Nucleus of the sun-uv-exposure cluster. UV has opposite-signed effects on different
outcomes: it causes skin cancer (harm, RCT + gold-MA grade) while sun-exposure habits associate
with lower all-cause and cardiovascular mortality (benefit, observational only). Both can be
true — they are different endpoints, so this is not a contradiction but a trade-off to net
(type-A). The honest answer is that the net is genuinely uncertain and stratum-dependent, and the
two arms are known with very different confidence.
The net-effect ledger
type-A — an emergent structure present in no single source:
| Arm | Outcome | Best evidence | Effect | Tier |
|---|---|---|---|---|
| HARM | cutaneous melanoma (+ NMSC) | Green RCT + Gandini gold MA | sunscreen HR 0.27-0.50; sunburn RR 2.03, intermittent RR 1.61 (different quantities, not poolable) | RCT + gold |
| BENEFIT | all-cause / CVD / noncancer-non-CVD mortality | Lindqvist single cohort | avoiders ~1.7-2x mortality (HR 0.6 categorical; rate “doubled”); life-expectancy loss 0.6-2.1 y | observational (confounded) |
The tiers are asymmetric, and this is the whole point. The harm is anchored on a randomized trial plus a gold meta-analysis. The benefit rests on a single observational cohort of Swedish women, wide open to reverse causation. Treating the two as equally established would be the error. (The harm arm’s two sources are complementary, not independent — Green authored studies inside Gandini’s pool, and an intervention HR and an etiologic RR are not poolable; see Melanoma and UV Exposure for the full harm arm.)
Parameter table (BLOCKING — the arms are not the same quantity, so this is a trade-off, not a tension):
| Parameter | Harm arm (Green / Gandini) | Benefit arm (Lindqvist) | Same quantity? |
|---|---|---|---|
| Outcome | melanoma incidence | all-cause + CVD mortality | NO — different endpoints |
| Exposure | UV pattern / sunscreen intervention | self-reported sun-habit score 0-4 | NO |
| Design | RCT + observational MA | single observational cohort | NO — different tiers |
Fourth column NO on every row -> the two findings do not contradict (not-joined check ii: different outcome/scope); they must be weighed, not reconciled away.
Why the benefit arm is weak — run the artifact check hard
The U-Shaped Association Artifact: the sun-avoiders in Lindqvist are older, poorer, less educated, far more comorbid (20.4% vs 6.7%), and less active than sun-seekers. Sick, frail people avoid the sun — classic reverse causation / confounding-by-frailty. The association survived adjustment for age, smoking, education, income, comorbidity, exercise, and BMI stratification, which raises but does not settle it; the authors concede (Lindqvist et al., 2016):
«First, it is not possible to differentiate between active sun exposure habits and a healthy lifestyle, and secondly, the results are of an observational nature; therefore, a causal link cannot be proven.»
There is no RCT on the mortality-benefit arm (G — needs a trial the literature does not have). The sun avoidance ~ smoking in magnitude framing (Lindqvist et al., 2016) is a life-expectancy comparison from this cohort, not a validated causal equivalence.
The vitamin-D mediator is probably NOT the mechanism (a key gap)
The intuitive mediator is vitamin D — but supplementation RCTs are largely null on hard outcomes (Deficiency Repletion vs Enhancement: VITAL found no mortality/cancer/CVD benefit in a replete population). If the mortality benefit were vitamin D, a pill would reproduce it; it does not. So the benefit, if causal at all, likely runs through non-vitamin-D UV pathways (nitric-oxide -> blood pressure, melatonin, skin beta-endorphin) rather than through 25(OH)D (Lindqvist et al., 2016):
«Whether the positive effect of sun exposure demonstrated in this observational study is mediated by vitamin D, another mechanism related to UV radiation, or by unmeasured bias cannot be determined from our results. Vitamin D levels might be just a marker of sun exposure.»
This matters for the decision: it means oral vitamin D is not a substitute for whatever (if anything) sun does here — a surrogate that has failed its own transmission test.
Same exposure pattern on both arms — a genuine trade-off, NOT a dovetail
The tempting move is to say harm and benefit attach to different exposure patterns (intermittent burning causes melanoma; regular habitual sun lowers mortality), so a single strategy could capture both. The sources do not support that. Lindqvist’s sun-exposure score is built entirely from recreational/intermittent items — summer sunbathing, winter/mountain sunbathing, holidays abroad to sunbathe, and tanning-bed use (Lindqvist et al., 2016). That is precisely Gandini’s harmful intermittent pattern (and tanning beds are themselves an established melanoma risk factor) — not the null chronic/occupational pattern. So the same sun-seeking behaviour is what raises melanoma in Gandini AND associates with lower mortality in Lindqvist: a real trade-off on one exposure, not two exposures that conveniently separate.
- The one exposure plausibly robust across both arms — regular, genuinely non-burning exposure — is a hypothesis, because it is not what Lindqvist measured; its score cannot distinguish non-burning habitual sun from recreational burning. Do not present it as an evidenced sweet spot.
- Avoiding sunburn remains the well-evidenced move (harm arm), but it does not, on this evidence, come free of the mortality-arm’s exposure — the two are entangled in the same recreational habit.
- Sunscreen is the partial escape: in Green it cut melanoma without reducing outdoor time (Green et al., 2011) — but Lindqvist cautions against reading sunblock as a licence to overexpose.
Stratum-dependence — the net shifts (mostly via baseline melanoma risk)
- Fair skin (Fitzpatrick I-II) + high ambient UV (e.g. N. Australia, UV index >=6 year-round): melanoma harm dominates; protection is the clear call.
- Darker skin and/or high latitude, low UV (e.g. Sweden, UV index <3 for 8-9 months): melanoma baseline risk is far lower, so the same relative harm is a much smaller absolute harm and the (weak, confounded) mortality-benefit arm carries relatively more weight; Lindqvist argues restrictive year-round sun-avoidance advice may not be beneficial there. (This is a baseline-risk shift, not a vitamin-D-repletion argument — the D route was discounted above.)
- Route (a) baseline-risk stratification does most of the work: the same relative melanoma effect is a much larger absolute harm in fair, high-UV populations.
Bottom line (do not collapse to a slogan)
- Avoiding sunburn is the robust, well-evidenced move (harm arm is RCT/gold-grade).
- The claimed all-cause mortality benefit of sun-seeking is observational, confounded, and unreplicated — plausible but far from established; oral vitamin D does not stand in for it.
- The net is stratum-dependent; the outcome menu (melanoma risk vs mortality vs the unmeasured quality-of-life trajectory) is the person’s to weight (layer 3). The loop is open — no source here measured the net of a sun-seeking vs sun-avoiding strategy on all-cause mortality in a trial.
- Opens the sun/UV domain -> candidate deliverable once a second mortality-arm source (or a Vitamin-D MR / VITAL-style analysis on the sun question) lands. AWAITS a second sun-exposure mortality cohort or Mendelian-randomization source — to test whether the Lindqvist all-cause benefit replicates outside one Swedish female cohort.