PHLT 8074 Week 4 Exposure Assessment Critique Example

What Could "Ever Lived or Worked on a Farm" Stand For? A Five-Question Critique of the Exposure Measure in an Earlier Composite Parkinson's Disease Study, and Whether Its Error Differed by Case Status

Student Name

Walden University

PHLT 8074: Environmental and Occupational Epidemiology

Instructor Name

Month Day, Year

What Could "Ever Lived or Worked on a Farm" Stand For? A Five-Question Critique of the Exposure Measure in an Earlier Composite Parkinson's Disease Study, and Whether Its Error Differed by Case Status

The Study and Its Measure

An earlier composite case-control study in the same valley enrolled 210 people with Parkinson's disease and 420 controls and reported an association between Parkinson's disease and farm exposure. Exposure was measured with two yes-or-no questions asked after diagnosis: "Have you ever lived on a farm?" and "Have you ever worked on a farm?" A yes to either counted as exposed. The study concluded that farm exposure, and by implication pesticides, may raise the risk of Parkinson's disease. This critique asks five questions of the measure and then asks whether its errors could have differed between cases and controls.

Question 1: What Content Does the Measure Capture?

"Farm" stands for many exposures at once: pesticides of many kinds, well water, animal contact, dust, solvents in machinery, rural childhood and diet. A yes cannot distinguish among them, so an association with farm living cannot be attributed to herbicides or to any single agent. Studies that asked about specific pesticides, such as the nested case-control study within the Agricultural Health Study that examined pesticides grouped by mechanism, could link risk to particular products in a way the farm question cannot (Tanner et al., 2011). The measure also treats a childhood summer on a relative's farm the same as forty years of spraying.

Question 2: Who Answered?

The study allowed proxies, usually spouses or adult children, to answer for participants who could not. Proxies were more common among cases, some of whom had cognitive impairment or speech difficulties, than among controls. Proxies may know whether someone grew up on a farm but are less likely to know about work done before the marriage or details of tasks. If proxies answer differently from participants, and proxies cluster among cases, the error differs by case status.

Question 3: What Instrument Was Used?

Two yes-or-no questions are quick and easy to answer, but they discard nearly everything that matters for exposure: duration, intensity, timing, tasks and products. Teschke et al. (2002) argue that case-control studies gain accuracy from structured questions about specific tasks and periods, because concrete questions are easier to answer accurately and less open to interpretation than general ones. The farm questions are the opposite of that approach.

Question 4: What Time Frame?

"Ever" has no time frame. Exposure in childhood, in working years and in the decade before diagnosis all count equally. For a disease preceded by years of subtle change before diagnosis, exposure in the final years may reflect the early illness rather than cause it, for example if people with early symptoms moved to quieter rural homes. A measure without timing cannot separate these possibilities.

Question 5: Was the Measure Checked?

The study reported no validation of the farm questions against any other source, such as residential histories, employment records or a second interview. Without a check, there is no estimate of how often answers were wrong or whether errors differed between groups.

Could the Error Differ Between Cases and Controls?

Crude measures are often defended on the grounds that their error is nondifferential, the same for cases and controls, and so would only weaken an association. Armstrong (1998) shows that nondifferential classical error in a continuous exposure generally attenuates associations, but the farm measure raises three reasons to doubt the error was nondifferential here.

Recall

People with a serious diagnosis often search their past for causes. A case who has read that pesticides are linked to Parkinson's disease may be more likely to remember and report a brief period on a farm than a control with the same history. Coughlin (1990) describes this pattern as recall bias, a form of differential misclassification that can create or exaggerate associations.

Proxies

As noted, proxy answers were concentrated among cases. If proxies underreport farm work done before they knew the participant, they would bias in the opposite direction, toward no association, for those cases.

Prodromal Moves

If some cases moved to rural or farm homes during the years before diagnosis, the measure would record exposure that followed the earliest changes of the disease rather than preceding it.

These biases point in different directions, so their net effect cannot be predicted. What can be said is that the measure does not support the assumption that error was the same for both groups, and therefore the study's estimate could be biased either toward or away from no association.

A Rough Sense of Scale

How much could differential recall matter? Suppose, for illustration, that 40% of controls truly had farm exposure and that cases did as well, so that the true odds ratio was 1.0. If cases who had any brief farm contact reported it 90% of the time while controls reported it 75% of the time, the observed odds ratio would rise above 1, roughly to 1.3 in a sample like this one, purely from reporting. The figures are composite, but the arithmetic shows that modest differences in recall can produce associations of the size the earlier study reported. This is why the critique cannot accept the assumption of nondifferential error without evidence.

What the Study Can Still Say

The study can say that people with Parkinson's disease in the valley more often reported farm living or work than controls did. It cannot say which farm exposure was involved, how much exposure matters or whether the difference reflects true exposure rather than reporting. The critique does not dismiss the finding; it narrows it to what the measure supports.

What a Better Measure Would Need

A replacement would ask about specific products and tasks year by year, use the participant wherever possible and record proxy status, set a time frame with a lag before diagnosis and include a validation step comparing reports among cases and controls. Week 10's assessment plan will build on these requirements. It will also need a budget that can pay for interviews long enough to collect year-by-year detail, which is the main reason short yes-or-no questions remain common in studies of this kind despite their weaknesses.

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

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

Tanner, C. M., Kamel, F., Ross, G. W., Hoppin, J. A., Goldman, S. M., Korell, M., Marras, C., Bhudhikanok, G. S., Kasten, M., Chade, A. R., Comyns, K., Richards, M. B., Meng, C., Priestley, B., Fernandez, H. H., Cambi, F., Umbach, D. M., Blair, A., Sandler, D. P., & Langston, J. W. (2011). Rotenone, paraquat, and Parkinson's disease. Environmental Health Perspectives, 119(6), 866-872. https://doi.org/10.1289/ehp.1002839

Teschke, K., Olshan, A. F., Daniels, J. L., De Roos, A. J., Parks, C. G., Schulz, M., & Vaughan, T. L. (2002). Occupational exposure assessment in case-control studies: Opportunities for improvement. Occupational and Environmental Medicine, 59(9), 575-594. https://doi.org/10.1136/oem.59.9.575