Proxy respondents and the validity of occupational and other exposure data. The Selected Cancers Cooperative Study Group

Our plain-English summary of Proxy respondents and the validity of occupational and other exposure data. The Selected Cancers Cooperative Study Group, by C.A. Boyle, E.A. Brann (American Journal of Epidemiology, 1992).

The study

• Authors: C.A. Boyle, E.A. Brann
• Journal: American Journal of Epidemiology
• Published: 1992;136(6):712-21
• Cited by: 63 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

As part of a multicenter cancer case-control study conducted in 1984-1988, a proxy interview was attempted for all cases who were initially interviewed for the study but who died during the 4-year data collection period. To assess the validity of using wives, other relatives, or other informants to obtain information about a subject, the authors compared occupational and other exposure data obtained from 270 male cancer cases and their proxy respondents. The primary focus of the case-control study was on Vietnam military service and exposure to phenoxy herbicides, but cases and their proxy respondents were also asked about occupational and other exposures relevant to the cancers. The accuracy of reporting for specific occupational exposures (e.g., asbestos and formaldehyde) and specific occupations (e.g., dry cleaning and meat packing or processing) was poor, although the latter improved somewhat when only case-spouse pairs were examined. Similarly, there was poor sensitivity in the reporting of herbicide exposure information in farming and other related occupations. In contrast, the reporting of certain demographic characteristics, childhood history characteristics, and use of alcohol and cigarettes was relatively good, and was even better when only case-spouse pairs were examined. The poor quality of proxy information for detailed exposure information suggests the need for careful use and interpretation of proxy information in epidemiologic studies.

In plain English

This was a methods study within a larger U.S. case-control study of six cancers in men, run in 1984-1988. When a man who had been interviewed died during the study, staff tried to interview a proxy, using the same questions reworded to refer to him, minus a few sensitive ones. They tried his wife first, and another knowledgeable person if she could not take part. The man's own answers were treated as correct. Proxy interviews were completed for 270 of the 310 men who died. Proxies reported general smoking history well: for ever having been a regular smoker, sensitivity and specificity were both 95 percent. Reporting of specific workplace exposures, including asbestos, was poor. Of 44 men who said they had worked with or around asbestos, proxies identified 25 percent. Wives were not better than other proxies at reporting workplace hazards. The authors concluded that the poor quality of proxy information on detailed exposures called for careful use and interpretation of such information. They noted that the men's own reports, used as the standard, may also contain errors.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Methodological (validation) substudy within a multicenter case-control study. Each case's own interview answers were compared with answers given by a proxy respondent after the case died, using sensitivity, specificity, positive predictive value and prevalence.
• Who was studied: 270 male cancer cases and their proxy respondents. These were the completed proxy interviews among 310 cases who died during the study period. The men were born 1929-1953 and first diagnosed between December 1984 and November 1988 with non-Hodgkin's lymphoma, Hodgkin's disease, soft tissue sarcoma, or liver, nasal or nasopharyngeal cancer, in eight population-based U.S. cancer registry areas. All had been interviewed themselves before they died. Staff tried the wife first; the proxies were 163 wives, 28 siblings, 29 parents, 6 children and 44 others. The proxy questions were the same as the case's, modified to refer to the case, and a few sensitive questions (such as illicit drug use and sex partners) were deleted.
• Exposure: Not an asbestos-disease study. The questionnaire's main focus was Vietnam military service and phenoxy herbicides. It also asked about 8 industries and 18 occupational exposures, including asbestos, insulation materials, formaldehyde and chemical solvents, as well as demographic, childhood, smoking and alcohol questions. No asbestos fiber types were specified.
• Main results:
  • Proxy reporting of specific occupational exposures and specific occupations was poor. With the exception of plywood and wood dust or sawdust, sensitivities were 40 percent or lower.
  • Asbestos (worked with or around): sensitivity was 25 percent among the 44 cases who reported asbestos exposure, and specificity was 95 percent among the 226 who did not. Positive predictive value was 48 percent. Asbestos exposure was reported by 16 percent of cases but by 9 percent of proxies.
  • Sensitivity was generally somewhat better when the case had held the exposure-related job for 5 or more years, but it stayed low. For asbestos, sensitivity was 17 percent for jobs under 5 years and 33 percent for longer jobs.
  • General smoking status was reported well by proxies: sensitivity and specificity were both 95 percent for ever having been a regular cigarette smoker. Ever having been a regular drinker had 78 percent sensitivity and 73 percent specificity. Details of quantity and duration were reported less accurately.
  • Wives were somewhat more accurate than other proxies about the jobs their husbands held. They were not better at reporting whether their husbands had worked with or around hazards such as solvents, insulation, or wood dust/sawdust.
  • In a worked example assuming a true risk ratio of 2.0, 250 cases and 500 controls, the proxy misclassification levels found here would give observed risk ratios of 1.8 (never married), 1.6 (less than 12 years of education), 1.4 (ever sprayed or mixed herbicides), 1.4 (more than 30 pack-years of smoking) and 1.2 (ever worked with cutting oils).
• The authors' conclusion: The authors concluded that proxy information on detailed exposures was of poor quality. They stated that proxy information needs careful use and interpretation in epidemiologic studies, and that occupational exposure data from proxies were greatly misreported.
• Limitations the authors state:
  • The case's own report was treated as the gold standard, although self-reports also contain errors.
  • Some disagreement may reflect poor reliability, especially for questions about the details of exposure.
  • Only proxies of cases who died during the study were interviewed, and the effect of the death on proxy accuracy is unclear.
  • Only cases were included, so the accuracy of proxies for control subjects could not be examined.

Topics

Medical terms in this study, explained in our Lexicon: Case-Control Study, Occupational Exposure.

Related studies

• Baseline predictors of negative and incomplete pleural cytology in patients with suspected pleural malignancy - Data supporting 'Direct to LAT' in selected groups — S. Tsim, S. Paterson, et al. (2019)
• INTEROCC case-control study: lack of association between glioma tumors and occupational exposure to selected combustion products, dusts and other chemical agents — A. Lacourt, E. Cardis, et al. (2013)
• The INTEROCC case-control study: risk of meningioma and occupational exposure to selected combustion products, dusts and other chemical agents — D.M. McElvenny, M. van Tongeren, et al. (2018)
• Design and conduct of occupational epidemiology studies: IV. The analysis of case-control data — N. Pearce, H. Checkoway, et al. (1989)
• Reliability of Family Proxy Data for Studies of Malignant Mesothelioma: Results from the ATSDR Pilot Surveillance — N. Melnikova, J. Wu, et al. (2013)
• Paclitaxel for malignant pleural mesothelioma: a phase II study of the EORTC Lung Cancer Cooperative Group — J. van Meerbeeck, C. Debruyne, et al. (1996)

Browse more research in American Journal of Epidemiology, or return to medical research.