As UV radiation accelerates skin aging and promotes skin cancer, novel photoprotective measures represent a promising area of research. Dietary polyphenols are natural antioxidant and anti-inflammatory agents that could protect skin against UV exposure. In our population of healthy non-smoking women, 12-week HFC intake was associated with significantly increased net skin elasticity but no significant change in MED. Participants did not report any significant clinical side-effects.
The current literature suggests that chronic ingestion of HFC may be photoprotective, but there are several methodological limitations. Only 1 controlled, double-blind, RCT evaluated the effect of HFC vs. LFC intake on skin sensitivity to UV radiation after 12 weeks, as measured by MED . 22 women and 8 men with Fitzpatrick skin phototypes II or III were included in that trial. After 12 weeks, mean MED remained stable in the LFC group compared to baseline. Statistically significant increases in mean MED were observed with HFC (within-group comparisons). Nevertheless, these authors did not report between-group comparisons. Moreover, women and men were included in their trial, but participants’ characteristics in each group were not elaborated. Finally, our study population was quite different from that of Williams’  as we included only women with skin type I or II.
Heinrich et al. enrolled 24 women in their RCT and evaluated the effect of HFC vs. LFC intake on MED in solar simulation radiation . After 12 weeks, women assigned to the HFC group showed significantly increased MED in comparison to the LFC group. It is important to note that the characteristics of women in each group were not presented, and no information was given regarding trial profile (loss-to-follow-up, intent-to-treat analysis). Finally, visually-assessed MED, considered to be the gold standard , was not available. Thus, direct comparison with other studies was not possible.
Several factors could have contributed to our finding of no statistically photoprotective effects of HFC vs. LFC measured by MED at 12 weeks. Season of participation differed in the 2 groups. Specifically, more participants in the LFC group started the trial during winter and finished at the 12-week endpoint during the spring season, potentially introducing a residual confounding effect. During this period, physiologically-increased MED is attributed to longer solar exposure [42, 43]. Indeed, exploratory analyses of our data demonstrated a boost in MED during spring, compared to its stability during winter and, inversely, a decrease during fall (Additional file 1: Figure S2). This extrinsic effect could partly explain the shape of the LFC group curve in Figure 2, in which it can be seen that MED increased from baseline to week 12. Consequently, the increment could have masked difference from the HFC group. Differing season of participation could also have introduced bias in the results, since roughly one-third of participants in each group completed the study during spring, when the natural increase in MED could have rendered the photoprotective effects of chocolate less visible. Indeed, multivariate analyses revealed that HFC group participants terminating during spring manifested lower changes in MED. Furthermore, it would have been interesting to study women over the age of 50 years, as our multivariate analyses disclosed greater changes in MED in this population, indicating that HFC may have significant photoprotective effects after menopause in comparison to control LFC. Finally, MED imbalance at baseline rendered our groups less comparable, which we might not have completely accounted for despite adjustments in our multivariate model.
Lack of compliance with chocolate consumption is another factor that could partially explain our non-significant results. In fact, the HFC group showed a stronger increase in MED at week 6 than at weeks 9 and 12 compared to baseline, mimicking the same tendency found in plasma polyphenol concentrations at these time periods. Women could have been slightly less compliant with chocolate consumption after 6 weeks, rendering its effects even less noticeable.
It would have been interesting to compare HFC with LFC low in theobromine, the primary alkaloid in cocoa. Theobromine concentrations in our LFC and HFC groups were similar. Although it has never been studied extensively in humans, sparse data indicate that it has been tested for the treatment of hypertension to exploit its vasodilator and smooth tissue-relaxing properties [44, 45]. Theobromine may contribute to the effect of dark chocolate on endothelium function. Therefore, it could have increased the microcirculatory delivery of flavanols in both groups (as the LFC group also showed significant flavanol elevation), and masked the isolated action of polyphenols.
Moreover, since MED was highly variable between women participating in our clinical trial (as relative SD was 25%), the non-optimal power of our study sample could partially explain the non-significant results. Furthermore, the variability of our results is highlighted by the large CIs revealed by multivariate analysis. In addition, it’s not possible to exclude that imbalance in numbers allocated to HFC and LFC partly explained by the 2 independently generated randomisation lists and differences between the 2 groups for skin type and date of entry could somewhat bias our results. Indeed, as specified in the section methods, after three months of recruitment, proportion of women with skin type 2 and age 50–65 were more prevalent than expected and a new independent list of randomisation was generated. Moreover, in the adjusted model, age group and skin type didn’t modify the results.Interestingly, after 3-week washout, the HFC group showed decreased MED with return to baseline. Conversely, the LFC group manifested continuously elevated MED. We would, therefore, have expected similar increment in the HFC group, instead of the observed decline. This differential tendency after removal of our intervention, well-illustrated in Figure 2, indicates loss of a clinical effect too small to be statistically significant due to the afore-mentioned factors contributing to lack of significance.
MED represents the lowest UV dose necessary to produce just perceptible erythema at 24 hours. The erythema reaction to UV is the endpoint of complex biological processes, including direct DNA, lipid and protein damage, activation of pro-inflammatory pathways and generation of free radicals, mainly reactive oxygen species. Polyphenols are expected to act mainly on the latter by supplementing the body’s natural, free-radical-quenching, antioxidant mechanisms [10–14] . A recently-published review  of photoprotection and antioxidants notes that the action spectrum of reactive oxygen species generation is predominately in the ultraviolet A (UVA) range, although there is some overlap with UVB. Thus, perhaps antioxidants play a larger role in protecting against UVA-induced production of free radicals. Consequently, measurement of the photoprotective effects of flavanols solely by MED after exposure to UVB might not be fully representative of their real biological potential in this regard .
A positive outcome of HFC consumption was noted in net temple elasticity at 12 weeks compared to LFC, but was not reproduced in arm elasticity, suggesting that flavanols might impact only sun-exposed skin for this outcome. Furthermore, the gain in elasticity was not lost after 3-week washout. The underlying mechanisms are not known, but augmented blood flow  can lead to heightened production of collagen and elastin, key structural proteins which diminish in the skin with aging. On the other hand, HFC consumption did not modify skin hydration. Heinrich et al. postulated improvements in skin density and thickness, stratum corneum hydration, transepidermal water loss, and skin surface roughness after 12 weeks of flavanol-rich cocoa drink consumption in healthy women .
HFC and LFC consumption did not significantly change BMI and blood glucose, a previously-reported finding . Lipid profile also was not affected, except for a slight but significant rise in high-density lipoprotein-cholesterol level in the HFC compared to the LFC group at week 12. Participants did not report any significant clinical side-effects. In all, chocolate consumption did not result in notable adverse events.