater attention to the quality of self-reported risk factor data in studies of bone loss and fractures.

The pill, parity, and rheumatoid arthritis

We report on a case-control study investigating the relationship of oral contraceptive pill (OCP) use and parity to the development of rheumatoid arthritis (RA). Women with RA were compared with 2 separate control groups, women with osteoarthritis (OA) and women randomly selected from a population-based electoral register. Nulliparity was found to be a risk factor for the development of RA, with age-adjusted odds ratios of 1.82 (95% confidence interval [CI] 1.09-3.03) versus the OA control group and 1.83 (95% CI 1.03-3.06) versus the population control group. Use of OCPs before the age of 35 was negatively associated with RA (odds ratio 0.56, 95% CI 0.29-1.12 versus the OA control group; odds ratio 0.6, 95% CI 0.30-1.17 versus the population control group). Some evidence of a duration-response effect was seen, although the numbers were small. The 2 variables were also multiplicative, with nulliparous non-OCP users having a 4-fold risk of RA compared with parous OCP users. These findings suggest that pregnancy and OCP use have a "protective effect" on the development of RA, although the mechanism remains unclear.

The protective effect of the oral contraceptive pill on rheumatoid arthritis: an overview of the analytic epidemiological studies using meta-analysis

The oral contraceptive pill (OCP) has been implicated as having a protective effect on the development of rheumatoid arthritis (RA). The results of 12 studies have now been reported and produced differing results and conclusions. Because of the discrepancy in results and the importance of the issue we undertook a review of the studies and performed a meta-analysis. In all, 9 independent studies satisfied the criteria for selection, 6 case-control design and 3 longitudinal. Using standard meta-analysis techniques, the overall pooled odds ratio for all the studies was 0.68 for the crude results (95% CI 0.58-0.78) and 0.73 for the adjusted results (95% CI 0.61-0.85). The graphical odd-man-out method produced a 94% interval of 0.70-0.72. The pooled odds ratio of the case-control studies was lower than for the longitudinal studies. However, subdividing studies by the type of case source produced a pooled odds ratio for studies using hospital-based cases of 0.49 (95% CI 0.39-0.63) which was considerably less than that of studies using population-based cases: 0.95 (0.78-1.16). This difference was unlikely to have explained by bias due to selection of controls. We suggest that OCP use may not have a "protective effect" on the development of RA but may prevent the progression to severe disease by modifying the disease process.

The Rochester Coronary Heart Disease Project: effect of cigarette smoking, hypertension, diabetes, and steroidal estrogen use on coronary heart disease among 40- to 59-year-old women, 1960 through 1982

A population-based case-control study of coronary heart disease (CHD) risk in young women attributable to cigarette smoking, hypertension, diabetes, and steroidal estrogen use was conducted among residents of Rochester, Minnesota. All newly diagnosed cases of CHD (sudden unexpected death [SUD], N = 18; myocardial infarction [MI], N = 90; and angina, N = 133) among female Rochester residents 40 to 59 years of age during the years 1960 through 1982 were identified, and two community control subjects were matched for age and duration of community medical record. The overall adjusted odds ratio (OR) for the association between steroidal estrogen use and definite CHD (MI and SUD) was 0.6 (95% confidence interval [CI] = 0.2 to 1.3). Smoking (OR = 5.1; 95% CI = 2.3 to 11.6), hypertension (OR = 4.8; 95% CI = 2.3 to 10.2), and diabetes (OR = 8.4; 95% CI = 1.6 to 44.5) were strong risk factors for CHD events. If considered causal, cigarette smoking accounted for 64% of all MIs and SUDs in the community, hypertension accounted for 45%, and diabetes accounted for 13%. Although steroidal estrogen exposure reduced CHD among these women by 14%, giving steroidal estrogens to all women in this age group might reduce the population rates of MI by as much as 45%.

Duration of relative protection of screening for cervical cancer

The reduction in cervical cancer among women with at least one previous negative Pap smear can be assessed in terms of the time elapsed since the last smear was taken. One indicator of the utility of screening is the duration of relative protection, commonly calculated as the inverse of the disease odds ratio. Most investigations of the extent of relative protection provided by Pap testing have relied on data from centrally organized screening programs or case-control studies. For geographic areas without mass screening programs or tumor registries, reliance on subject recall of Pap tests is required. We conducted a case-control investigation of cervical cancer and interviewed 153 Maryland women with invasive disease and two control groups: 153 case-nominated controls and 392 randomly selected controls. The duration of relative protection of screening for disease was 4-6 years for both control groups [relative protection (RP) = 4.30, 95% confidence interval (CI) = 1.5-12.7 for neighborhood controls, RP = 3.63, 95% CI = 1.4-9.6 for random controls]. These findings held after adjusting for education, ever treated for a sexually transmitted disease, smoking, age at menarche and at first sexual intercourse, number of pregnancies, lifetime contraceptive use, and utilization of obstetrician-gynecologist services; the RPs increased upon adjustment. However, there was a decline in the RP with increased duration. Our findings are directly comparable to reports where smears have been verified, suggesting that self-reports of previous tests may be reliable as a method to evaluate the utility of screening.

The independent associations of parity, age at first full term pregnancy, and duration of breastfeeding with the risk of breast cancer. Cancer and Steroid Hormone Study Group

Although the important influence of a woman's reproductive history on her risk of breast cancer is widely recognized, it is not clear whether this is wholly accounted for by the age at her first full-term pregnancy, or whether there are additional, independent influences of breastfeeding or number of children. To examine the respective contributions to the risk of breast cancer of these reproductive factors, we used logistic regression methods to analyze data from a multicenter case-control study, the Cancer and Steroid Hormone Study. Included in the analysis were 4599 women, 20-55 years of age, identified as having an initial diagnosis of breast cancer by one of eight collaborating population-based cancer registries. The 4536 controls were women of similar ages selected by random dialing of households with telephones in the same eight areas. As expected, age at first full-term pregnancy exerted a strong influence on the risk of breast cancer. However, after it and other potentially confounding factors had been controlled for, parity and duration of breastfeeding also had a strong influence on the risk of breast cancer. Compared with women of parity one, women of parity seven or greater had an adjusted relative risk of breast cancer of 0.59 (95% CL, 0.44-0.79). Compared with parous women who never breastfed, women who had breastfed for 25 months or more had an adjusted relative risk of 0.67 (0.52-0.85). These results do not support the supposed preeminent importance of age at first full-term pregnancy among the reproductive determinants of breast carcinogenesis. Resolution of this issue may have important implications for elucidating hormonal influences on breast cancer and for projecting future trends in the disease.

A comparison of physician and patient reports of Pap smear histories

We describe the level of agreement between patients and their physicians with respect to their use of the Pap smear, the symptoms present, and the smear results. The data derived from a case-control interview study of 181 newly diagnosed invasive squamous cell cervical cancer cases and 905 age-matched controls, a second case-control interview study of 250 cases of cervical dysplasia and 500 age-matched controls, and the medical records of attending physicians for all patients. Cases and controls in both studies reported more smears over the previous 5 years than their medical records indicated; also patients reported their last smear as more recent than noted in physician files. Cancer cases were somewhat more accurate on frequency and timing of smears than the other patient groups. Patients tend to report more symptoms at interview than recorded in their files. Disagreements on smear results were probably because of differences in terminology, especially in distinguishing invasive from pre-cancerous conditions; cancer cases tended to report some dysplasias as cancer, but the controls in both studies under-reported dysplasia. If high physician response rates can be obtained, we would prefer to use physician records as the basis of evaluation of screening programs, especially when accurate dating of screening events is required. However it is less clear whether physician records are to be preferred, if the evaluation is to take symptoms into account.

Repeatability of interview-derived socio-demographic and medical information

In this report, the repeatability of socio-demographic and medical information elicited by means of an interviewer-administered questionnaire is examined. One hundred and seven women who were controls in case-control studies of breast disease were re-interviewed approximately one year after initial interview. Both interviews addressed the same time period, namely that which preceded the first interview. For most of the quantitative variables studied, at least 70% of individuals provided two estimates (original and repeat) which were within one unit of each other. Also, comparison of the categorization of individuals on repeated measurements of qualitative variables (reports of events, operations and habits) revealed that most individuals were categorized consistently. This strong agreement between the original and repeat estimates was also reflected in the relatively high values of the summary measures of agreement (the intraclass correlation coefficient for quantitative variables and the Kappa statistic for qualitative variables) for most of the variables studied. The results of the present study on repeatability complement the results of previous research into the criterion validity of questionnaire-derived information and, taken together, suggest that information of relatively high quality may be obtained from such sources.