Critically evaluating the evidence: risk versus benefit for sun exposure

The past 10 years have seen an explosion of interest and research into possible beneficial effects of sun exposure – primarily through its role in vitamin D production. The accruing evidence of benefit is now challenging the prevailing sun protection paradigm, but has not yet resulted in clear public health guidelines on safe sun exposure.

Minimizing exposure to the sun to avoid the adverse health effects of UV radiation (UVR) has been the predominant paradigm for almost 50 years. In light of increasing skin cancer rates, clear research evidence of a causal role of sun exposure and the threat of stratospheric ozone depletion, have led public health authorities, cancer councils globally (but in particular those in high UVR locations, such as Australia and New Zealand) to vigorously promote sun protection. However, even during this time, several isolated research reports hypothesized that higher levels of sun exposure were associated with a lower incidence of multiple sclerosis [1], colon cancer mortality [2] and coronary heart disease [3] (i.e., there were beneficial effects as well as adverse effects from sun exposure). As evidenced in the review by Grant in this issue [4], this has become a highly productive research area. Indeed, there are now calls to rethink sun-exposure policy or to promote vitamin D supplementation.

However, any change in sun-exposure policy requires careful prior examination of the evidence, using the principles of evidence-based medicine to assess whether there is sufficient evidence to infer a protective effect of sun exposure for various diseases. Furthermore, few chronic diseases have a single cause; rather, multiple causal factors increase risk of disease development additively or in an interactive manner. The population attributable fraction (PAF) estimates the proportion of the disease risk that is attributable to the specified exposure and could, thus, be eliminated if exposure (in this case inadequate sun exposure) was reduced to zero. The PAF depends on both the strength of the association between the exposure, and the outcome and on the prevalence of the exposure in the population (i.e., avoiding an uncommon exposure, even one with a very strong causal association with an outcome, may have only a minimal effect on disease incidence).

In evidence-based medicine, there is a hierarchy of evidence [5]. In general, the strongest evidence comes from well-conducted, blinded, randomized clinical trials (RCTs) and systematic reviews of the results of such trials. However, RCTs are not always possible for both practical and ethical reasons, particularly for exposures that may alter the incidence of cancers, with origins many years before diagnosis. Furthermore, care must be taken when interpreting outcomes from such trials that were not designed as end points and, therefore, may not have been rigorously measured (e.g., episodic self-reports of colds or influenza in a trial primarily examining the effect of vitamin D supplementation on bone loss in postmenopausal women) [6].

Individual-level, observational epidemiological studies fall in the next broad level of evidence (case–control and cohort studies). These are noninterventional studies examining naturally occurring associations between various exposures of interest and disease outcomes. These studies are based on comparing sets of individuals, in each of which, information has been collected on both exposures and disease development. As nonexperimental studies, they are subject to bias and confounding factors, so that careful appraisal of the study design, conduct and analysis (including adjustment for possible confounders) is necessary for the results to be considered valid. Lower levels of evidence, from individual study subjects, are case series and anecdotal reports.

Other observational studies, sometimes called ecological studies, examine correlations between variables measured at the group level [7]. Such studies have value for the exploration of disease patterns within and between populations, enabling both hypothesis generation and, in limited situations, hypothesis testing. They reflect the experience of whole populations and allow comparisons between populations. However, a frequent weakness with such studies in assessing whether there is a causal role for the exposure of interest is the lack of certainty that those who had the ‘exposure’ were the same individuals who developed the disease. Relatedly, the adjustment for possible confounding factors typically uses population-level averages, often of some proxy measure of the confounding variable, such as using lung cancer rates as an index of smoking prevalence.

Grant has marshaled much compelling evidence to support an important role for greater sun exposure or vitamin D supplementation for better health. However, for the moment, much of the available evidence is of the observational, population-level category – with its inherent limitations. This has two important implications. First, the plausibility of any causal inference that might be drawn needs especially critical scrutiny. Second, that there is clear need, from this point, for a widened spectrum of research that draws proportionally more on individual-level studies and, where possible, experimental studies.

In the assessment of whether an association is causal, Bradford Hill’s nine ‘criteria’ are a commonly used framework [8]. Notably, strength of the evidence refers not only to the number of studies published that demonstrate the relationship, but takes account of the strength of the study design (as outlined previously), the internal validity of the study and appropriate analysis of the results (with control for possible confounding factors). The findings should be consistent across well-designed individual-level studies and be consistent with patterns evident in ecological studies. Biological plausibility depends on the current knowledge. Vitamin D is postulated to stimulate immune function to protect against both viral and bacterial (with a clear effect only recognized for tuberculosis) infections, but have a Th1 immunosuppressant effect that may protect against autoimmune diseases such as Type 1 diabetes and multiple sclerosis. Although UVR exposure initiates vitamin D production, it has independent immunosuppressive effects: depressing the cellular response to vaccination, promoting the reactivation of latent viral infections, suppressing Th1 immune function and stimulating T regulatory cells. Further clarification of the roles of vitamin D and UVR exposure in immune function is required.

Of Bradford Hill’s nine criteria, only that of temporality is considered essential for causality (i.e., that the exposure precedes the disease and by an interval that is biologically plausible for disease development). Thus, the finding of a protective effect of modest vitamin D supplementation over 1–4 years for the development of cancer may not be consistent with work suggesting that cancer can take “15–40 years to progress from initiation to detection or death” [4]. Temporality is most clearly established in prospective (RCT and cohort) studies, but can be inferred from case–control studies if particular attention is paid to this requirement in the study design. Ecological studies, examining disease patterns using population-level data, may demonstrate changing exposure and disease patterns over time but often cannot clearly demonstrate temporality.

What is the exposure of interest?

In evaluating the evidence presented to encourage change in sun-exposure policy or vitamin D supplementation it is important to critically examine the ‘exposure’ measure used. Ideally the ‘exposure’ is measured vitamin D status, usually serum 25-hydroxyvitamin D (25[OH]D) or personal UV dose (a combination of ambient UVR [a function of latitude, altitude, atmospheric ozone levels and time of year], skin exposed [influenced by behavioral, cultural and clothing practices], and skin pigmentation [with a smaller effective dose to underlying structures in dark compared with light skin]). Individual-level studies use latitude or ambient UVR of residence, history of sunburns, history of time in the sun at various ages, dietary and supplemental vitamin D intake and others as proxy measures. More recently, at this individual level, our ability to accurately measure UV dose has improved with the use of silicone rubber casts of the back of the hand as an objective measure of cumulative sun damage. Current vitamin D status is assessed by measuring serum 25(OH)D and this test is widely available. Ecological studies must focus on ecological measures, such as latitude, ambient UVR (or UVB radiation, generally calculated from satellite data), or population averages of sun-related diseases (e.g., skin cancers). These are proxies for the exposure of real interest, and there are drawbacks to each in inferring that a relationship with this proxy is a relationship with personal UV dose or vitamin D status.

While ambient UVR varies approximately by latitude (with further variation by altitude, levels of atmospheric ozone and pollutants and other factors), so do a variety of other possible etiological factors, including temperature, diet, exposure to infectious agents and, possibly, physical activity levels: that is, there is no specificity for ambient UVR. Furthermore, under any level of ambient UVR, personal UV dose may vary greatly depending on sun exposure behavior (itself moderated by ambient temperature), clothing and skin pigmentation. Although the incidence of squamous cell carcinoma of the skin may be a useful proxy for lifetime UV dose (as it appears to be caused by cumulative sun exposure), this is unlikely to be true for cutaneous malignant melanoma or basal cell carcinoma where the risk may be related to intermittent high-dose exposure during childhood and adolescence. Unfortunately, nonmelanoma skin cancer incidence is not routinely recorded in cancer registries so that data availability is patchy. Without universal registration of skin cancers it is difficult to interpret ‘incidence’ rates: those skin cancers that are registered may be systematically different (e.g., more aggressive or in unusual locations), from those that are more commonly removed by destructive means without histological diagnosis. Skin cancer mortality is rare (particularly for squamous and basal cell carcinomas) and although partly related to incidence, is also strongly affected by access to healthcare and patient education. As such, skin cancer mortality may be a poor proxy of personal UV dose.

Importantly, there is no empirical evidence that any of these measures are proxies for vitamin D status. Indeed, as Grant notes, in addition to the likely evolution of lighter skin in response to lower ambient UVR and the need to maintain vitamin D levels, populations living at high latitude traditionally eat diets high in vitamin D to avoid the consequences of vitamin D deficiency [4]. This would tend to dissociate latitude/ambient UVR and vitamin D status at least at high latitudes. There is, to date, no research evidence that population levels of serum 25(OH)D are correlated with latitude or with ambient UVR. Although it seems logical that there should be an association between vitamin D status and ambient UVR, there are important moderating effects from diet and indoor/outdoor lifestyle, clothing and skin pigmentation.

That there are both adverse and beneficial effects of sun exposure on health, either independently or mediated through enhanced vitamin D production, is clear. While the adverse effects are relatively well-defined, evidence is still accumulating on possible beneficial effects. In addition, there is considerable evidence that there may be independent beneficial effects of UVR exposure that are not mediated by vitamin D [9].

Although much of the early evidence on the adverse effects of sun exposure arose from ecological studies, subsequent observational studies supported an important role of sun exposure as a cause of skin cancers, cataracts of the eye, pterygium and other adverse effects. Interestingly, ecological studies did not demonstrate a clear-cut latitudinal pattern in melanoma occurrence – a result of confounding by skin pigmentation and patterns of sun-exposure behavior, such that fair-skinned populations living at high latitude (low-ambient UVR) experiencing episodic high UV exposure were at greater risk than darker-skinned populations inhabiting lower latitudes. To date, much of the evidence for beneficial effects arises from ecological studies, generally with observational studies supporting some findings (e.g., a beneficial effect of vitamin D, particularly combined with calcium on colorectal cancer and osteoporosis), but with mixed results for other outcomes.

It remains unclear what the balance is between adverse and beneficial effects. Diffey suggests a flattened U-shaped association between disease burden and UV dose [10]. Although this suggests that there is a rather broader level of ‘safe’ UVR exposure, compared with an earlier figure by Lucas and Ponsonby [11], other work suggests threshold effects in regard to vitamin D levels and some autoimmune diseases [12,13] perhaps a reverse J-shaped relationship with regard to overall disease burden [14] or a steeply rising, near-linear association (at least initially) for both adverse [15] and beneficial effects.

So, where are we now? How should we be guiding our patients and the public more generally? In 2007, the Cancer Council Australia issued a new position statement [16] (updated from 2005) that was based on the meeting and subsequent deliberations of a working group composed of representatives involved in both the adverse and beneficial effects of sun exposure. The group comprehensively reviewed the current state of the evidence to provide a balanced evaluation and subsequent revision of sun protection messages. Internationally, in 2005 the WHO convened a similar meeting in Munich (Germany) to consider the evidence. Both meetings agreed that there are beneficial effects of sun exposure – for colon cancer and osteoporosis. But further research is required before the evidence is considered ‘sufficient’ for diseases such as multiple sclerosis, Type 1 diabetes and other solid cancers. One critical question remaining is whether all of the beneficial effects of sun exposure can be obtained from taking vitamin D supplements, or whether there are independent beneficial effects of UVR exposure itself. The Cancer Council Australia now recommends that below a UV index of 3 (the UV index is part of the daily weather forecast in Australian newspapers), modest sun exposure without sun protection is safe. When the UV index is 3 or greater, sun protection is advised.

This may or may not be sufficient. While some research suggests that only brief exposures to 10–15% of the body surface (face and arms) 2–3 times per week is sufficient to maintain vitamin D levels, this does not fit with other evidence demonstrating that hypovitaminosis D is widespread. What is critical is that only UVB wavelengths initiate vitamin D synthesis – this is different from the action spectrum for erythema and the UV index (which include a long UVA tail) and the UV index may not be a sufficient guide to appropriate sun exposure for vitamin D synthesis. In addition, sun-protection messages urging people to stay out of the sun during the middle hours of the day (10 am–2 pm) and to get their sun exposure outside of those times, may be ensuring that such exposure is to UVA wavelengths – possibly increasing skin cancer risk with little change in vitamin D synthesis [17].

There appears to be many outstanding unknowns in this debate. It now seems clear that the level of serum 25(OH)D previously considered to denote adequacy (>50 nmol/l), is in fact inadequate, and a new lower normal limit of 75 nmol/l is now being widely accepted. However, it is still not clear how much sun exposure is required to reach and maintain vitamin D adequacy – taking account of latitude/ambient UVR, time of day and year, skin pigmentation, and position of the individual, such as standing or walking. Most importantly, further research is required to achieve sufficient evidence for many of the hypothesized causal associations. And even where the evidence that low vitamin D or sun exposure has a causal role is considered sufficient, we need to assess the PAF – a causal role may be proven, but the exposure may be a minor risk factor of little significance.

Financial & competing interests disclosure

Dr Lucas is supported by a NHMRC ‘Capacity Building Grant, Environment and Population Health: Research Development from Local to Global, 2003–2007’ (No. 224215), a Macquarie Bank MS Research Australia Fellowship and a RACP Cottrell Fellowship. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.