Potential sources of bias and confounding in environmental epidemiologic studies of pregnancy outcomes.

Biases familiar to all epidemiologists take on new importance when studying spontaneous abortion because of the significance of the timing of exposure during gestation, and because the gestational age at which the pregnancy is recognized may be related to risk factors under study. In this paper, we systematically review the principal biases that might affect epidemiologic studies of pregnancy loss, in the context of a series of studies that found associations between adverse reproductive outcomes and prenatal use of tap or bottled water. These biases relate to availability of subjects for study, exposure opportunity, recall of exposure, recall of disease, and confounding. In one of the studies, the rate of bottled water use changed dramatically over the course of the study period, providing an opportunity to test indirectly for biased recall of exposure. We conclude that a less complete recall of water intake among women with normal pregnancies, as compared with those with spontaneous abortions, may have produced, or increased the magnitude of, the associations seen in these studies.     

Recall bias in epidemiologic studies

The factors which contribute to bias due to differential recall between cases and controls in retrospective studies have been little studied. A review of the literature on recall accuracy suggests that the extent of inaccurate recall is related to characteristics of the exposure of interest and of the respondents, though a distinction must be drawn between recall which is biased and that which is simply inaccurate. Interviewing technique and the study protocol, including the design of questionnaires and the motivation of respondents, play a central role and are under the control of the investigator. The results of validation studies carried out to date suggest that the likelihood of recall bias may be greater when recall is poor in general.     

The potential impact of incomplete radiography histories in epidemiologic studies that rely on prepaid health plan records

Members of prepaid health plans are often considered a group for which complete histories of exposure to medically-related risk factors are known. In a population-based case-control study of radiography as a risk factor for acute myelogenous leukemia, extensive retrospective exposure data from an average of 71% of subjects' health care providers were collected. It was determined that at least 21% of prepaid health plan members received diagnostic imaging procedures outside of their plans over the 10-y period of interest and that these "outside" procedures constituted 43% of their total red bone marrow dose from all radiographic imaging. Procedures significantly more likely to have been received outside of prepaid health plans were conventional spine x rays (p = 0.004) and coronary angiograms (p < 0.0001). All "outside" coronary angiograms and 50% of "outside" spine x rays were received at hospitals. Epidemiologic studies that restrict exposure to that received within prepaid health plans may miss a substantial portion of exposure, resulting in underestimated risk estimates and a loss of statistical power.     

Geographic bias related to geocoding in epidemiologic studies

Background  

This article describes geographic bias in GIS analyses with unrepresentative data owing to missing geocodes, using as an example a spatial analysis of prostate cancer incidence among whites and African Americans in Virginia, 1990–1999. Statistical tests for clustering were performed and such clusters mapped. The patterns of missing census tract identifiers for the cases were examined by generalized linear regression models.     

Models for retrospective quantification of indoor radon exposure in case-control studies

In epidemiologic studies on lung cancer risk due to indoor radon the quantification of individual radon exposure over a long time period is one of the main issues. Therefore, radon measurements in one or more dwellings, which in total have been inhabited by the participants for a sufficient time-period, are necessary as well as consideration of changes of building characteristics and ventilation habits, which influence radon concentration. Given data on 1-y alpha-track measurements and personal information from 6,000 participants of case-control studies in West and East Germany, an improved method is developed to assess individual radon exposure histories. Times spent in different rooms of the dwelling, which are known from a personal questionnaire, are taken into account. The time spent outside the house (average fraction 45%) varies substantially among the participants. Therefore, assuming a substantially lower radon exposure outside the dwelling, the residence time constitutes an important aspect of total radon exposure. By means of an analysis of variance, important determinants of indoor radon are identified, namely constant conditions such as type of house (one family house or multiple dwelling), type of construction (half-timbered, massive construction, lightweight construction), year of construction, floor and type of basement, and changeable conditions such as heating system, window insulation, and airing habits. A correction of measurements in former dwellings by factors derived from the analysis is applied if current living conditions differ from those of the participants at the time when they were living in the particular dwellings. In rare cases the adjustment for changes leads to a correction of the measurements with a factor of about 1.4, but a reduction of 5% on average only. Exposure assessment can be improved by considering time at home and changes of building and ventilation conditions that affect radon concentration. The major concern that changes in ventilation habits and building conditions lead to substantial errors in exposure (and therefore risk) assessment cannot be confirmed in the data analyzed.     

A review of residential radon case-control epidemiologic studies performed in the United States.

Lung cancer is the leading cause of cancer death in the United States for both men and women. Although most lung cancer deaths are attributable to tobacco usage, even secondary causes of lung cancer are important because of the magnitude of lung cancer incidence and its poor survival rate. This review summarizes the basic features and major findings from the published U.S. large-scale residential radon case-control studies performed in New Jersey, Iowa, and Missouri (two studies). The methodology from an unpublished study covering Connecticut, Utah, and Southern Idaho is also presented. Overall, the higher categorical risk estimates for these published studies produced a positive association between prolonged radon exposure and lung cancer. Two studies (Missouri-II and Iowa) that incorporated enhanced dose estimates produced the most compelling evidence suggesting an association between prolonged residential radon exposure and lung cancer. The prevailing evidence suggests that the statistically significant findings may be related to improved retrospective radon exposure estimates. The general findings from the U.S. studies, along with extrapolations from radon-exposed underground miners, support the conclusion that after cigarette smoking, prolonged residential radon exposure is the second leading cause of lung cancer in the general population.     

Potential for gene-gene confounding bias in case-parental control studies

PURPOSE: To show the potential for gene-gene confounding bias in case-parental control studies. METHODS: The authors quantify the magnitude of gene-gene confounding bias using simple mathematical equations. They also demonstrate the potential problems of such a bias with hypothetical (but realistic) examples. RESULTS: The degree of bias resulting from gene-gene confounding was found to be quite substantial under certain conditions (two genes are very closely linked and/or the study was performed in a recently admixed population). CONCLUSION: In this post-genomic era more and more encounters of the gene-gene confounding will be expected, if the one-gene-at-a-time approach continues to be adopted.     

Intercomparison of retrospective radon detectors.

We performed both a laboratory and a field intercomparison of two novel glass-based retrospective radon detectors previously used in major radon case-control studies performed in Missouri and Iowa. The new detectors estimate retrospective residential radon exposure from the accumulation of a long-lived radon decay product, (210)Pb, in glass. The detectors use track registration material in direct contact with glass surfaces to measure the alpha-emission of a (210)Pb-decay product, (210)Po. The detector's track density generation rate (tracks per square centimeter per hour) is proportional to the surface alpha-activity. In the absence of other strong sources of alpha-emission in the glass, the implanted surface alpha-activity should be proportional to the accumulated (210)Po, and hence to the cumulative radon gas exposure. The goals of the intercomparison were to a) perform collocated measurements using two different glass-based retrospective radon detectors in a controlled laboratory environment to compare their relative response to implanted polonium in the absence of environmental variation, b) perform collocated measurements using two different retrospective radon progeny detectors in a variety of residential settings to compare their detection of glass-implanted polonium activities, and c) examine the correlation between track density rates and contemporary radon gas concentrations. The laboratory results suggested that the materials and methods used by the studies produced similar track densities in detectors exposed to the same implanted (210)Po activity. The field phase of the intercomparison found excellent agreement between the track density rates for the two types of retrospective detectors. The correlation between the track density rates and direct contemporary radon concentration measurements was relatively high, considering that no adjustments were performed to account for either the residential depositional environment or glass surface type. Preliminary comparisons of the models used to translate track rate densities to average long-term radon concentrations differ between the two studies. Further calibration of the retrospective detectors' models for interpretation of track rate density may allow the pooling of studies that use glass-based retrospective radon detectors to determine historic residential radon exposures.     

Exposure assessment of ultrafine particles in epidemiologic time-series studies

The health effects of fine-particulate air pollution (PM2.5, aerodynamic diameter <2.5 microm) observed in epidemiologic time-series studies may partly be due to the high number concentration of ultrafine particles (aerodynamic diameter <0.1 microm) in urban air. The key uncertainty is how well daily variations in the ultrafine particle concentration measured at a central site correlate with the variations in average personal exposure. Due to a lack of research data, this correlation has been estimated indirectly in this review on the basis of studies on the sources, ambient air levels, spatial variability, indoor air levels, and lung deposition of ultrafine particles. It is concluded that central site monitoring may give a somewhat worse proxy for human exposure to ultrafine particles than to PM2.5 in time-series studies.     

The usability of wood as a volume trap for the purpose of retrospective radon exposure assessment.

The possibility of using wood samples as a retrospective radon monitor was investigated. Retrospective radon monitors are based on the analysis of the enhanced 210Po content of the bulk of the sample as a result of radon exposure. Several wood samples in different sizes and shapes were tested for their radon penetrability and the speed at which radon can diffuse throughout the material. Also, the volume ratio of air to solid material and the natural occurring 210Po background were determined. It was seen that only in fairly exceptional cases wood can be used as an acceptable volume trap. This is mostly due to the high and variably occurring natural 210Po background in wood samples with respect to the expected increase in radon-related 210Po due to common radon concentrations in dwellings.     

Exposure assessment in industry specific retrospective occupational epidemiology studies.

Quantitative estimation of exposure for occupational epidemiology studies has received increasing attention in recent years and, as a result, a body of methodological literature has begun to take form. This paper reviews the generic issues in the methodology of exposure assessment, particularly methods for quantitative retrospective assessment studies. A simple framework, termed an exposure data matrix (EDM), for defining and analysing exposure data is proposed and discussed in terms of the definition of matrix dimensions and scales. Several methods for estimation, interpolation, and extrapolation, ranging from subjective ratings to quantitative statistical modelling are presented and discussed. The various approaches to exposure assessment based on the EDM concept are illustrated with studies of lung disease among coal miners and other dust and chemically induced chronic occupational diseases. The advantages of validated statistical models are emphasised. The importance of analysis and control of errors in exposure assessments, and integration of the exposure assessment and exposure-response processes, especially for emerging occupational health issues, is emphasised.     

Potential for bias in case-control studies of oral contraceptives and breast cancer

The discrepancies between the findings of the 6 large case-control studies to study the association between oral contraceptive (OC) use and breast cancer diagnosed in the 1980s may be due to chance or bias. The likelihood that chance played a role is suggested by the large numbers of subgroups examined in each study, the inconsistencies in the findings of different studies, and the wide confidence intervals around most of the relative risks. The most serious potential problem in case-control studies is that the procedures used to select cases and controls may produce groups that are not truly comparable. Recall bias is likely to contaminate information about the duration and type of part OC use. In addition, the more frequent examination of the breasts of women using OCs can produce surveillance bias. 6 procedures are recommended to minimize bias in future case-control studies of OC use and breast cancer: 1) whenever possible, cases and controls should be selected from an entire community; 2) if hospital controls need to be used, there should be explicit criteria for selecting them and the proportions of OC users in each diagnostic group should be presented; 3) women interviewed should not be aware of the study's hypotheses; 4) interviewers should be kept blind as to whether a subject is a case or control; 5) the possibility of recall bias should be investigated by comparing contraceptive histories from a sample of cases and controls with an independent source of information (preferably recorded before cancers were diagnosed); and 7) the interview should include questions about the frequency of breast examinations, so that any effects of more frequent surveillance of OC users can be controlled for.     

Problems found using a radon stripping algorithm for retrospective assessment of air filter samples

An evaluation of a large number of air sample filters was undertaken using a commercial alpha and beta spectroscopy system employing a passive implanted planar silicon (PIPS) detector. Samples were only measured after air flow through the filters had ceased. Use of a commercial radon stripping algorithm was implemented to discriminate anthropogenic alpha activity on the filters from the radon progeny. When uncontaminated air filters were evaluated, the results showed that there was a time-dependent bias in both average estimates and measurement dispersion of anthropogenic activity estimates with the relative bias being small compared to the dispersion, indicating that the system would not give false positive indications for an appropriately set decision level. By also measuring environmental air sample filters simultaneously with electroplated alpha sources, use of the radon stripping algorithm demonstrated a number of substantial unexpected deviations from calibrated values indicating that the system would give false negative indications. Use of the current algorithm is, therefore, not recommended for general assay applications. Use of the PIPS detector should only be utilized for gross counting without appropriate modifications to the curve-fitting algorithm. As a screening method, the radon stripping algorithm might be expected to see elevated alpha activities from pure Pu on air sample filters (not due to radon progeny) around the 200 disintegrations per minute level.     

Potential bias due to prevalent diseases in prospective studies

In prospective studies, subjects found to have the disease under investigation at the initial screening examination are commonly excluded from analyses. However, the possibility of bias due to prevalent conditions other than the disease of interest is usually not considered. In the present study, an algebraic development enables analysis of the effects of inclusion and exclusion of subjects with certain prevalent conditions upon risk estimates. Hypothetical data are presented for which an association between a risk factor and an incident disease could become null or even reversed after removing subjects with certain prevalent diseases. Bias appears even when the only association present is between risk factor and total disease incidence. Data from the Honolulu Heart Study also have been used to illustrate this finding, examining the association between coronary heart disease (CHD) incidence and smoking. Decisions regarding the inclusion or exclusion of subjects with prevalent diseases requires prior knowledge of alteration of usual risk factors levels by individuals with these diseases. Simply removing all subjects with prevalent diseases might on the contrary create bias. Therefore, people with prevalent diseases should be screened for potential alteration of their risk factor levels as a result of the diseases. The situation becomes still more complex when several risk factors and prevalent diseases need to be considered at the same time as it happens in multivariate analyses. Because this situation represents a bias, and not confounding or effect modification, controlling for the effect of prevalent diseases is not appropriate.     

职业性回顾队列研究中的接触估计

本文介绍了职业性回顾队列研究中进行接触估计的基本步骤以及针对不同资料特征采用的具体估计方法,为阐明剂量－反应（接触－反应）关系提供证据。