Labelled sunscreen SPFs may overestimate protection in natural sunlight

Brian Diffey*a and Uli Osterwalderb
aDermatological Sciences, Institute of Cellular Medicine, University of Newcastle, UK. E-mail:; Tel: +44 (0)1395 519378
bDSM Nutritional Products, PO Box 2676, 4002 Basel, Switzerland

Received 10th July 2017 , Accepted 10th August 2017

First published on 11th August 2017

Limited evidence exists to indicate that sunscreen protection factors determined in the laboratory are higher than those in natural sunlight. In this article we propose an explanation for this difference and estimate the expected SPFs of sunscreen products in natural sunlight and those expected from laboratory testing. Our results indicate that the labelled SPF, determined by in vivo assay using a UV solar simulator, overestimates the SPF that would be expected in natural sunlight to the extent that for products labelled SPF50+, it may not be possible to achieve a protection against sunlight of more than 25-fold. The popular interpretation of the SPF that it can be thought of as how much longer skin covered with sunscreen takes to burn in sunlight compared with unprotected skin, can no longer be defended.


The concept of the Sun Protection Factor (SPF) was popularised by the Austrian scientist Franz Greiter in the 1970s and subsequently adopted by many regulatory authorities and the cosmetic and pharmaceutical industries as a measure of the protection provided by topical sunscreens against sunburn.

There is often a perception that topical sunscreens do not provide the degree of protection in natural sunlight that might be expected from the labelled SPF. This is partly accounted by variations in application thickness and uniformity of spreading coupled with removal during the exposure period due to factors such as sweating and rubbing.1

Another factor that can contribute to this mismatch between expected and delivered protection is the difference between the spectral emission of UV solar simulators that are used in the laboratory determination of SPF and the spectral power of terrestrial sunlight.

There are a few reports2,3 indicating that SPFs determined using UV solar simulating radiation in the laboratory are higher than those measured using natural sunlight as the source, but the data obtained, especially on modern sunscreens with SPFs of 15 and higher, are inadequate to give a clear indication of the magnitude of the difference.

This is not surprising since determining SPFs using natural sunlight is problematic. With high factor sunscreens, in particular, the exposure periods required would be an hour or more. During this time not only does the spectral irradiance of sunlight change on a minute-to-minute basis as the solar altitude, and possibly cloud cover, changes, but perturbations in air temperature, humidity and subject movement, for example, will also act as confounding factors. So while outdoor testing under actual use conditions might be considered the most realistic method, it is both impractical and unreliable.

In this paper we propose an explanation for the disparity between SPFs determined in the laboratory and those expected in natural sunlight, and estimate the degree to which sunscreens applied in natural sunlight may be underperforming in respect of the labelled SPF.


The ambient spectral irradiance of clear sky, midday summer sunlight at mid-latitudes (around 40 degrees), calculated using a computer model of solar spectral irradiance,4 is shown in Fig. 1. Whilst it is possible to configure the spectral output of a xenon arc lamp to provide a close match to this spectrum over much of the ultraviolet, visible and infrared wavebands, the output power of the lamp would need to be limited so as not to exceed the small area heat pain threshold,5 which would necessitate exposure times of several minutes in order to elicit delayed erythema.
image file: c7pp00260b-f1.tif
Fig. 1 The UV spectral irradiance of clear sky, midday summer sunlight at mid-latitudes (blue) and an Oriel® Sol-UV Series UV solar simulator (red).

To overcome this limitation, solar simulators designed specifically for SPF determination incorporate a UV-transmitting, visible-light absorbing filter in the output beam in order to remove the visible light component and so allow much higher irradiances in the UV waveband compared with natural sunlight.6 Since wavelengths in the UVB and short-wave UVA (UVAII) region are the most effective at inducing erythema, the design goal is to match the relative shape of the spectral output of a UV solar simulator as closely as possible to sunlight in this region.

The spectral output of an Oriel® Sol-UV Series UV Solar Simulator is compared with summer sunlight in Fig. 1, where we see that, unlike sunlight, the spectral output of the UV solar simulator falls rapidly at wavelengths beyond 370 nm and emits negligible radiation beyond 400 nm. The spectral irradiance of the UV solar simulator shown in this figure is much greater in the UV region than is sunlight, as reflected by the scaling on the respective ordinates, resulting in a typical erythemal irradiance at the skin surface of about 90 SED h−1, compared with mid-latitude summer sunlight, which has a maximum erythemal irradiance around midday of 7–9 SED h−1. This means that whilst it takes about 20–25 minutes of sun exposure on an unshaded, horizontal surface to receive sufficient solar UV to elicit a minimal erythema in unacclimatised white skin a few hours later, the same reaction can be achieved with the UV solar simulator in about 2 minutes. The total unweighted irradiance from the UV solar simulator is about 0.35 kW m−2, which is around one-third of the full spectrum irradiance from sunlight.

In order to know whether the removal of visible light from UV solar simulators is of concern, we need to consider the erythema response of the skin to different wavelengths of UV and visible radiation.

The erythema action spectrum shown in Fig. 2 was obtained from an experimental study using a 5000 W Xe–Hg compact arc lamp optically coupled to a holographic grating monochromator in which the minimal erythema dose (MED) was determined at several wavelengths between 250 and 435 nm at 24 h after exposure.7 Erythema was not observed at the highest exposure dose used at 435 nm and so the relative sensitivity lies below the data point, as indicated by the arrow in Fig. 2. This does not mean that skin will not develop delayed erythema in response to visible light, simply that the exposure dose given was insufficient to induce a reaction.

image file: c7pp00260b-f2.tif
Fig. 2 The erythema action spectrum7 extrapolated into the visible region. Erythema was not observed at the highest exposure dose used at 435 nm and so the relative sensitivity lies below the data point, as indicated by the arrow. The half-value wavelengths of the upper and lower broken grey lines were 200 nm and 10 nm, respectively, with the central broken red line having a half-value wavelength of 54 nm.

In this analysis, we postulate that erythema induction does not cease at around 400 nm but continues at wavelengths into the visible region, albeit with an ever-decreasing effectiveness as wavelength increases. As there are no data on the action spectrum for delayed erythema beyond 405 nm, the best we can do is to extrapolate the curve on the same downward slope as between 380 and 405 nm, shown by the central broken red line in Fig. 2.

The SPF of a sunscreen may be calculated as:

image file: c7pp00260b-t1.tif

E(λ) is the spectral power distribution at wavelength λ nm of the radiation source used in the determination, S(λ) is the erythema action spectrum shown in Fig. 2 and T(λ) is the spectral transmission through the sunscreen. The upper limit of the summation, λmax, is, theoretically, the maximum wavelength in terrestrial sunlight but it can be truncated at 700 nm without affecting the accuracy of the SPF determination.

By means of a commonly-used sunscreen simulator,8,9 we consider two sunscreens that we calculated to result in an SPF of 30 when the radiation source is a UV solar simulator shown by the red curve in Fig. 1. Product A contains active ingredients (6% benzophenone-3, 7.5% ethylhexyl methoxycinnamate, 4% phenylbenzimidazole sulfonic acid) that absorb mainly UVB radiation and is typical of sunscreens used in the 1980s and 1990s. Product B (active ingredients: 3% butyl methoxydibenzoylmethane, 5% octocrylene, 4% methylene bis-benzotriazolyl tetramethylbutylphenol, 2.5% phenylbenzimidazol sulfonic acid) is a modern, broad-spectrum sunscreen that provides balanced protection across the UV spectrum.

The monochromatic protection factors (mPF), which show the protection on a wavelength-by-wavelength basis throughout the UV and visible regions, are shown in Fig. 3. At a given wavelength, λ, the mPF is equal to the reciprocal of the transmission, T(λ). For product A, the mPFs are effectively one (i.e. no protection) at all wavelengths in the visible region, whereas product B exhibits weak protection at wavelengths in the blue and blue/green regions. This visible light absorption is attributed to the active ingredient methylene bis-benzotriazolyl tetramethylbutylphenol that is present in this product.

image file: c7pp00260b-f3.tif
Fig. 3 The monochromatic protection factors (mPF) of the two SPF30 sunscreens (product A – red curve; product B – blue curve).


From the equation above, the SPF expected from a UV solar simulator for either product is calculated to be 30 and remains unchanged no matter what wavelength we choose for λmax above 400 nm, the reason being that the UV solar simulator has negligible output at wavelengths above 400 nm.

If now, we use the spectrum of natural sunlight (black curve, Fig. 1) in the expression and summate up to 700 nm we find the SPF for product B has fallen from 30 to 15, since natural sunlight has much greater spectral output in the visible region compared with the UV region and this compensates to some extent for the falling erythemal sensitivity of skin in the visible region, which we have assumed follows the broken red line in Fig. 2. For product A, the expected SPF in sunlight is even lower at 11.

By changing the concentrations of the active ingredients appropriately, we can estimate SPFs that would be expected from the UV solar simulator and the corresponding SPFs expected in sunlight. Note that because the absorption of the active UV filters in product A are limited in the UVA region, it is not possible to achieve SPFs higher than 30 at allowable concentrations of the active ingredients used in these products.

We see from Fig. 4 that the mismatch between laboratory determination of SPF using a UV solar simulator and that expected in natural sunlight increases as the SPF increases such that for products labelled SPF50+, it is not possible to achieve a protection against solar radiation of more than 25-fold. Also, for a given labelled SPF, the degree of mismatch is greater for sunscreens, such as product A, that provide sub-optimal broad spectrum protection.

image file: c7pp00260b-f4.tif
Fig. 4 The variation of SPF expected in natural sunlight with laboratory determined (or labelled) SPF using a UV solar simulator for product A (red curve) and product B (blue curve).

Uncertainty analysis

The solar spectrum we chose is representative of midday mid-summer sunlight under clear skies at latitudes around 40° when the solar altitude is 70°–75°. Furthermore, we assume that the erythema action spectrum falls through the visible spectrum with the same downward slope as between 380 and 405 nm. This slope is equivalent to erythemal sensitivity reducing by a factor of two every 54 nm; we term this fall in sensitivity the half-value wavelength (HVW).

We repeated our calculations using a clear sky solar spectral irradiance4 for solar altitudes of 30°, 60° and 90°, keeping the ozone layer thickness fixed at 3.2 mm. We first postulated that the erythemal sensitivity fell rapidly at wavelengths beyond 405 nm (the longest wavelength where sensitivity could be measured7) with a slope equivalent to a HVW of 10 nm, shown by the lower grey broken curve in Fig. 2. We then repeated our calculations by postulating that erythemal sensitivity fell very slowly throughout the visible spectrum with a slope equivalent to a HVW of 200 nm (upper grey broken curve in Fig. 2). In every case, we calculated the expected SPF in sunlight for product B when the SPF determined in the laboratory was calculated to be 30. The results are shown in Table 1.

Table 1 Expected SPF in sunlight for a broad spectrum sunscreen having a nominal SPF of 30
Half-value wavelength Solar altitude
90° 60° 30°
10 nm 25.0 23.8 18.7
54 nm 15.8 14.1 8.8
200 nm 9.5 8.2 4.7

We see that as the sun falls lower in the sky (decreasing solar altitude) the expected SPF in sunlight also falls, since the UVB component reduces relative to the long wave UV and visible components. At low solar altitudes around 30°, the SPF may appear alarmingly low compared with the labelled value, but the UV index at this solar altitude is only about 1–2 and so there is little risk to health and hence the need for sunscreen.

A further observation is that the more slowly the erythemal sensitivity falls through the visible region, i.e. the greater the half-value wavelength, the less effective is the sunscreen at protecting against sunburn.


The fundamental premise of our argument is that visible light, present in natural sunlight but not in UV solar simulators, contributes towards a sunburn reaction. There are very few experimental studies on cutaneous responses to visible light10 and although broad band visible light has been shown to produce immediate erythema that fades within 2 hours or so and thought to be due to a thermal, and not photochemical, response,10 there are no data on the induction of delayed (24 h) erythema in normal skin following visible light irradiation.

Using a light source emitting 98.3% visible light, 1.5% infrared radiation, and 0.19% UVAI, Mahmoud et al.11 observed no delayed erythema in subjects of skin type II at the highest dose used of 480 J cm−2. However, by combining the spectral emission of their light source (Fig. 1 in their paper11) with the action spectrum for erythema extrapolated through the visible region (Fig. 2), we calculate that in order to observe a minimal erythema at 24 h, an exposure dose of 1860 J cm−2 would be necessary, that is, almost 4-fold higher than the maximum dose used Mahmoud et al.11

There are no published data on the cutaneous impact of a dose of visible light of this magnitude and so it is not surprising that delayed erythema in normal skin has not been demonstrated by exposure to visible light. However, in the photosensitivity disorder chronic actinic dermatitis (CAD), erythemal responses at 24 h following exposure with an irradiation monochromator have been observed at wavelengths in the visible region up to 600 nm.12,13

Furthermore the action spectrum in the UV region for induction of an erythemal response at 24 h in CAD has been shown to be the same shape as that for normal sunburn in fair-skinned individuals.14 This suggests that an endogenous chromophore(s), responsible for initiation of the abnormal reaction to light in CAD may be the same as, or similar to, that/those responsible for erythema in normal skin. We suggest, therefore, that, at sufficient exposure doses, radiation at different wavelengths throughout the visible region is capable of inducing delayed erythema. Whilst this postulate does not confirm our projected action spectrum (Fig. 2), there are no observational data that refute it.

On the basis of the extrapolated action spectrum shown by the broken red line in Fig. 2, we estimate that for clear sky, midday summer sunlight at mid-latitudes, the contribution to an erythemal reaction from UVB (290–320 nm), UVA (320–400 nm) and visible light (400–780 nm) is about 78%, 17% and 5%, respectively.

We show in this analysis that SPFs expected in sunlight are lower than those estimated using a laboratory UV solar simulator, which we hypothesize is because of the, albeit very low, erythemal sensitivity of the skin to visible light present in sunlight but not in UV solar simulators used for SPF assay. There are no data on how the erythemal sensitivity of the skin changes with wavelength in the visible region and in the absence of this information, we defend our extrapolation. However, as we show in Table 1, should the wavelength sensitivity differ from our extrapolated curve, so will the calculated SPFs expected in sunlight, although they will always be below the values determined using a UV solar simulator.

An important observation from Fig. 4 is the non-linear relationship between laboratory (or labelled) SPFs and the SPF expected in sunlight. For example, a broad spectrum sunscreen, like product B, may be available as SPF15 and SPF30, i.e. a factor of two difference in protection. But the expected SPFs in sunlight for these two products are SPF10 and SPF15, respectively, i.e. a factor of 1.5 difference in protection.

Failure in reciprocity is another factor that might be proposed to explain the lower SPF values observed in sunlight, due to its greater total irradiance compared with a UV solar simulator. However, we do not believe that this is likely. A summary of reciprocity experiments carried out in human and mouse skin were reviewed by Martin et al.15 and in every case, reciprocity for erythema was shown to hold. Of particular relevance to solar simulators and sunlight, where exposure times to induce minimal erythema will vary between a few minutes with a UV solar simulator but up to half-an-hour or more for sunlight, Meanwell & Diffey16 showed that exposure to polychromatic radiation from a xenon arc lamp for time periods ranging from 1-second to 1-hour induced degrees of delayed erythema ranging from minimal to marked that depended only on dose and not dose rate.

This study has demonstrated that labelled SPFs may not reflect the magnitude of protection expected in sunlight and that the popular interpretation of the SPF that it can be thought of as how much longer skin covered with sunscreen takes to burn in sunlight compared with unprotected skin, can no longer be sustained.

Conflicts of interest

There are no conflicts of interest to declare.


  1. B. L. Diffey, Photodermatol., Photoimmunol. Photomed., 2009, 25, 233–236 CrossRef PubMed.
  2. R. Sayre, D. Desrochers, E. Marlowe and F. Urbach, Arch. Dermatol., 1978, 114, 1649–1651 CrossRef CAS.
  3. D. Lott, Cosmet. Toiletries Mag., 2013, 128, 638–647 Search PubMed.
  4. B. L. Diffey, Photochem. Photobiol., 2015, 91, 553–557 CrossRef CAS PubMed.
  5. H. Bücker, Strahlentherapie, 1960, 111, 404 Search PubMed.
  6. D. S. Berger, J. Invest. Dermatol., 1969, 53, 192–199 CrossRef CAS PubMed.
  7. J. A. Parrish, K. F. Jaenicke and R. R. Anderson, Photochem. Photobiol., 1982, 36, 187–191 CrossRef CAS PubMed.
  8. BASF Sunscreen Simulator. Available at: (accessed 30 June 2017).
  9. B. Herzog and U. Osterwalder, Pure Appl. Chem., 2015, 87, 937–951 CrossRef CAS.
  10. L. R. Sklar, F. Almutawa, H. W. Lim and I. Hamzavi, Photochem. Photobiol. Sci., 2013, 12, 54–64 CAS.
  11. B. H. Mahmoud, E. Ruvolo, C. L. Hexsel, Y. Liu, M. R. Owen, N. Kollias, H. W. Lim and I. H. Hamzavi, J. Invest. Dermatol., 2010, 130, 2092–2097 CrossRef CAS PubMed.
  12. W. Frain-Bell, Clin. Exp. Dermatol., 1979, 4, 1–29 CrossRef CAS PubMed.
  13. I. A. Magnus, Dermatological Photobiology, Blackwell Scientific Publications, Oxford, 1976, p. 191 Search PubMed.
  14. H. du P. Menagé, G. I. Harrison, C. S. Potten, A. R. Young and J. L. M. Hawk, Photochem. Photobiol., 1995, 62, 976–979 CrossRef.
  15. J. W. Martin, J. W. Chin and T. Nguyen, Prog. Org. Coat., 2003, 47, 292–311 CrossRef CAS.
  16. E. F. Meanwell and B. L. Diffey, Photodermatology, 1989, 6, 146–148 CAS.

This journal is © The Royal Society of Chemistry and Owner Societies 2017