PHLT 8074 Week 11 Exposure Metric Defense Paper Example

Defending Lagged Cumulative Herbicide Handling Days Against the Charge That Cases Remember More: Four Lines of Evidence From a Composite Valley Case-Control Study of Parkinson's Disease

Student Name

Walden University

PHLT 8074: Environmental and Occupational Epidemiology

Instructor Name

Month Day, Year

Defending Lagged Cumulative Herbicide Handling Days Against the Charge That Cases Remember More: Four Lines of Evidence From a Composite Valley Case-Control Study of Parkinson's Disease

The Metric and the Objection

The composite valley study's primary exposure metric is lagged cumulative weighted herbicide handling days: days on which a participant personally mixed, loaded or applied the herbicide, weighted by task and protective equipment, summed over the years more than a decade before the reference date and counted through a structured calendar interview checked against a crop-exposure matrix. In the composite results, the odds of Parkinson's disease rose across tertiles of this metric, with an odds ratio of 1.9 in the highest tertile compared with participants who never handled the herbicide. All results in this paper are composite.

The likeliest critic will say the association is an artifact of recall. People diagnosed with Parkinson's disease search their past for causes, have often heard that pesticides are suspected and may therefore report more handling than controls with the same history. Recall bias of this kind is differential misclassification, and it can create an association where none exists (Coughlin, 1990). The objection is serious, and this paper sets it out in full before answering it with four lines of evidence.

The Objection at Full Strength

Three features of the study make recall bias plausible. Cases were interviewed after diagnosis, often after reading about environmental causes. The interview asked about handling decades earlier, where memory is weakest and most open to reconstruction. And the herbicide's suspected link to Parkinson's disease has been reported in the news, so cases may have a specific reason to remember it. If cases reported, on average, even 20% more handling days than controls with identical histories, the observed odds ratio could be substantially inflated.

Line 1: The Validation Substudy Found Similar Agreement in Cases and Controls

The substudy compared interview reports with farm spray logs, crew records and coworker interviews for 60 cases and 60 controls. Agreement between interview-reported and record-based handling days was similar in both groups: in the composite results, the intraclass correlation was 0.71 among cases and 0.68 among controls, and the ratio of interview to record days averaged 1.06 among cases and 1.04 among controls. If cases systematically overreported, their ratio would be noticeably higher. It was not. Spiegelman (2010) describes how such validation data can be used to correct exposure estimates; applying a correction for the small observed difference changed the highest-tertile odds ratio from 1.9 to 1.8.

Line 2: The Recall-Free Matrix Gave a Similar Result

The crop-exposure matrix assigns exposure from job title, crop and period, values that cannot be affected by how participants remember handling days. Using the matrix estimate in place of the interview, the highest-tertile odds ratio was 1.7. Because the matrix carries mostly Berkson-type error, which tends to reduce precision rather than inflate associations under common conditions (Armstrong, 1998), an association of similar size from the matrix argues against the interview result being a product of recall.

Line 3: The Association Held at Longer Lags and Weakened at None

If recall drove the result, the association should be strongest for recent handling, which is easiest to remember. The opposite occurred. With no lag, the highest-tertile odds ratio was 1.4; with a ten-year lag it was 1.9; with a fifteen-year lag it was 1.8. The pattern fits a causal exposure in the distant past, combined with reduced handling by cases during the prodromal years, and it does not fit a pattern driven by recall of recent events.

Line 4: A Comparison Exposure Showed No Association

The interview asked about handling a fungicide used on the same crops and in the same way, but not suspected of any link to Parkinson's disease. If cases simply reported more farm chemical work of every kind, they would report more fungicide handling as well. In the composite results, the highest-tertile odds ratio for fungicide handling days, adjusted for herbicide handling, was 1.05. Cases were not reporting more of everything; they reported more of the one product under suspicion.

Why These Four Lines and Not Others

Other checks were possible, and several were considered. Asking participants directly whether they had heard of a link between pesticides and Parkinson's disease would identify those most likely to overreport, but asking the question could itself prompt recall, so it was placed at the end of the interview and used only descriptively. Comparing reports from participants with those from proxies was set aside because proxies were few. The four lines chosen share one feature: each could have contradicted the main result. A defense built only from checks that cannot fail would persuade no one.

What the Defense Does and Does Not Establish

Taken together, the four lines make recall bias an unlikely explanation for the association. They do not prove causation. Other threats remain: confounding by unmeasured factors that travel with herbicide handling, such as well water or other pesticides; selection bias if controls who handled herbicides were less likely to participate; and residual error in the metric. Each is addressed elsewhere in the dissertation. Checkoway et al. (2004) note that occupational studies gain credibility by showing consistency across methods of exposure assessment, and the agreement among interview, matrix and validation data is the strongest feature of this study.

Conclusion

The study keeps its lagged, weighted count of handling days as the main measure. Its critic's objection is reasonable, and the design anticipated it: a validation substudy split by case status, a recall-free matrix, a lag structure that would expose recall of recent events and a comparison exposure that would reveal general overreporting. All four point the same way. If future work in other valleys, with other records and other validation samples, finds the same pattern, the case for a causal role will strengthen; if it does not, the explanation will need to be sought in what was particular to this study. Either outcome would be informative, and the study is designed so that either can be recognized. That is the standard it asks to be judged by.

References

Armstrong, B. G. (1998). Effect of measurement error on epidemiological studies of environmental and occupational exposures. Occupational and Environmental Medicine, 55(10), 651-656. https://doi.org/10.1136/oem.55.10.651

Checkoway, H., Pearce, N., & Kriebel, D. (2004). Research methods in occupational epidemiology (2nd ed.). Oxford University Press.

Coughlin, S. S. (1990). Recall bias in epidemiologic studies. Journal of Clinical Epidemiology, 43(1), 87-91. https://doi.org/10.1016/0895-4356(90)90060-3

Spiegelman, D. (2010). Approaches to uncertainty in exposure assessment in environmental epidemiology. Annual Review of Public Health, 31, 149-163. https://doi.org/10.1146/annurev.publhealth.012809.103720