Sampler of findings from the 1986 national mortality followback survey on risk factors, disability, and health care

The National Center for Health Statistics conducted a mortality followback survey of a national probability sample drawn from all deaths of U.S. adults in 1986 and an oversampling of deaths of persons with selected characteristics. Responses were received from the next of kin or other close relatives of 16,598 adult decedents (88.6 percent). Data were collected through a mail questionnaire, followed by telephone or personal interviews with nonrespondents. Data were also collected from the hospitals and other health care facilities used by the decedent in the last year of life. Illustrative results are presented on the four major subject areas studied: risk factors for premature death, disability and care in the last year of life, socioeconomic differentials, and the reliability of selected items reported on the death certificate. Researchers are encouraged to explore the data tape to pursue indepth epidemiologic studies.

An overview of demographic analysis as a method for evaluating census coverage in the United States

"Since the 1950 U.S. census, demographic methods based on the fundamental balancing equation of demography have played an important role in the evaluation of the census net undercount. Application of this set of methods, called demographic analysis, results in national estimates of the net undercount for age-race-sex groups. Although results of demographic analysis are readily available in Bureau of the Census publications, the procedures used to estimate each of the components of population change are less well-known. In this paper we review the historical foundation of demographic analysis, beginning with Coale's 1950 census evaluation project and concluding with the recent evaluation of the 1990 census. We examine each of the components of the method, how their estimation has changed over time, and how they were estimated for the 1990 census."

Oral contraceptives and breast cancer. A review with some comments on mathematical models

The relationship between oral contraceptive use and breast cancer is discussed on the basis of information given in review articles, meta-analyses and editorials emphasizing methodological problems related to bias and confounding. Over the last few years a shift in opinion has taken place. Most reviewers now consider that long-term use of oral contraceptives is associated with an increased risk of premenopausal breast cancer and no effect among postmenopausal breast cancer. This result is compatible with an additive effect (in rate measure scale) of oral contraceptive use on breast cancer risk.

Occupational hazards and pregnancy outcomes

This study examined the association between exposure to occupational hazards and pregnancy outcomes using data from a case-control study conducted in 29 hospitals in Shanghai, China. The sample included 1,875 perinatal deaths and newborns with birth defects and the same number of controls. Information on mother's exposure to occupational radiation, chemicals, noise, and pesticides was investigated. Logistic regression analysis controlling for potential confounders showed that exposure to radiation before/during pregnancy was associated with antepartum fetal death, birth defects, small-for-gestational-age (SGA), and threatened abortion. Exposure to chemicals before/during pregnancy was associated with antepartum fetal death, early neonatal death, birth defects, preterm birth, and threatened abortion. Women exposed to pesticides during pregnancy had an increased risk of SGA and threatened abortion. Exposure to occupational noise during pregnancy increased the risk of antepartum fetal death. Furthermore, higher than expected numbers of congenital anomalies in the central nervous system (CNS) were identified among women exposed to chemicals before pregnancy and to pesticides during the first trimester of pregnancy. No significant association was found between occupational exposure and intrapartum fetal death. Although recall bias may be possible in our study, the findings encourage further research.

Estrogen replacement therapy and the risk of breast cancer: results from the case-control surveillance study

To examine the relation of noncontraceptive estrogen use to the risk of breast cancer among postmenopausal women, the authors conducted a case-control study: 1,686 cases were compared with 2,077 hospital control subjects, of whom 1,120 had non-gynecologic cancers and 957 had nonmalignant (also non-gynecologic) conditions. Data were obtained from 1980 to 1986, by interview of subjects in hospitals in the United States and Canada. The relative risk estimate for any use of replacement estrogens unopposed by progestogens was 1.2 (95% confidence interval (Cl) 1.0-1.4), after adjustment for age and type of menopause; when all known risk factors for breast cancer were taken into account in a multivariate analysis, the estimate was similar. For use of at least 15 years duration, the estimate was 0.9 (95% Cl 0.5-1.9). Most of the unopposed use was of conjugated estrogens: overall, the relative risk (95% Cl) was 1.3 (1.0-1.6); for durations of 15 or more years, it was 0.9 (0.4-1.9); for use of 5 years followed by a latent interval of 15 or more years, it was 1.3 (0.7-2.4); and for current use it was 1.1 (0.7-1.6). There was no evidence of increased breast cancer risk when the conjugated estrogen users were divided according to dose. There was little use of estrogens opposed by progestogens; the relative risk estimate was 1.7 (95% Cl 0.9-3.3). The results of this large study provide no evidence that the use of unopposed conjugated estrogens increases the risk of breast cancer, even after long durations of use or long latent intervals, but the possibility of a modest increase (less than a doubling) could not be excluded. There were insufficient data to evaluate the effects of nonconjugated estrogens and of combined estrogen and progestogen therapy.

Further analysis of contraceptive failure of the ovulation method

Reanalysis of data on the ovulation method of natural family planning collected by the World Health Organization yields the following conclusions. The method is effective during perfect (correct and consistent) use, with a first-year probability of failure of 3.4%. However, it is extremely unforgiving of imperfect use, with a first-year probability of failure of 84.2% if the method is not used correctly. During the initial year, 87% of the cycles were characterized by perfect use. Nevertheless, the 13% of cycles characterized by imperfect use had a tremendous impact on the overall failure rate. During the first year of typical use 22.5% of the women in the clinical trial became accidentally pregnant.

Total error in PES estimates of population

"We describe a methodology for estimating the accuracy of dual systems estimates (DSE's) of population, census estimates of population, and estimates of undercount in the census. The DSE's are based on the census and a post-enumeration survey (PES). We apply the methodology to the 1988 dress rehearsal census of St. Louis and east-central Missouri and we discuss its applicability to the 1990 [U.S.] census and PES. The methodology is based on decompositions of the total (or net) error into components, such as sampling error, matching error, and other nonsampling errors. Limited information about the accuracy of certain components of error, notably failure of assumptions in the 'capture-recapture' model, but others as well, lead us to offer tentative estimates of the errors of the census, DSE, and undercount estimates for 1988. Improved estimates are anticipated for 1990." Comments are included by Eugene P. Ericksen and Joseph B. Kadane (pp. 855-7) and Kenneth W. Wachter and Terence P. Speed (pp. 858-61), as well as a rejoinder by Mulry and Spencer (pp. 861-3).

Modelling digit preference in fecundability studies

Retrospective studies of fecundability, in which women are asked how many cycles they required to become pregnant, are often affected by problems of digit preference. A probability model for such digit preference is proposed in which misreporting favours 6 or 12 (and possibly also 3) cycles. It is assumed that in the absence of misreporting the number of cycles follows a beta-geometric distribution. The model is applied to two data sets, with clear-cut results: The inclusion of additional parameters to model the misreporting can lead to substantial improvements in fit, but causes little change to the estimated parameters of the underlying beta-geometric distribution. In some cases misreporting parameters may be regarded as nuisance parameters, while in others they may be of interest. We have found estimates of these parameters to vary between different categories of women in an interpretable manner. The models may also be used to estimate the percentage of couples in any study that misreport their conception waiting time.

Response bias in a case-control study: analysis utilizing comparative data concerning legal abortions from two independent Swedish studies

Independent reports of legal abortions in two Swedish epidemiologic studies of breast cancer in young women, covering the same women and overlapping the same time period, have been compared in order to estimate a putative response bias. One study used case-control methods and obtained data by retrospective interviews from 317 cases and 512 controls. The other study was based on objectively documented information froma nationwide registry covering legally induced abortions. Analysis demonstrated a ratio between the odds ratios from the two studies of 1.5 (95 percent confidence interval 1.1-2.1) and an observed ratio of 22.4 (p less than 0.007) between underreporting of previous induced abortions among controls relative to overreporting among cases. This response bias may explain the tendency toward increased risk of breast cancer which, according to several case-control studies, appears to be associated with induced abortion.

Her-2/neu and INT2 proto-oncogene amplification in malignant breast tumors in relation to reproductive factors and exposure to exogenous hormones

In previous studies in southern Sweden, early use of oral contraceptives has been found to be accompanied by an increased risk of developing premenopausal breast cancer, and the tumors developing in these patients have shown a more aggressive behavior. In the present study, amplification of the proto-oncogenes Her-2/neu (also known as ERBB2) and INT2 was studied in primary tumor specimens from 72 premenopausal women and was related to starting age of oral contraceptive use and other reproductive risk factors. Amplification of Her-2/neu was more common among early oral contraceptive users (i.e., those starting at less than or equal to 20 years of age) than among nonusers or late users (odds ratio [OR], 5.3; 95% confidence interval [CI], 1.6-16.7), whereas INT2 amplification did not differ significantly among those groups (OR, 0.9; 95% CI, 0.1-5.0). The likelihood of INT2 amplification was greater among users of progestins and those with a history of abortions before the first full-term pregnancy (OR, 9.0; 95% CI, 1.3-51.7; and OR, 18.6; 95% CI, 2.2-165.8, respectively). No significant relationships were found between proto-oncogene amplification and the variables of parity, age at first full-term pregnancy, or late abortion. The increased ORs persisted after adjustment for age at diagnosis and other risk factors. The findings suggest that the higher rate of Her-2/neu amplification among early oral contraceptive users is an effect of the oral contraceptive use per se rather than of the relative youth of the users. Moreover, the relationship between progestin use and early abortion and amplification of the INT2 gene is biologically plausible.

Adjustment of provisional mortality series: the dynamic linear model with structured measurement errors

The authors "consider the problem of adjusting provisional time series using a bivariate structural model with correlated measurement errors. Maximum likelihood estimators and a minimum mean squared error adjustment procedure are derived for a provisional and final series containing common trend and seasonal components. The model also includes measurement errors common to both series and errors that are specific to the provisional series. [The authors] illustrate the technique by using provisional data to forecast ischemic heart disease mortality."

Breast cancer risk and oral contraceptive use: results from a large case-control study

The association between oral contraceptive use and breast cancer risk was examined using data from a case-control study of breast cancer in Long Island, New York. Cases were defined as female residents of Nassau and Suffolk Counties between the ages of 20 and 79, diagnosed with breast cancer between January 1, 1984 and December 31, 1986. Age- and county-matched controls were selected from driver's license files. Among all women under age 70 at diagnosis, there was no association between oral contraceptive use and breast cancer; there was, however, a positive association in the subgroup ages 20-49 (adjusted odds ratio = 1.68, 95% CI: 1.16-2.42). Risk increased with increasing duration of use, but did not differ between women who first used oral contraceptives before the first pregnancy and those who first used them later, or between women who first used oral contraceptives before age 25 and those who first used them at a later age. Risk also appeared to increase with number of years of use before the first pregnancy or before age 25, although numbers were small. History of benign breast disease did not influence risk. The association of breast cancer risk with oral contraceptive use appeared stronger in women from Suffolk County than Nassau County.

Poisson regression with misclassified counts: application to cervical cancer

"Errors in disease classification can give misleading inferences for covariate effects when the probability of error is itself related to the covariates. More accurate inferences are possible using supplemental data on both true and fallible disease counts at various covariate levels. We present a method for incorporating such supplemental data into disease rate regression and use it to show that, although observed intercountry differences in European cervical cancer mortality rates are exaggerated by errors in death certification, they are not completely explained by such errors." The data concern Belgium, England and Wales, France, and Italy.

The impact of forecasting methodology on the accuracy of national population forecasts: evidence from the Netherlands and Czechoslovakia

"This study considers the accuracy of national population forecasts of the Netherlands and the Czechoslovak Socialist Republic.... We look at the demographic components employed in each forecast, the procedure to extrapolate fertility and the level at which assumptions for each component are formulated. Errors in total population size, fertility, mortality and foreign migration, and age structure are considered. We discuss trends in errors and methodology since 1950 and compare the situations in the two countries. The findings suggest that methodology has only a very limited impact on the accuracy of national population forecasts."

Reporting errors in time-to-pregnancy data collected with a short questionnaire. Impact on power and estimation of fecundability ratios

Few tools exist in reproductive epidemiology for studying adverse effects on fertility. Data on time to pregnancy (the number of menstrual cycles required to conceive) can be used to estimate fecundability ratios, a sensitive endpoint for identifying factors associated with reduced fertility. Time-to-pregnancy data can be collected in detailed interviews. The accuracy of data collected on brief, self-administered questionnaires is not known. In a study of occupational exposures to dental assistants conducted in 1987-1988, 523 women provided time-to-pregnancy data both on a short, mailed questionnaire and in a detailed telephone interview. The correlation between the two measures was 0.82. Assuming that the detailed data were accurate, reporting errors in data from the short form were distributed nondifferentially with respect to most covariates of interest in fecundability analyses. Simulation studies were conducted to estimate bias and loss of power from the misclassification. Bias was toward the null. Substantial power was lost in detecting weak exposures. However, exposures that reduce fecundability by 50 percent (equivalent to adding about three cycles to the median time to pregnancy) could still be detected with 80 percent power in samples of about 100 women (half of them exposed to a possible toxin). The authors conclude that time-to-pregnancy data collected with a few self-administered questions can be useful in a variety of epidemiologic studies, including occupational and environmental surveillance programs.

Oral contraceptives and breast cancer: review of the epidemiologic literature

Oral contraceptives have caused little or no overall increase in risk of breast cancer in women in developed countries, even in long-term users or after more than two decades since initial exposure. Limited evidence suggests a possible small increase in risk in users of oral contraceptives in developing countries, and further studies of breast cancer and oral contraceptives in low risk populations are warranted. Oral contraceptives may enhance risk of breast cancer in young women with a history of benign breast disease, and this possibility should also be investigated further. Multiple investigations have fairly consistently shown risk of breast cancer in women under age 45 years to be increased slightly in long-term users of oral contraceptives. Further studies should be conducted to elucidate the mechanism of this apparent phenomenon.