Biomechanical allometry in hominoid thoracic vertebrae.

Considerable differences in spinal morphology have been noted between humans and other hominoids. Although comparative analyses of the external morphology of vertebrae have been performed, much less is known regarding variations in internal morphology (density) and biomechanical performance among humans and closely related non-human primates. In the current study we utilize density calibrated computed tomography images of thoracic vertebral bodies from hominoids (n=8-15 per species, human specimens 20-40 years of age) to obtain estimates of vertebral bone strength in axial compression and anteroposterior bending and to determine how estimates of strength scale with animal body mass. Our biomechanical analysis suggests that the strength of thoracic vertebral bodies is related to body mass (M) through power law relationships (y proportional, variant M(b)) in which the exponent b is 0.89 (reduced major axis) for prediction of axial compressive strength and is equal to 1.89 (reduced major axis) for prediction of bending strength. No differences in the relationship between body mass and strength were observed among hominoids. However, thoracic vertebrae from humans were found to be disproportionately larger in terms of vertebral length (distance between cranial and caudal endplates) and overall vertebral body volume (p<0.05). Additionally, vertebral bodies from humans were significantly less dense than in other hominoids (p<0.05). We suggest that reduced density in human vertebral bodies is a result of a systemic increase in porosity of cancellous bone in humans, while increased vertebral body volume and length are a result of functional adaptation during growth resulting in a vertebral bone structure that is just as strong, relative to body mass, as in other hominoids.

Estimates of historical exposures by phase contrast and transmission electron microscopy in North Carolina USA asbestos textile plants.

OBJECTIVES: To develop a job-exposure matrix (JEM) for fibre exposures in three asbestos textile plants and to develop estimates of fibre size-specific exposures. METHODS: Historical dust samples from three North Carolina, USA asbestos textile plants were obtained. Plant specific samples were used to express impinger dust concentrations as fibre concentrations by phase contract microscopy (PCM). Mixed models were used to estimate PCM exposures by plant, department, job and calendar time. Archived membrane filter samples were analysed by transmission electron microscopy (TEM) to determine the bivariate diameter/length distribution of airborne fibres by plant and operation. RESULTS: PCM fibre levels estimated from the models were very high in the 1930s, with some operations having in excess of 200 fibres/ml, and decreased appreciably over time. TEM results for 77 airborne dust samples found that only a small proportion of airborne fibres were measured by PCM (>0.25 microm in diameter and >5 microm in length) and the proportion varied considerably by plant and operation (range 2.9% to 10.0%). The bivariate diameter/length distribution of airborne fibres demonstrated a relatively high degree of variability by plant and operation. PCM adjustment factors also varied substantially across plants and operations. CONCLUSIONS: These data provide new information concerning airborne fibre levels and characteristics in three historically important asbestos textile plants. PCM concentrations were high in the early years and TEM data demonstrate that the vast majority of airborne fibres inhaled by the workers were shorter than 5 microm in length, and thus not included in the PCM-based fibre counts.

Lung cancer mortality and fibre exposures among North Carolina asbestos textile workers.

OBJECTIVE: To describe mortality among workers exposed to chrysotile asbestos and evaluate the relationship between lung cancer and asbestos fibre exposure. METHODS: Workers employed for at least 1 day between 1 January 1950 and 31 December 1973 in any of four plants in North Carolina, USA that produced asbestos textile products were enumerated. Vital status was ascertained through 31 December 2003. Historical exposures to asbestos fibres were estimated from work histories and 3578 industrial hygiene measurements taken in 1935-1986. Mortality of the cohort was compared with that of the national population via standardised mortality ratios (SMRs). Exposure-response relationships for lung cancer were examined within the cohort using Poisson regression to compute adjusted mortality rate ratios. RESULTS: Follow-up of 5770 workers included in the cohort resulted in 181 640 person-years of observation, with 2583 deaths from all causes and 277 from lung cancer. Mortality from all causes, all cancers and lung cancer was significant higher than expected, with SMRs of 1.47 for all causes, 1.41 for all cancer and 1.96 (95% CI 1.73 to 2.20) for lung cancer. SMRs for pleural cancer, mesothelioma and pneumoconiosis were also elevated. The risk of lung cancer and asbestosis increased with cumulative fibre exposure (RR 1.102 per 100 fibre-year/ml, 95% CI 1.044 to 1.164, and RR 1.249 per 100 fibre-year/ml, 95% CI 1.186 to 1.316, respectively, for total career exposure). CONCLUSIONS: This study provides further evidence that exposure to chrysotile asbestos in textile manufacturing is associated with increased risk of lung cancer, asbestosis cancer of the pleura and mesothelioma.

Occupational injuries among aides and nurses in acute care.

BACKGROUND: Occupational injuries are common among nursing personnel. Most epidemiologic research on nursing aides comes from long-term care settings. Reports from acute care settings often combine data on nurses and aides even though their job requirements and personal characteristics are quite different. Our objective was to assess risk of work-related injuries in an acute care setting while contrasting injuries of aides and nurses. METHODS: A retrospective cohort of aides (n = 1,689) and nurses (n = 5,082) working in acute care at a large healthcare system between 1997 and 2004 were identified via personnel records. Workers' compensation filings were used to ascertain occupational injuries. Poisson regression was used to estimate rate ratios (RR) and 95% confidence intervals (95% CI). RESULTS: Aides had higher overall injury rates than nurses for no-lost work time (RR = 1.2, 95% CI: 1.1-1.3) and lost work time (RR = 2.8, 95% CI: 2.1-3.8) injuries. The risk of an injury due to lifting was greater among aides compared to nurses for both non-lost work time and lost work time injuries. Injury rates among aides were particularly high in rehabilitation and orthopedics units. Most of the injuries requiring time away from work for both groups were related to the process of delivering direct patient care. CONCLUSIONS: Our findings illustrate the importance of evaluating work-related injuries separately for aides and nurses, given differences in injury risk profiles and injury outcomes. It is particularly important that occupational safety needs of aides be addressed as this occupation experiences significant job growth.

A case crossover study of triggers for hand injuries in commercial fishing.

OBJECTIVES: Previous studies estimate hand and wrist injuries are common in commercial fishing. Risk factors including working with catch, handling gear and slips or falls, have been identified from activity and injury contact reports, but no studies have examined the influence of transient risk factors, or triggers. This case crossover study design was conducted to investigate triggers for acute hand trauma in commercial fishing. METHODS: A case crossover study was nested within a previously established prospective cohort of 217 southeastern United States commercial fishermen followed from April 1999 through October 2001. Hand injury cases and controls were matched using three control selection strategies. Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated to determine if transient risk factors such as glove use, engaging in more than one type of fishing, maintenance activities and other covariates of interest increased the risk of occupational traumatic hand/wrist/digit injuries. RESULTS: 21% (46/217) of fishermen reported one or more hand/wrist/digit injuries, yielding 65 eligible cases. Performing maintenance work (any vs none) (OR 3.1, 95% CI 1.8 to 5.5) and using multiple types of fishing equipment in comparison to using only one type (OR 1.9, 95% CI 0.9 to 3.8) were associated with increased risk of hand/wrist/digit injury. There was no evidence glove use was protective (any vs none) for hand/wrist/digit injury (OR 0.9, 95% CI 0.5 to 1.7). CONCLUSIONS: The case crossover design is a useful method to determine triggers of commercial fishing-related hand/wrist/digit injuries. Maintenance work was strongly associated with hand/wrist/digit injury for these fishermen.

Development of a fibre size-specific job-exposure matrix for airborne asbestos fibres.

OBJECTIVE: To develop a method for estimating fibre size-specific exposures to airborne asbestos dust for use in epidemiological investigations of exposure-response relations. METHODS: Archived membrane filter samples collected at a Charleston, South Carolina asbestos textile plant during 1964-8 were analysed by transmission electron microscopy (TEM) to determine the bivariate diameter/length distribution of airborne fibres by plant operation. The protocol used for these analyses was based on the direct transfer method published by the International Standards Organization (ISO), modified to enhance fibre size determinations, especially for long fibres. Procedures to adjust standard phase contrast microscopy (PCM) fibre concentration measures using the TEM data in a job-exposure matrix (JEM) were developed in order to estimate fibre size-specific exposures. RESULTS: A total of 84 airborne dust samples were used to measure diameter and length for over 18,000 fibres or fibre bundles. Consistent with previous studies, a small proportion of airborne fibres were longer than >5 microm in length, but the proportion varied considerably by plant operation (range 6.9% to 20.8%). The bivariate diameter/length distribution of airborne fibres was expressed as the proportion of fibres in 20 size-specific cells and this distribution demonstrated a relatively high degree of variability by plant operation. PCM adjustment factors also varied substantially across plant operations. CONCLUSIONS: These data provide new information concerning the airborne fibre characteristics for a previously studied textile facility. The TEM data demonstrate that the vast majority of airborne fibres inhaled by the workers were shorter than 5 mum in length, and thus not included in the PCM-based fibre counts. The TEM data were used to develop a new fibre size-specific JEM for use in an updated cohort mortality study to investigate the role of fibre dimension in the development of asbestos-related lung diseases.

Breast cancer mortality among female electrical workers in the United States.

BACKGROUND: Previous epidemiologic studies have suggested that exposure to electric or magnetic fields in occupational and residential environments may cause cancer. Recent experimental findings provide some support for the hypothesis that exposure to extremely low-frequency electromagnetic fields reduces the pineal gland's nocturnal production of the hormone melatonin, thereby increasing susceptibility to sex hormone-related cancers such as breast cancer. PURPOSE: Our purpose was to assess the evidence that cancer of the female breast might be associated with exposure to extremely low-frequency electromagnetic fields. METHODS: Records of women who had breast cancer as the underlying cause of their death (ICD-9 174) and control subjects (four per case) were selected from computer files of U.S. mortality data for the years 1985-1989. Women 20 years and older at the time of their death were eligible for inclusion if they were residents of and died in one of the 24 states that provided death certification records with occupation and industry codes to the National Center for Health Statistics for at least 1 year during the study interval. Data from death certificates were used to classify the case and control subjects with regard to potential occupational exposure to electric and magnetic fields. Control subjects were a random sample of women who died of any other underlying cause, excluding leukemia and brain cancer. RESULTS: The data analysis contrasted 68 women with breast cancer and 199 controls, all with electrical occupations, with 27,814 women with breast cancer and 110,750 controls, all of whom had other occupations. Electrical workers had excess mortality from breast cancer relative to other employed women [odds ratio (OR) = 1.38; 95% confidence interval (CI) = 1.04-1.82]). Adjusted ORs for specific electrical occupations were 1.73 (95% CI = 0.92-3.25) for electrical engineers, 1.28 (95% CI = 0.79-2.07) for electrical technicians, and 2.17 (95% CI = 1.17-4.02) for telephone installers, repairers, and line workers. There was no excess of breast cancer, however, in seven other occupations held more frequently by women and also involving potentially elevated electrical exposures, including telephone operators, data keyers, and computer operators and programmers. CONCLUSIONS: In light of the limitations inherent in death certification data and the design of this study, any conclusions regarding the hypothesis that exposure to extremely low-frequency electromagnetic fields causes breast cancer among women must be limited. Nevertheless, our findings are broadly consistent with that hypothesis and encourage further investigation with improvements in study design and data quality.

Poisson regression analysis of ungrouped data.

BACKGROUND: Poisson regression is routinely used for analysis of epidemiological data from studies of large occupational cohorts. It is typically implemented as a grouped method of data analysis in which all exposure and covariate information is categorised and person-time and events are tabulated. AIMS: To describe an alternative approach to Poisson regression analysis using single units of person-time without grouping. METHODS: Data for simulated and empirical cohorts were analysed by Poisson regression. In analyses of simulated data, effect estimates derived via Poisson regression without grouping were compared to those obtained under proportional hazards regression. Analyses of empirical data for a cohort of 138 900 electrical workers were used to illustrate how the ungrouped approach may be applied in analyses of actual occupational cohorts. RESULTS: Using simulated data, Poisson regression analyses of ungrouped person-time data yield results equivalent to those obtained via proportional hazards regression: the results of both methods gave unbiased estimates of the "true" association specified for the simulation. Analyses of empirical data confirm that grouped and ungrouped analyses provide identical results when the same models are specified. However, bias may arise when exposure-response trends are estimated via Poisson regression analyses in which exposure scores, such as category means or midpoints, are assigned to grouped data. CONCLUSIONS: Poisson regression analysis of ungrouped person-time data is a useful tool that can avoid bias associated with categorising exposure data and assigning exposure scores, and facilitate direct assessment of the consequences of exposure categorisation and score assignment on regression results.

Incidence of allergy and allergy symptoms among workers exposed to laboratory animals.

BACKGROUND AND AIMS: Few studies have described relations between exposure to laboratory animals and the incidence of laboratory animal allergy (LAA). Studies that have found exposure-response relations have been cross sectional in design or have focused on exposure to rats and mice. This study used longitudinal data collected over a 12 year period to describe the relations between indices of exposure to laboratory animals and the development of LAA and LAA symptoms. METHODS: Data were obtained from questionnaires and serological laboratory results from a dynamic cohort of workers exposed to a variety of laboratory animals in a pharmaceutical manufacturing company. Poisson regression was used to model the incidence rate ratios of species specific and general LAA and LAA symptoms at different levels of exposure. RESULTS: The 12 year incidence rates of LAA symptoms and LAA for all workers were 2.26 (95% CI 1.61 to 2.91) and 1.32 (95% CI 0.76 to 1.87) per 100 person-years, respectively. Higher rate ratios were seen with increasing reported hours of exposure to tasks that required working with animal cages or with many animals at one time. The most common symptoms were related to rhinitis rather than to asthma. CONCLUSIONS: This study suggests that the risk of LAA increases with duration of exposure to animals and work in animal related tasks. Incidence might be reduced by limiting hours per week of exposure to laboratory animals.

Population-based case-control study of occupational exposure to electromagnetic fields and breast cancer.

PURPOSE: This population-based case-control study examined occupational exposure to electromagnetic fields in relation to female breast cancer incidence among 843 breast cancer cases and 773 controls. METHODS: Exposure was classified based on work in the two longest-held jobs, and indices of cumulative exposure to magnetic fields based on a measurement survey. RESULTS: Female breast cancer was not associated with employment as an office or industrial worker. For the total study population, cumulative exposure over the entire career, and in the past 0-10 and 10-20 years generally showed odds ratios (ORs) close to the null. Moderately elevated risks were found for intermediate but not high levels of cumulative exposure accumulated 20 or more years ago (OR = 1.5; 95% CI = 1.1-2.0). Associations were stronger for premenopausal women (OR = 1.7; 95% CI = 1.1-2.7) in the past 10-20 years, and those with estrogen-receptor positive (ER+) breast tumors (OR = 2.06; 95% CI = 1.1-4.0). No consistent dose-response patterns were observed. CONCLUSIONS: These findings give little support to the hypothesis that electromagnetic fields cause cancer of the female breast.

Mortality and ambient fine particles in southwest Mexico City, 1993-1995.

++Epidemiologic studies have focused attention on the health effects of fine particulate air pollutants <2.5 microm in diameter (PM2.5). To further characterize the potential effects of fine particles, we investigated the relationship of air pollution to mortality in Mexico City during 1993-1995. The concentration of PM2.5 was measured on a 24-hr integrated basis; concentrations of NO2 and ozone were measured hourly and reduced to 24-hr means. Daily mortality was determined from death registration records, and Poisson regression was used to model daily death counts as a function of air pollutant levels on the same and previous days, while controlling for temperature and periodic cycles. Without taking other air pollutants into account, a 10 microg/m3 increase in the level of PM2.5 was associated with a 1.4% increase in total mortality, both on the current day and 4 days after exposure [95% confidence interval (CI), 0.2-2.5]. An equivalent increase in PM2.5 was also associated with somewhat larger excesses of deaths among people over 65 years of age and from cardiovascular and respiratory causes, which occurred after a lag of 4 days. The mean concentration of ozone over a 2-day period was associated with a 1.8% increase in mortality from cardiovascular diseases. NO2 was not consistently related to mortality. Fine particles had an independent effect on mortality when modeled simultaneously with other pollutants, and the association of ozone with cardiovascular mortality was strengthened after adjusting for NO2 and PM2.5. These results support previous findings that urban air pollution at current levels leads to excess mortality and suggest that fine particles may play a causal role in producing that excess.

Association between air pollution and intrauterine mortality in São Paulo, Brazil.

The associations among daily counts of intrauterine mortality and pollutant concentrations (NO2, SO2, CO, O3, and particulate matter (3/4)10 microm) were investigated for the period ranging from January 1991 to December 1992 in the city of São Paulo, Brazil. We used Poisson regression techniques, adjusted for season and weather. The association between intrauterine mortality and air pollution was strong for NO2 (coefficient = 0.0013/ microg/m3; p<0.01) but lesser for SO2 (coefficient = 0.0005/ microg/m3; p<0.10) and CO (coefficient = 0.0223/ppm; p<0.10). A significant association was observed when an index that combined these three pollutants was considered in the models instead of considering each pollutant individually (p<0.01). These associations exhibited a short time lag, not over 5 days. In addition, some evidence of fetal exposure to air pollution was obtained by disclosing a significant association between the levels of carboxyhemoglobin of blood sampled from the umbilical cord and ambient CO levels in children delivered by nonsmoking pregnant women in the period from May to July 1995. Our results suggest that air pollution in São Paulo may promote adverse health effects on fetuses.

Informal jobs: another occupational hazard for women's mental health?

OBJECTIVES: The hypothesis that informal jobs, which imply absence of formal labour contracts, instability and the absence of fringe benefits, are positively associated with psychiatric symptoms was evaluated in a poor urban area of Brazil. METHODS: With data from a cross-sectional study, the association between informal jobs and high number of psychological symptoms was estimated. The study population was composed of 327 women randomly selected from a community in the city of Salvador, Brazil. Women who reported having a job without a formal contract were classified as informal workers. Psychological symptoms were collected through a validated questionnaire, the QMPA. RESULTS: A positive association between informal work and a high number of psychological symptoms was found (crude prevalence ratio = 1.88, 95% confidence interval [CI]: 1.24-2.85). More than 4 hours of housework a day and being a family head were confounders, although adjustment for these variables did not significantly change the results (adjusted prevalence ratio = 1.97, 95% CI: 1.26-3.09). CONCLUSIONS: These findings are suggestive that informal work may be a risk for mental symptoms. Reinforcement of universal labour rights coverage and improvement in housework sharing are recommended.

Uncertainty in the relation between exposure to magnetic fields and brain cancer due to assessment and assignment of exposure and analytical methods in dose-response modeling.

Incomplete scientific knowledge ensures that, in every study, uncertainty will enter the processes of exposure estimation and exposure-response modeling. In the light of the heated debate about the health effects of magnetic fields resulting from power production and usage, we undertook a sensitivity analysis to evaluate uncertainty related to key decisions in a previous study of brain cancer and occupational exposure to magnetic fields. The findings appeared to be relatively insensitive to most variations in the methods of exposure assessment, exposure assignment, and data analysis. The results can be visualized by defining bands of uncertainty about a best-bet estimate of the association based on our original study. These bands of methodological uncertainties were similar in magnitude to the conventional 95% confidence interval, but they provide a measure of the potential range of systematic bias in the results, rather than reflecting statistical variability alone. The methodology employed here can be applied to other studies, and other researchers are encouraged to conduct sensitivity analysis in order to estimate methodological uncertainty as an alternative to statistical confidence intervals.

Is social class standardisation appropriate in occupational studies?

Social class standardisation has been proposed as a method for separating the effects of occupation and "social" or "lifestyle" factors in epidemiological studies, by comparing workers in a particular occupation with other workers in the same social class. The validity of this method rests upon two assumptions: (1) that social factors have the same effect in all occupational groups in the same social class, and (2) that other workers in the same social class as the workers being studied are free of occupational risk factors for the disease of interest. These assumptions will not always be satisfied. In particular, the effect of occupation will be underestimated when the comparison group also has job-related exposures which cause the disease under study. Thus, although adjustment for social class may minimise bias due to social factors, it may introduce bias due to unmeasured occupational factors. This difficulty may be magnified when occupational category is used as the measure of social class. Because of this potential bias, adjustment for social class should be done only after careful consideration of the exposures and disease involved and should be based on an appropriate definition of social class. Both crude and standardised results should be presented when such adjustments are made.

Use of protective equipment in a cohort of rugby players.

PURPOSE: To describe the level of usage of protective devices and equipment in a cohort of New Zealand rugby players. METHODS: Male and female players (N = 327) from a range of competitive grades were followed over the course of the season. Participants were interviewed weekly about their participation in rugby and use of protective equipment. The main outcome measure was percentage of all player-weeks of follow-up for which each equipment item was used. RESULTS: Mouthguards, the most commonly used equipment item, were worn for 64.9% of player-weeks. Mouthguard usage ranged from 55.0% of player-weeks in Schoolgirls grade to 72.9% of player-weeks in Senior A competition. The next most common item was taping of body joints (23.7% of player-weeks). The sites most commonly taped were the ankle, knee, and hand. Overall usage for the other protective equipment items studied (shin guards, padded headgear, head tape, support sleeves, and grease) was below 15%. In general, forwards had higher usage of protective equipment than backs, and male players had higher usage than female players. The most common self-reported reasons for using protective equipment were to prevent injury and because of a past injury. Players exhibited considerable week-to-week variation in their usage of protective equipment. CONCLUSIONS: In general, equipment usage was highest in those at greatest risk of injury, namely, forwards, male players, and the senior grades. The high voluntary use of mouthguards is encouraging and indicative of a base of player support for their role in this sport.

The impact of exposure categorisation for grouped analyses of cohort data.

BACKGROUND: Poisson regression is routinely used in occupational and environmental epidemiology. For typical Poisson regression analyses, person-time and events are tabulated by categorising predictor variables that were originally measured on a continuous scale. In order to estimate a dose-response trend, a researcher must decide how to categorise exposures and how to assign scores to exposure groups. AIMS: To investigate the impact on regression results of decisions about exposure categorisation and score assignment. METHODS: Cohort data were generated by Monte Carlo simulation methods. Exposure categories were defined by quintiles or deciles of the exposure distribution. Scores were assigned to exposure groups based on category midpoint and mean exposure levels. Estimated exposure-disease trends derived via Poisson regression were compared to the "true" association specified for the simulation. RESULTS: Under the assumption that exposures conform to a lognormal or exponential distribution, trend estimates tend to be negatively biased when scores are assigned based on category midpoints and positively biased when scores are assigned based on cell specific mean values. The degree of bias was greater when exposure categories were defined by quintiles of the exposure distribution than when categories were defined by deciles of the exposure distribution. CONCLUSIONS: The routine practice of exposure categorisation and score assignment introduces exposure misclassification that may be differential with respect to disease status and, consequently, lead to biased exposure-disease trend estimates. When using the Poisson regression method to evaluate exposure-disease trends, such problems can be minimised (but not necessarily eliminated) by forming relatively refined exposure categories based on percentiles of the exposure distribution among cases, and by assigning scores to exposure categories that reflect person-time weighted mean exposure levels.

Deindustrialisation and the long term decline in fatal occupational injuries.

AIMS: To examine the extent to which deindustrialisation accounts for long term trends in occupational injury risk in the United States. METHODS: Rates of fatal unintentional occupational injury were computed using data from death certificates and the population census. Trends were estimated using Poisson regression. Standardisation and regression methods were used to adjust for the potential effect of structural change in the labour market. RESULTS: The fatal occupational injury rate for all industries declined 45% from 1980 to 1996 (RR (rate ratio) 0.55, 95% CI 0.52 to 0.57). Adjustment for structural changes in the workforce shifted the RR to 0.62 (95% CI 0.60 to 0.65). Expanding industries enjoyed more rapid reduction in risk (-3.43% per year, 95% CI -3.62 to -3.24) than those that contracted (-2.65% per year, 95% CI -2.88 to -2.42). CONCLUSIONS: Deindustrialisation contributed to the decline of fatal occupational injury rates in the United States, but explained only 10-15% of the total change.

Estimation of particulate matter from visibility in Bangkok, Thailand.

Lack of daily data on airborne particles has been a common problem in an air pollution research. To deal with this problem, a regression model was developed to estimate daily PM10 concentration using visibility in Bangkok from 1992 to 1997, based on 1092 visibility/PM10 pair-observations on low humidity days (humidity < or = 76.5%). Visibility was significantly and inversely associated with PM10 (r = 0.71), after adjusting for minimum temperature and winter indicator variable. The R2 of the model was 0.51.

Effect of incomplete exposure assessment on epidemiologic dose-response analyses.

OBJECTIVES: When potentially hazardous agents have multiple environmental sources, failure to include all exposure sources can constitute a type of measurement error. In addition, the effects of exposure from one source can also be confounded by exposure to other sources of the same agent. In this study clarification of these concepts is sought, and the direction and magnitude of the resulting bias in epidemiologic measures of association are examined. METHODS: The bias in dose-response functions when the exposure data omit some sources of the agent was estimated with linear and log-linear models to compute risk differences and risk ratios under different assumptions about the magnitude and correlation of exposures from measured and unmeasured sources. RESULTS: With unmeasured exposure of constant magnitude, there is no bias when a measure of association of the appropriate form (difference measures for additive dose-response processes, ratios for multiplicative ones) is selected. When the magnitude of unmeasured exposure varies, the results is nondifferential measurement error that can bias observed dose-response relations upward or downward, depending on the pattern of measurement error and the measure of association. CONCLUSIONS: Failure to measure all sources of exposure to an agent and account for them in the analysis can bias the results of epidemiologic studies. When it is not feasible to measure all exposure sources, the magnitude of bias can be predicted by estimating the distribution of omitted exposures from external data or substudies. Sensitivity analyses are particularly useful for estimating the direction and magnitude of potential bias from incomplete exposure assessment.