On the Nature of Evidence and 'Proving' Causality: Smoking and Lung Cancer vs. Sun Exposure, Vitamin D and Multiple Sclerosis

Abstract

If environmental exposures are shown to cause an adverse health outcome, reducing exposure should reduce the disease risk. Links between exposures and outcomes are typically based on 'associations' derived from observational studies, and causality may not be clear. Randomized controlled trials to 'prove' causality are often not feasible or ethical. Here the history of evidence that tobacco smoking causes lung cancer-from observational studies-is compared to that of low sun exposure and/or low vitamin D status as causal risk factors for the autoimmune disease, multiple sclerosis (MS). Evidence derives from in vitro and animal studies, as well as ecological, case-control and cohort studies, in order of increasing strength. For smoking and lung cancer, the associations are strong, consistent, and biologically plausible-the evidence is coherent or 'in harmony'. For low sun exposure/vitamin D as risk factors for MS, the evidence is weaker, with smaller effect sizes, but coherent across a range of sources of evidence, and biologically plausible. The association is less direct-smoking is directly toxic and carcinogenic to the lung, but sun exposure/vitamin D modulate the immune system, which in turn may reduce the risk of immune attack on self-proteins in the central nervous system. Opinion about whether there is sufficient evidence to conclude that low sun exposure/vitamin D increase the risk of multiple sclerosis, is divided. General public health advice to receive sufficient sun exposure to avoid vitamin D deficiency (<50 nmol/L) should also ensure any benefits for multiple sclerosis, but must be tempered against the risk of skin cancers.

Keywords: association; causality; epidemiology; lung cancer; multiple sclerosis; smoking; sun exposure; vitamin D.

Figure 1

The hierarchy of evidence, increasing from the base of the pyramid to the gold standard for establishing causality.

Figure 2

Death rate from lung cancer and consumption of tobacco and cigarettes over time in the United Kingdom (redrawn using data from [17]).

Figure 3

Incidence of a first clinical diagnosis of central nervous system demyelination, as a marker of multiple sclerosis incidence, in Australia in 2003–2006, based on data from the Ausimmune Study [20] (dotted lines represent latitude, annual incidence (per 100,000 population) is circled).

Figure 4

Data from Sweden on age-adjusted incidence (per 100,000 person years) of MS overlaid with the prevalence of 25(OH)D level ≥75nmol/L (plotted from data available in [24] and [25]) in time periods from 1946–2005.

Figure 5

Odds of being a case with a first demyelinating event (a precursor of MS) by quintile of serum 25(OH)D, with the highest quintile as the reference category. Only the lowest quintile of 25(OH)D concentration is associated with a significant increase in risk (replotted from data from the Ausimmune Study [32]).