Page 90 - Vitamin D and Cancer

4 The Epidemiology of Vitamin D and Cancer Risk 77

4.2.4 Case–Control and Cohort Studies of Sun Exposure

Self-reported sun exposure or surrogates such as region of residence and number
of sunburns can be used in epidemiologic studies. A number of ecologic studies
have examined the vitamin D and cancer hypotheses at the population level, but
some case–control and cohort studies, which assess exposure and outcome at the
individual level are now available. In principle, confounding may be better
controlled because typically more detailed information can be assessed on other
covariates in analytic studies. In addition, the study population may be relatively
homogeneous, which may reduce the potential for residual or uncontrolled
confounding. An additional strength of such studies is that exposure is actually
assessed for the individual, whereas in ecologic studies, exposure is inferred –
for example, presumably living in sunnier regions may allow for greater oppor-
tunity for sun exposure, but actual exposure will depend on individuals’
behaviors.
The sun exposure studies have some strengths and some limitations. They do not
directly assess vitamin D exposure, and some surrogates that have been used (such
as sunburns) may represent acute short-term exposures to sun rather than chronic
exposures, which may be more relevant for vitamin D synthesis. There also may be
measurement error and perhaps recall bias in case–control studies in assessing past
exposures. Some objective methods to assess sun exposure, such as the use of
reflectometry, may be useful. One important advantage of these studies is that most
blood-based and dietary cohorts are in middle-aged individuals, and the assessment
of past sun exposures allows the possibility of estimating vitamin D status at points
earlier in life. For some cancers, it is plausible that these earlier time periods may
be most relevant.

4.2.5 Randomized Trials

A double-blinded, placebo-controlled, randomized intervention is the “gold stan-
dard” in establishing a causal association because in theory, confounding can be
largely eliminated as an explanation of a positive result. Because of their expense,
these studies have been rarely done in the context of vitamin D and cancer.
In practice, these studies have practical limitations, including selection of the
effective dose, varying baseline levels of the exposure of interest, poor compli-
ance, contamination by the placebo group adopting the change, and the unknown
but presumably long induction period for cancer. Thus, when these studies show a
null association, caution must be given not to overinterpret the results. Besides the
absence of a true association, one or more of the limitations mentioned above
could produce a null association. If a significant association is found, such studies
are the strongest evidence of a causal association.