Photodermatology, Photoimmunology & Photomedicine
COMMENTARY
Comment on Sallander et al. ‘Vitamin D levels after UVB radiation: effects by UVA additions in a randomized controlled trial’ Mary Norval1 & Frank R. de Gruijl2
1
Biomedical Sciences, University of Edinburgh Medical School, Edinburgh, UK. 2 Department of Dermatology, Leiden University Medical Centre, Leiden, The Netherlands.
Correspondence: Professor Mary Norval, DSc, Biomedical Sciences, University of Edinburgh Medical School, Teviot Place, Edinburgh EH8 9AG, UK. Tel: +44 131 6503167 e-mail: [email protected]
Accepted for publication: 18 March 2014
Conflicts of interest: None declared.
We read the paper by Sallander and colleagues (1), recently published in this journal, with considerable interest. They conducted a randomized controlled trial to investigate the effectiveness of different combinations of UVA and UVB radiation on raising the level of 25-hydroxyvitamin D [25(OH)D] in serum. We wish to comment on several aspects of the study. 1. Although clearly not intended, the small number of subjects and the overrepresentation of skin types V and VI in the UVA group compared with the UVB and UVAB groups (UVA n = 10, 30% types V and VI; UVB n = 23, 0% types V and VI; UVAB n = 23, 4% types V and VI) make the robustness of the results uncertain, which perhaps led to some unsubstantiated conclusions, such as skin type not affecting 25(OH)D production. 2. Three types of UV lamps were used, called UVB, UVAB, and UVA. However there was both UVA and UVB wavelengths in all three: the UVB source contained 70% 176
UVA, the UVA source contained 99.3% UVA, and the UVAB source contained 95.3% UVA. Thus, none of the sources emitted solely UVB or UVA wavebands. Exposures from various sources were arbitrarily set at 85 J/m2 UV, weighted with the Commission International de l’Éclairage (CIE) vitamin D action spectrum. Other than this a priori spectral weighting, additivity was not considered or unbiasedly tested. The UVB exposure was delivered from six UV6 lamps in 61 s, and the UVAB exposure in 60 s from the same six lamps with 20 UVA lamps added. Was the switch difference of only 1 s in the cabin accurate, and was the change in the output of the lamps in heating up over 1 min accounted for in the dose? A 1-s difference in UVB exposure is not likely to result in a measurable increase in 25(OH)D level. Hence, the comparison between the UVB and UVAB exposures appears uninformative, with a negligible contribution from the UVA band. In addition, the authors stated that a single grating spectroradiometer (SolaHazard) was used to measure the outputs but that this instrument underestimates low UVB output in the presence of high UVA by a factor of 2 in comparison with other calibrations, where one would expect an overestimation because of stray light. It is not clear if the stated 0.7% UVB content in the UVA lamp was in fact corrected. As the 0.7 % UVB is not accurately defined and as it determines in large part the effective dose of the ‘UVA’ source, its spectral accuracy is crucial for the ‘UVA’ exposure. 3. The authors wrote that no increase in 25(OH)D level occurred as a result of a short UVA exposure. However, the chosen UVA dose was insufficient to detect a significant 25(OH)D increase in combination with dominating UVB in a short exposure time. The dose in the long exposure from the ‘pure’ UVA source was greater than 10-fold compared with the UVB and UVAB doses, and led to a mean increase of 4.7 nmol/L in 25(OH)D in the group with the smallest number of subjects and the highest proportion © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd doi:10.1111/phpp.12122
Commentary
with skin types V and VI. A proportionally smaller change from the short UVA exposure would not be apparent within the much larger increases in 25(OH)D that followed irradiation with UVB (mean increase 11.6 nmol/L) and UVAB (mean increase 13.6 nmol/L); the effect of the added small UVA exposure would fall well within the error range. Furthermore, the authors suggested that some photodegradation could occur following longer exposure to UVA. However, leaving aside uncertainty about the ‘UVA’ dosimetry, no experiments were included in the study to substantiate this suggestion, which was apparently based solely on the validity of spectral weighting with the CIE vitamin D action spectrum (and only seven type I–III individuals in the UVA group). 4. The authors proposed that the negative correlation they found (like many others before them) between 25(OH)D levels before (at t0) and the increase after exposures (at t2) could be ‘caused by the regression to the mean’. To substantiate this, they calculated an alternative correlation – 25(OH)D increase vs. mean value before at t0 and after at t2. Not surprisingly, this lowered the correlation coefficient to the point of losing significance: the mean value narrows the range of levels, given that the largest differences occur from the lowest initial levels, and thus any correlation will decrease. Regression of extremes to the mean is inherent to random variation from sample to sample. If it is strong and dominant, one can assume that ranking individuals according to levels before, at t0, has no significant correlation with ranking them according to levels after exposure at t2. Thus, a rank test would seem more interesting, appropriate, and robust than the statistical methods used in the study. We tried this rank testing on data from a mid-winter 8-week course of either sunbed use (sub-sunburn doses 3×/week) or vitamin D
supplementation (1000 IU/day) (2). Interestingly, we indeed found that ranking before and after sunbed use was not significantly correlated (with n = 35 Kendall’s τ = 0.22, P = 0.067). However, ranking before and after vitamin D supplementation was significantly correlated (with n = 36 Kendall’s τ = 0.35, P = 0.002). Thus, the authors may be correct in proposing that randomization in increases in 25(OH)D levels after UV exposures is greater (i.e. more variability in responses) than after vitamin D supplementation (although in this 8-week midwinter experiment, the irradiation scheme was controlled by subjects themselves according to their skin responses, adding a source of variability in the treatment not present in vitamin D supplementation and in the irradiation protocol of Sallander et al.). The authors concluded that when UVB irradiation was short and intense, the UVA dose delivered in an equally short time did not matter and did not influence the vitamin D level. However, as we point out, the results in their study were based on a comparison of UVB and UVAB exposures that may not be well founded. The additivity of UVB and UVA in increasing 25(OH)D levels was not tested. Instead, and assuming that their dosimetry was correct, the authors showed that spectral weighting with the CIE vitamin D action spectrum was incorrect in predicting increases in 25(OH)D from equally ‘vitamin D-effective doses’ from UVB- vs. UVA-dominated sources (their Table 4). Suggesting that this discrepancy was due to photodegradation of vitamin D by UVA radiation is not substantiated by their data. Thus, overall, the conclusions drawn by Sallander et al. may not be justified, given the uncertainties in the dosimetry and inadequacies in the experimental design. Further detailed investigations are merited in this important and topical area.
REFERENCES 1. Sallander E, Wester U, Bengtsson E, Wiegleb Edstrom D. Vitamin D levels after UVB radiation: effects by UVA additions in a randomized controlled trial. Photodermatol Photoimmunol Photomed 2013; 29: 323–329.
2. de Gruijl FR, Pavel S. The effects of a midwinter 8-week course of sub-sunburn sunbed exposures on tanning, vitamin D status and colds. Photochem Photobiol Sci 2012; 11: 1848–1854.
Photodermatol Photoimmunol Photomed 2014; 30: 176–177 © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
177
COMMENTARY
Comment on Sallander et al. ‘Vitamin D levels after UVB radiation: effects by UVA additions in a randomized controlled trial’ Mary Norval1 & Frank R. de Gruijl2
1
Biomedical Sciences, University of Edinburgh Medical School, Edinburgh, UK. 2 Department of Dermatology, Leiden University Medical Centre, Leiden, The Netherlands.
Correspondence: Professor Mary Norval, DSc, Biomedical Sciences, University of Edinburgh Medical School, Teviot Place, Edinburgh EH8 9AG, UK. Tel: +44 131 6503167 e-mail: [email protected]
Accepted for publication: 18 March 2014
Conflicts of interest: None declared.
We read the paper by Sallander and colleagues (1), recently published in this journal, with considerable interest. They conducted a randomized controlled trial to investigate the effectiveness of different combinations of UVA and UVB radiation on raising the level of 25-hydroxyvitamin D [25(OH)D] in serum. We wish to comment on several aspects of the study. 1. Although clearly not intended, the small number of subjects and the overrepresentation of skin types V and VI in the UVA group compared with the UVB and UVAB groups (UVA n = 10, 30% types V and VI; UVB n = 23, 0% types V and VI; UVAB n = 23, 4% types V and VI) make the robustness of the results uncertain, which perhaps led to some unsubstantiated conclusions, such as skin type not affecting 25(OH)D production. 2. Three types of UV lamps were used, called UVB, UVAB, and UVA. However there was both UVA and UVB wavelengths in all three: the UVB source contained 70% 176
UVA, the UVA source contained 99.3% UVA, and the UVAB source contained 95.3% UVA. Thus, none of the sources emitted solely UVB or UVA wavebands. Exposures from various sources were arbitrarily set at 85 J/m2 UV, weighted with the Commission International de l’Éclairage (CIE) vitamin D action spectrum. Other than this a priori spectral weighting, additivity was not considered or unbiasedly tested. The UVB exposure was delivered from six UV6 lamps in 61 s, and the UVAB exposure in 60 s from the same six lamps with 20 UVA lamps added. Was the switch difference of only 1 s in the cabin accurate, and was the change in the output of the lamps in heating up over 1 min accounted for in the dose? A 1-s difference in UVB exposure is not likely to result in a measurable increase in 25(OH)D level. Hence, the comparison between the UVB and UVAB exposures appears uninformative, with a negligible contribution from the UVA band. In addition, the authors stated that a single grating spectroradiometer (SolaHazard) was used to measure the outputs but that this instrument underestimates low UVB output in the presence of high UVA by a factor of 2 in comparison with other calibrations, where one would expect an overestimation because of stray light. It is not clear if the stated 0.7% UVB content in the UVA lamp was in fact corrected. As the 0.7 % UVB is not accurately defined and as it determines in large part the effective dose of the ‘UVA’ source, its spectral accuracy is crucial for the ‘UVA’ exposure. 3. The authors wrote that no increase in 25(OH)D level occurred as a result of a short UVA exposure. However, the chosen UVA dose was insufficient to detect a significant 25(OH)D increase in combination with dominating UVB in a short exposure time. The dose in the long exposure from the ‘pure’ UVA source was greater than 10-fold compared with the UVB and UVAB doses, and led to a mean increase of 4.7 nmol/L in 25(OH)D in the group with the smallest number of subjects and the highest proportion © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd doi:10.1111/phpp.12122
Commentary
with skin types V and VI. A proportionally smaller change from the short UVA exposure would not be apparent within the much larger increases in 25(OH)D that followed irradiation with UVB (mean increase 11.6 nmol/L) and UVAB (mean increase 13.6 nmol/L); the effect of the added small UVA exposure would fall well within the error range. Furthermore, the authors suggested that some photodegradation could occur following longer exposure to UVA. However, leaving aside uncertainty about the ‘UVA’ dosimetry, no experiments were included in the study to substantiate this suggestion, which was apparently based solely on the validity of spectral weighting with the CIE vitamin D action spectrum (and only seven type I–III individuals in the UVA group). 4. The authors proposed that the negative correlation they found (like many others before them) between 25(OH)D levels before (at t0) and the increase after exposures (at t2) could be ‘caused by the regression to the mean’. To substantiate this, they calculated an alternative correlation – 25(OH)D increase vs. mean value before at t0 and after at t2. Not surprisingly, this lowered the correlation coefficient to the point of losing significance: the mean value narrows the range of levels, given that the largest differences occur from the lowest initial levels, and thus any correlation will decrease. Regression of extremes to the mean is inherent to random variation from sample to sample. If it is strong and dominant, one can assume that ranking individuals according to levels before, at t0, has no significant correlation with ranking them according to levels after exposure at t2. Thus, a rank test would seem more interesting, appropriate, and robust than the statistical methods used in the study. We tried this rank testing on data from a mid-winter 8-week course of either sunbed use (sub-sunburn doses 3×/week) or vitamin D
supplementation (1000 IU/day) (2). Interestingly, we indeed found that ranking before and after sunbed use was not significantly correlated (with n = 35 Kendall’s τ = 0.22, P = 0.067). However, ranking before and after vitamin D supplementation was significantly correlated (with n = 36 Kendall’s τ = 0.35, P = 0.002). Thus, the authors may be correct in proposing that randomization in increases in 25(OH)D levels after UV exposures is greater (i.e. more variability in responses) than after vitamin D supplementation (although in this 8-week midwinter experiment, the irradiation scheme was controlled by subjects themselves according to their skin responses, adding a source of variability in the treatment not present in vitamin D supplementation and in the irradiation protocol of Sallander et al.). The authors concluded that when UVB irradiation was short and intense, the UVA dose delivered in an equally short time did not matter and did not influence the vitamin D level. However, as we point out, the results in their study were based on a comparison of UVB and UVAB exposures that may not be well founded. The additivity of UVB and UVA in increasing 25(OH)D levels was not tested. Instead, and assuming that their dosimetry was correct, the authors showed that spectral weighting with the CIE vitamin D action spectrum was incorrect in predicting increases in 25(OH)D from equally ‘vitamin D-effective doses’ from UVB- vs. UVA-dominated sources (their Table 4). Suggesting that this discrepancy was due to photodegradation of vitamin D by UVA radiation is not substantiated by their data. Thus, overall, the conclusions drawn by Sallander et al. may not be justified, given the uncertainties in the dosimetry and inadequacies in the experimental design. Further detailed investigations are merited in this important and topical area.
REFERENCES 1. Sallander E, Wester U, Bengtsson E, Wiegleb Edstrom D. Vitamin D levels after UVB radiation: effects by UVA additions in a randomized controlled trial. Photodermatol Photoimmunol Photomed 2013; 29: 323–329.
2. de Gruijl FR, Pavel S. The effects of a midwinter 8-week course of sub-sunburn sunbed exposures on tanning, vitamin D status and colds. Photochem Photobiol Sci 2012; 11: 1848–1854.
Photodermatol Photoimmunol Photomed 2014; 30: 176–177 © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
177
