Epidemiological studies confounding variables

We are not certain which comorbid risk factors cause mortality independent of sleep effects, and therefore, we cannot be certain whether we controlled too much or too little for comorbidities. For example, since short sleep or long sleep may cause a person to be sick at present or to get little exercise or to have heart disease (17), diabetes (18), etc., controlling for these possible mediating variables may have incorrectly minimized the hazards associated with sleep durations. This would be overcontrol. The hazard ratios for participants who were rather healthy at the time of the initial questionnaires were unlikely to be overcontrolled for initial illness. Since the 32-covariate models and the hazard ratios for initially healthy participants were similar, this similarity reduced concern that the 32-covariate models were overcontrolled. On the other hand, there may have been residual confounding processes that caused both short or long sleep and early death that we could not adequately control in the CPSII data set, either because available control variables did not adequately measure the confound or because the disease did not yet manifest itself. Depression, sleep apnea, and dysregulation of cytokines are plausible confounders that were not adequately controlled. It may be impossible to be confident that all conceivable confounds are adequately controlled in epidemiological studies of sleep. 

Table 9-2 also presents data from occupational and epidemiologic studies that indicate that the respiratory system is the primary target for sulfur dioxide. There was variability in the study findings that probably resulted from a lack of adequate analytical measurements (use of area sampling rather than personnel monitoring) the multiplicity of confounding, concurrent exposures to other chemicals and participates and the study indices investigated. However, some reasonable correlations between effects reported and exposure bounds can be determined. 

Case reports can be very useful and have been the mainstay in identification of human teratogens. Usually there is no information on confounding variables, and the actual cause and effect relationship cannot be established by the case report alone. Many known human teratogens were established by a consensus of case reports, later to be confirmed by epidemiologic evaluation. Correlation studies seek relationships between geographic location, time of exposure, personal characteristics, and pregnancy... 

A thorough review of markers of DNA repair and susceptibility to cancer in humans from 2000 [64] has summarized the results from 64 epidemiologic studies that addressed the association of cancer susceptibility with a putative defect in DNA repair capacity. The authors concluded that the vast majority of studies showed a difference between cancer case subjects and although this observation is compatible with a chromosomal instability due to the cancer itself, it was notable that impaired mutagen sensitivity was also observed in healthy relatives of cancer subjects. There were a variety of functional tests used that only indirectly addressed DNA repair and these showed high variability in their expression. Finally, the issue of confounding was almost totally unexplored. 

On occasion, such a survey may indicate the need for a more formal epidemiological study. Skin complaints may, for example, be widespread but unusually hard to explain. Such investigations should never be undertaken by clinicians without previous epidemiological and statistical consultation. Coenraads and Nater (1987) have published a useful introduction to the problems that may arise, including true prevalence estimation, bias, confounding variables and sample size. Questions of disease definition and inter-observer variability are not necessarily familiar to clinicians, who may therefore need to seek epidemiological advice at the earliest opportunity. Questionnaires are frequently designed that ask for far more detail than can possibly be usefully analysed statistically, and they should always be piloted first in order to achieve validity. 

The frustrations expressed by Ward and Dye stem from at least two major difficulties in the interpretation of these epidemiological data. The first is the issue of confounding variables of alcohol and age. Across all studies, in approximately 80% of the cases where THC was detected, alcohol was also found. Also, marijuana is primarily used by young males, who are over-represented in fatal crashes and are associated with socially high risk-taking behaviors,... 

Probably the most crucial difference between studies of lead effects, and now the most debated subject in epidemiological studies, is the statistical treatment of data, and the appropriate control of confounding variables. 

---