The timing of exposure relative to disease occurrence is among the most underappreciated aspects of exposure assessment and thus merits special emphasis. Some exposures are constant over time, such as genetic constitution or attributes defined at birth. However, exogenous exposures such as diet, medication use, social circumstances, and chemical pollutants vary over time, often substantially. Any choice of exposure measure implicitly or explicitly includes an assumption about the exposure interval that is relevant to disease etiology. The critical time window may be based on calendar time, age at exposure, or exposure relative to the occurrence of other exposures or events (Rothman, 1981). A corollary to the need to isolate the biologically relevant exposure is to pinpoint the etiologically relevant time window of exposure so that we can ignore etiologically irrelevant exposure that occurs outside that window. Inclusion of irrelevant exposure constitutes exposure misclassifica-tion relative to the ideal measure.
In examining the role of cigarette smoking in the causation of lung cancer, for example, we recognize that there is some interval between exposure and the occurrence of disease that is not relevant to its causation. The number of cigarettes smoked on the day of diagnosis is clearly not relevant, for example, nor are the cigarettes smoked during the period in which the tumor was present but undiag-nosed. Under some hypothesized mechanisms, the exposure for months or years prior to the diagnosis may be irrelevant. In the face of such uncertainty, Roth-man (1981) has argued for flexibility in evaluating temporal aspects of exposure. A series of reasoned hypotheses may be put forward based on alternative theories about disease causation.
In some instances, different timing of exposure corresponds to entirely different etiologic pathways. The role of physical exertion in relation to myocardial infarction appears to include an acute adverse effect, such that intense exertion is followed by some relatively brief period of increased risk (Siscovick et al., 1984; Albert et al., 2000), as well as a chronic beneficial effect, such that regular exercise over periods of months or years reduces risk of a myocardial infarction (Rockhill et al., 2001). The timing of the intense activity levels relative to the long-term history of activity may also be relevant, with a change from long-term inactivity to a higher level of activity a possible threat to health. Any attempt to measure physical activity in order to evaluate its association with risk of myocardial infarction would require carefully formulated etiologic hypotheses that address the temporal aspects of exposure and effect.
In other etiologic pathways, the critical exposure window may be defined not in relation to the timing of disease but based on stage of development. Regardless of when congenital malformations come to attention, precisely timed developmental events indicate the days and weeks of gestation in which certain defects can be produced by external insults (Kline et al., 1989). Similarly, it has been hypothesized that physical activity among adolescent girls is influential on their long-term risk of osteoporosis (US DHHS, 1996). For illustrative purposes, assume that this window of adolescence is the only period in which physical activity is pertinent to osteoporosis. A study that measured lifetime physical activity or physical activity from ages 40 to 49 would suffer from misclassification and observe the expected inverse association only to the extent that physical activity at those measured ages corresponded to physical activity in adolescence.
Was this article helpful?