Phuroihtvriirrri ond Phurohiolngv. Vol. 29. pp. 559.566, 003 I -8655’79/0301-0559802.

00/0
0 Pergsrnon Prrcs Ltd. 1979. Pnnied in Grcai Brilsm

A COMPARISON OF I N V I V O AND I N VITRO

TESTING OF SUNSCREENING FORMULAS
ROBERTM. SAYRE,*PATRICIA
POHAGIN, GORDON
J. LEVEEand EDWARDMARUIWE
Dept. of Photobiology Research, Schering-Plough, Inc., 3030 Jackson Avenue, Memphis, T N 38 151, U.S.A.

(Received 22 March 1978; accepted 4 August 1978)

Abstract-Seven commercially available sunscreens were compared by three different methods. Absor-
bance spectra were measured for each product in isopropanol solution and also on hairless mouse
epidermis. In uiuo tests were performed on human volunteers using a Xe arc solar simulator. Sun
Protection Factors (SPF) were calculated by each method for each product tested and the results
compared. By all methods used, the combination of 7% octyl dimethyl para-aminobenzoic acid and
3% oxybenzone provided the most protection from U.V.light. While estimates of the effectiveness of
all products were much too high when calculated by the isopropanol solution method, the hairless
mouse epidermis technique seems to be an accurate tool for predicting product efficacy in uiuo.

INTRODUCTlON on the solution absorbance data alone. Freeman et

Determining the Sun Protection Factor (SPF) is the al. (1966) have used this procedure to predict different
preferred technique for estimating sunscreen efficacy erythemic potentials for different solar angles. They
on human subjects, whether measured outdoors in have shown that the erythemic contribution and the
natural sunlight or indoors using a Xe arc solar shape of the convolution integral changes with the
simulator (Willis and Kligman, 1970; Cripps, 1974; solar angle.
Sayre et a/., 1978). The Protection Factor is a bio- By using the dilute-solution method, Kreps
logical measurement of the transmission of erythem- (1963a,b) reported that a product containing 104 gly-
ically effective light which can be expressed as S P F = ceryl dimethyl p-aminobenzoic acid ester would
l/Te~fectivc.This is particularly useful in comparing theoretically have an SPF of 10, while one containing
test results generated by different testing systems. By 2% of that ingredient would have an SPF of 100.
using this formula. any measurement of light In practice such high protection factors are not
transmission should be directly convertible into SPF observed (Cripps, 1974; Willis and Kligman, 1970;
units. Because outdoor human testing requires much Sayre et al., 1978). Work by Kreps and by others
time, many personnel, careful planning and good (Cumpelik, 1972; Kahn et al., 1969) suggests that
weather conditions, many investigators have sought almost all commercially available sunscreening
alternative and faster methods to predict the perform- products should be total blocks of sunlight, consider-
ance of sunscreens. By studying the transmission of ing the amounts of sunscreen ingredients they con-
U.V. light by in uifro spectroscopic techniques, they tain. Based on their studies, only a fraction of the
have hoped to use the data obtained to effectively amounts now used should result in highly efficacious
estimate sunscreen ingredient and product efficacy. products. None of these authors have attempted to
A number of in uitro testing procedures have been compare their predictions with in uiuo product tests,
suggested (Kahn and Wilcox, 1969; Kahn et al., 1969; however. Recent studies in our laboratories have
Kreps, 1955; Master et al., 1966; Groves, 1975) for shown that products containing from 3 to 7% of
estimating the human efficacy of sunscreening p-aminobenzoic acid esters result in S P F values of
products. Two major types of in uitro techniques are less than 10 (Sayre et al., 1978). It addition, solution
now in use. The most widely used type of in vitro methods are unable to demonstrate the contribution
test is the determination of the absorption character- to product efficacy provided by the presence of physi-
istics of sunscreening agents based on spectrophoto- cal sunscreens such as talc, titanium dioxide and zinc
metric analysis of dilute solutions. Both absorption oxide, because such ingredients are not soluble in sol-
and transmission curves have appeared in the litera- vents suitable for spectral analysis (Master et a[.,
ture to support claims of product eficacy (Kreps, 1966).
1955, 1963b; Robertson and Groves, 1972). The most The second major method involves the examination
elegant presentation of these techniques was made by of absorption or transmission data derived from thin
Kreps (1955) and more recently has been used by film sandwiches of ingredients or products between
Cumpelik (1972). These investigators have used the quartz plates (Master et al., 1966; Groves, 1973).
solution method to predict sunscreen efficacy based Robertson and Groves (1972) have modified earlier
attempts to use this thin film method by using the
*To whom correspondence should be addressed. Lambert-Beer Law to relate data obtained on thick
PAP 29/3- H 559
560 RORERTM. SAYRE, POH AGIN. GORIWN
PATRICIA J. LEVEEand EDWARD
MARLOWE

films t o thinner films as would actually be used on strument simultaneously. All exposures in this study were
the skin. Theoretically, both of these methods should based on a geometric progression each 25‘?; greater than
the previous one. i.e.
be valid, assuming that absorption of UV light by the
sunscreen product while in actual use does follow the TI = 1.25 To
Lambert-Beer relationship and that the interaction Tn= 1.25 T, _ I
of the product with human skin is negligible. These The advantages of such incremental exposures are de-
assumptions, however, are not entirely correct as this scribed by Hoppe er al. (1975) and by Van der leun (1966).
Products* tested in the stud!):
study will show. (A) 72, octyl dimethyl p-aminobenzoic acid ester, 30/,,
Neither of the currently used iri vitro test systems oxybenzone;
are able accurately to predict the efficacy of sun- (B) 100/6 2-hydroxy-4-methoxybenzophenone-5-sulfonic
screening ingredients in products designed t o prevent acid ;
(C) 3”/a octyl dimethyl p-aminobenzoic acid ester;
sunburn. Both methods ignore skin surface phenom- 3% glyceryl p-aminobenzoic acid ester;
ena as well as skin-vehicle interactions, important (D) 3.304, octyl dimethyl p-aminobenzoic acid ester:
considerations in sunscreen testing. (E) 4% amyl dimethyl p-aminobenzoic acid ester;
In the current study we have used the intensity (F) 3% oxybenzone;
spectrum of the solar simulator to predict sunscreen 3% dioxybenzone;
(G) 5% p-aminobenzoic acid.
efficacy from in uitro data in the same manner in Evaluation of sunscreen products in solution. Each
which Freeman et a/. (1966) used the solar spectrum. product was diluted in isopropanol to 2 pl/ml. Isopropanol
A similar approach has been used by Cripps et al. was used as the reference solution. Scans of the sunscreen
(1974) in tests of product efficacy a t 305nm with product in solution were run from 450 to 250nm using
monochromatic radiation. In this study the authors 0.1 cm quartz cuvettes in a Beckman Acta MVI spectro-
photometer. The absorption data obtained were then multi-
assumed that contributions to product effectiveness plied by 10 to correct for the short path length used, result-
and the erythemic potential outside this wavelength ing in data equivalent to 2 pl of product per cubic centi-
were negligible. The work of Freeman et al. (1966) meter of volume.
as well as the work presented here indicates that this In vivo human sunscreen testing. For each product tested,
8 to 15 healthy volunteers were selected who were free
is not the case. of any conditions that might abnormally affect test results.
The methods introduced in the current study make Informed consent was obtained from each volunteer before
it possible quickly to examine the potential efficacy beginning the sunscreen test. The methods used in this
of sunscreen products and ingredients. The first tech- study follow those previously described (Sayre et d., 1978).
nique employs human volunteers to determine the Two tesr sites which were uniform in pigmentation and
free of any observable defects were selected on the back
products’ Sun Protection Factors indoors at 2 pf/cmZ of each volunteer. On the first site a series of five graded
with a Xe arc solar simulator. The second examines exposures measuring 1 x 1 cm each was administered with
the transmission of light through a solution in alcohol the Xe arc solar simulator. This graduated series of expc-
of the product a t a concentration equivalent to an sures given on untreated, unprotected skin was used to
determine the subject’s minimal erythema1 dose (MED).
application density of 2 p//cm2. The third involves The results of these exposures were evaluated the next day
the transmission of light by forward scattering of each to determine the lowest exposure which produced a mini-
product when applied a t 2 p f / c m 2 t o the epidermis mally perceptible redness.
of a hairless mouse. All of the results cin then be After the MED had been determined. an area 5 x 10 cm
was outlined in ink on the second test site. 100pl of the
compared directly not only because of the interrela-
product being tested was applied as uniformly as possible
tionship between the SPF and the transmission of by finger, resulting in an overall average application of
erythemically effective light, but also because in all 2 pf/cm2. The product was then allowed to dry for 15 min.
cases, identical amounts of product have been used After this period, five graded exposures were given using
providing a ready means to compare in vivo and in the solar simulator to determine an MED for the protected
area. These exposures were likewise read the next day and
uitro test data. an SPF calculated. A Sun Protection Factor (SPF) is the
ratio of the product-treated MED to the untreated (con-
MATERIALS AND METHODS trol) MED. SPF results for each product were then aver-
aged for all volunteers tested.
Solar sirnulator light source. The solar simulator used Hairless mouse studies. Female HRS/J hairless mice
in this study consisted of a 2500W Xe arc filtered by a (Jackson Laboratories, Bar Harbor, Maine) 5-6 weeks old,
dichroic mirror to remove visible and infra-red radiation were sacrificed by cervical dislocation and the skin
and by a secondary cut-off filter (1.0 mm WG-320) to shape removed from the dorsal area of the body. After immersion
the short wavelength portion of the spectrum similar to in 60°C water for 30s. the epidermis was removed intact
that of natural sunlight. Five independent electronically from the dermal layer by careful blunt dissection following
timed I cmz exposures can be administered with this in- the method of Blank (1957). The epidermal layer was then
floated onto a quartz carrier plate and the excess moisture
carefully removed. Each piece of epidermis used ranged
*(A) Improved Super Shade 15. Schering-Plough Corp; from 5 cm2 to over 10 cm’. Forward scattering scans were
(B) Uval Sunscreen Lotion, Dorsey Laboratories; (C) run from 400 to 250 nm in the Beckman Acta MVI record-
Eclipse Lotion, G. S. Herbert Laboratories; (D) Sun- ing spectrophotometer fitted with a diffuse reflectance
down Sunscreen, Johnson & Johnson; (E) Block-Out sphere. Additional 2mm UG-5 filters were used in both
Cream Lotion, Sea & Ski (Smith-Kline Corp.); (F) Solbar, the sample and the reference beams to remove visible fluor-
Person & Covey: (G) Presun Lotion, Westwood Pharma- escence from the epidermis and from some of the sunscreen
ceu t ica Is. formulas tested.
 Testing of sunscreening formulas 561

For each piece of epidermis examined. the first measure- Table 2. Sun Protection Factors using the Xe arc solar
ment obtained was a scan of the unprotected epidermis. simulator
The carrier plate with the epidermis was then removed
from the reflectance sphere and the epidermal area Number of Statistical
measured. The sunscreen to be tested was applied at a Product volunteers SPF range
concentration of 2 p//cmZ and the forward scattering scan
repeated for the epidermis with its applied product. To A 11 17.41 f 3.17 14.58-22.79
determine the absorption of the sunscreening product C 11 13.03 & 4.09 7.46-18.26
alone, the absorption of the untreated mouse epidermis G 10 12.01 k 3.11 7.48-18.23
was subtracted from the values for the product-treated epi- B 10 8.20 k 2.69 4.79-1 1.68
dermis point by point at 5nm intervals. E 10 8.13 f 2.42 5.97-1 1.67
For each formula tested, at least three separate scans D 10 6.70 f 1.65 4.78-9.34
were run on different epidermal sections. The absorption F 11 4.60 1.02 3.0&5.98
values were then averaged and the difference spectra plot-
ted so that the absorption spectra could be compared.
Predictiori of sunscreen efficacy ,from in vitro data. Theis shown and a statistical range given. The formulas
SPF of a sunscreen is equivalent to the reciprocal of the have been arranged in order of descending protection.
transmission of erythemic light by a sunscreen film, at leastThese results demonstrate the actual effectiveness of
in the monochromatic approximation. The model used in each sunscreen on human skin.
this study is based on the following equations:
After recording the spectra of each product in iso-
SPF = 1/T propanol at a concentration of 2 p’/m/, Figs. I A and
1B were prepared, exhibiting the absorption spectra
of the products from 400 to 250nm. Figure I A in-
cludes all of the products containing p-aminobenzoic
acid esters or combinations of these esters. From the
solution data, the combination of 776 octyl dimethyl
where
p-aminobenzoic acid and 3:4 oxybenzone (A) appears
EE(I) = erythemal efficiency spectrum to be the most effective at all wavelengths examined.
(Sayre et al 1966)
In Fig. IB, Product A is compared to the products
I(1.) = Solar simulator intensity spectrum as tested containing benzophenone derivatives as well as
measured with a calibrated spectro-
radiometer. the 5% p-aminobenzoic acid product (G). Product A
absorbs much better in the UVB range (290-320 nm)
EE(E.) x I(1) = 1.0(normalized) than any of these products. It is interesting to note
that in solution, the absorption peak of p-aminoben-
= 290 to 320 nm in 5 nm increments.
~(i) = 10-ab*(A)
zoic acid (G) occurs at 290 nm and falls off thereafter.
The benzophenone products exhibit 2 peaks, the first
where at about 285 nrn and a second smaller one at 330 nm.
abs(i) = The spectroradiometer measure of In solution, Product A absorbs more light at 285 nrn
sunscreen product absorbance. than either B or F. Only B absorbs more than A
Table I shows the normalized values of the product in the UVA wavelength area above 320 nm. The spec-
function used in these studies. As can be seen, a sunscreen tra of all the products presented suggest that A would
product which transmits all of the erythemic light must be the best product for sunburn protection as well
have an SPF of 1.0, while one which absorbs all of the as the best broad-spectrum product in this study.
light must have an SPF of infinity.
Apparent SPF values can be calculated from the
data obtained by the solution technique by using the
RESULTS AND DISCUSSION relationship SPF = t/Trrrcctivr. The results of these
Table 2 shows the results of identical human testi calculations are presented in Table 3, compared to
performed on each of the seven sunscreening formulas the human test results. In all cases the predicted pro-
examined in this study. For each, an average SPF tection offered by each product is in error by one
to several orders of magnitude when the same amount
Table 1. The normalized product function used in the cal- of product is tested. On the skin, the ingredients may
culation of SPF data. EE = erythemal efficiency spectrum; shield one another, may be absorbed into the skin,
I = solar simulator intensity spectrum
or may exhibit local concentrations different from
Wavelength EE x I (Normalized) that of the product in solution alone. The behavior
of the active ingredients and the vehicle on the skin
290 0.0 150 is quite different than that of the product in solution.
295 0.0817 Figure 2 demonstrates the method used to derive
300 0.2874
305 0.3278 a product absorption spectrum for a product alone
310 0.1864 when studied on hairless mouse epidermis. In Fig. 2A.
31s 0.0839 there are two spectra: (1) That of the mouse epi-
320 0.0 180 dermis, and (2) that of the mouse epidermis plus the
applied product. To obtain the “product alone” spec-
= 1.0000
trum, the absorbance values in Fig. 2A are subtracted
 VI
h)
m
UVA

I T
n

10

15

g 12
2
aa
$ 9
2

3 iI
0

250 270 290 310 330 350 370 3904

WAVELENGTH
NM

Figure tA. Comparison of the spectra of PABA-ester containing sunscreens at a dilution of 2pl/ml
in isopropanol.
Figure 1 B. Comparison of the spectra of the benzophenone-derivative containing sunscreens and
p-aminobenzoic acid (G) at a dilution of 2 pl/ml in isopropanol. Note the maximum absorbance peak
of G at 290nm.
 Testing of sunscreening formulas 563

Table 3. Comparison of the Sun Protection Factors pre- Although earlier attempts to correct some of the
dicted by the solution method to the SPFs obtained in problems in the thin film method were made (Master
human testing et al., 1966), this technique still has serious deficien-
SPF from SPF predicted Difference in cies. Riegelman and Penna (1960) realized that similar
human in orders of problems existed for the dilute-solution technique.
Product testing isopropanol magnitude To eliminate all visible fluorescence, UG-5 filters
were inserted in the spectrophotometer on both the
1.74 x 10' 3.9 x 10'8 lo1'
1.3 x 10' 1.4 x 109 1O8
sample and the reference sides for all of the hairless
1.2 x 10' 2.8 x 10' lo1 mouse studies. Without these filters, substantial fluor-
8.2 x 10' 8.0 x 10' 107 escence could be observed, as shown in Fig. 3. The
8.1 x I O O 7.9 x 108 1os spectrum of the mouse epidermis with and without
6.7 x 10' 6.5 x lo6 1O6 the filters is virtually unchanged, whereas the spectra
4.6 x 10' 6.6 x lo5 105
of the epidermis plus the product are very different.
Without the filters to remove visible fluorescence, a
point by point, resulting in the data in Fig. 2B. This much less accurate SPF would have been obtained
was done for each product, as described in Methods. for many of the products tested. Note the great differ-
In their use of the thin film technique, Robertson ence in the spectra in the UVB (29G-320) solar erythe-
and Groves (1972) neglected to account for fluor- ma1 range. These filters still allow some fluorescence
escence, with the result that many products appeared in the UVA range to come through, however, so that
to be many times more or less effective than can be the SPFs calculated for products or ingredients with
observed by human testing. The thin film method is intense UVA fluorescence might be expected to be
also subject to several other criticisms by its very somewhat variable. Also, inserting the filter into the
nature: (1) The uniformity of the film even when spectrophotometer causes a slight change in the spa-
applied by the most reliable' method is always ques- tial geometry of the instrument. Regardless of these
tionable: (2) the surface tension of the film may affect problems, the values obtained for predicted SPFs
the quartz plates and hence the film's thickness; (3) based on the hairless mouse data are reasonably close
no authors have claimed that the thin film in any to those obtained in human testing. Products exhibit-
way resembles the film when applied to human skin, ing fluorescence are A, C, E, and D, each of which
because penetration, evaporation, temperature and contains fluorescent PABA derivatives.
pH changes have not been taken into account. Table 4 presents the SPF values obtained by apply-

Figure 2. The product absorbance speotrum of A alone is shown on the right. This spectrum was
obtained by subtracting the absorbance of the hairless mouse epidermis alone from the spectrum
of the mouse epidermis with the applied product, as described in Methods.
5 64 ROBERTM. SAYKE,
PATRICIA
POHAGIN, GORDON
J. LEVEEand EDWAKD
MARLOWE

UVB
uvc cnnica

2.1

1.4

"
y1
14

Z
4
m
a 1.1
I r
0
v)
m
Q 0.1

0.1

I I l l
0.:

1 . -SKIN ALONE

t
I
-SKIN L UV FILTER

Figure 3. The fluorescence which interferes in the determination of accurate SPF values is demonstrated
above. Note that without the filter, E appears to absorb much less intensely in the UVB range than
the sample examined with the filter in place, while the spectrum of the skin alone remains unchanged.

ing a product to hairless mouse epidermis. In all exhibit maximal absorbance in the correct range, in
cases, the data obtained falls within the statistical both in uitro methods, they are much less absorbant
range given in Table 2 for the in uiuo (human) testing. than A at the same concentration. Additionally, A
Upon examining the product spectra obtained on also provides a better degree of UVA protection than
hairless mouse epidermis (Figs. 4A, B), it is evident any product except B in the 330-370nm range. B.
that A still provides the best broad-range protection however, does not absorb well in the critical "sun-
from uv light, both U V B and UVA. Figure 4A also burn'' wavelengths of 29&320 nm. It is important to
demonstrates a shift in the absorption spectrum peak remember that all of these products were applied
of G (p-aminobenzoic acid), from the peak shown in identically at 2 p//cm2 when making these compari-
solution (Fig. 1B). PABA has always performed much sons. From the in uiuo test results, A is the only
better in human testing than was predicted by itz vitro product which falls into the "ultra protective" range
methods (Willis and Kligman, 1970; Cripps, 1974). of SPFs as categorized by the Food and Drug Admin-
O n hairless mouse epidermis, we show one reason istration (1978).
for this occurrence. The true absorption peak can
now be seen as a broad peak between 295 and 320 nm.
Evidently, p-aminobenzoic acid on the skin is a much Table 4. Comparison of Sun Protection Factors obtained
by human testing vs that obtained on hairless mouse
better sunscreen than can be predicted from its solu- epidermis
tion spectrum. B, while providing good UVA absorp
tion, is still less protective in the critical UVB range Human Hairless mouse
than A. When compared to the other products con- Product testing skin
taining p-aminobenzoic acid esters, A again is the 17.41 3.17 22.8 + 10.1
most absorbant product for all of the UV wavelengths
(Figure 4B).
13.03
12.01
4.09
* 3.12.69
1
12.4
6.8 ** 9.3
1.5

*It*
To be effective in preventing sunburn and other 8.20 & 10.2 0.02
skin damage, a sunscreening product should have a 8.13 & 2.42 9.5 4.9
wide range of absorbance, which should peak in the
6.70 + 1.65 6.8 1.2
4.60 k 1.02 3.4 0.5
290-320nm wavelength area. While C, E and D do
 uvc UVB - UVA
7- -
HAIRLESS MOUSE SKIN
2lWCm’ SUNSCREEN
1.4 -

.-

c
-
270 310 330 350 370 390400
WAVELENGTH
NM
Figure 4A. By using the technique shown in Fig. 3, the spectra of the benzophenone-derivative sun-
screen products and that of PABA were determined on hairless mouse epidermis. Notice the change
in these spectra from that shown in Fig. lB, especially the shift in the maximum absorbance peak
of G (PABA) to 305nm.
Figure 4B. Shown above are the product spectra of the PABA-derivative containing sunscreen products
tested in this study. Compare these spectra to those shown in Fig. 1A. the isopropanol solution data.
In the wavelengths above 320 nm (UVA), A provides good protection while maintaining its maximum
peak of absorbancy at 305-310nm (UVB).
566 RORERTM. SAYRE.
PATRICIA
POHAGIN, GORDON
J. LEVEEand EDWARDMARLOWE

The combination of the p-aminobenzoic acid ester in the protection which an individual obtains when
(7:; octyl dimethyl p-aminobenzoic acid) and the ben- uiing a sunscreen formula outdoors. While outdoor
zophenone derivative (3% oxybenzone) accounts for sunscreen testing in actual use conditions is necessary
the broad spectrum protection offered by Product A. to assess the effectiveness of sunscreen formulas, in
It performs in a superior fashion to either family of uitro testing can aid the investigator in predicting
ingredients (PABA derivatives, benzophenones) when which sunscreen formulas will be of potential value
used alone, as seen in Fig. 4 and Tables 1 and 4. when in actual use. Unlike field studies, laboratory
This unique combination gives A characteristics of assays can be done under controlled conditions to
both classes of ingredients, in contrast to C or F, give reproducible data. The need for this has been
which each contain combinations of two ingredients presented in the OTC Review Panel's Monograph on
from the same chemical family. Product A exhibits Sunscreeens (1978). Work is continuing in our labora-
overlapping absorption qualities which give it greater tories to better simulate actual usage conditions in-
uv absorption derived from the two sunscreening in- doors.
gredients. From the data presented, it is apparent that the
The current study presents an in cirro method of hairless mouse epidermis method for predicting sun-
predicting in uiuo results under standard laboratory screen efficacy is a good one, closely approximating
conditions. Outdoors, many environmental factors the results obtained by in uiuo testing. Although some
(temperature, humidity. wind, sweating, etc.) are problems of fluorescence in the uv wavelengths still
known to be factors in overall sunscreen effectiveness need to be overcome, this method is superior to any
(Pathak et ul., 1969; Langner and Kligman, 1972; previously presented in virro technique for testing
Sayre et a/., 1978). The interaction of the vehicle with sunscreen effectiveness.
the skin and also with water plays an important part

REFERENCES
Blank, I. A., R. D. Griesemer and E. Could (1957) J . Invest. Dermatol. 29. 299-309.
Cripps, D. J . and S . Hegedus (1974) Arch. Dermarol. 109, 202-204.
Cumpelik. Boris M. (1972) J . Soc. Cosm. Chew. 23, 333-345.
Freeman, R. G., D. W. Owens. J. M. Knox and H. T. Hudson (1966) J . Inwst. Derniatol. 47, 586-592.
Groves, G. A. (1973) Aust. J . Derrnhtol. 14, 21-34.
Groves, G. A. (1975) Aust. Soc. Cosm. Chem. 90,36-50.
Hoppe, U., H. J. Kopplon and A. Wiskernann (1975) Arzneim. Forsch. 25, 817-825.
Kahn, G. and G. Wilcox (1969) J . Soc. Cosm. Chem. 20, 807-823.
Kahn, G., G. Wilcox, B. W. Wheeler and W. Atchison (1969) Br. J . Dermutol. 81, 731-736.
Kreps, Saul 1. (1955) Proc. Sci. Sect. Toilet Goods Assoc. 23, 13-26.
Kreps, Saul I. (1963a) J . Soc. Cosm. Chem. 14, 625-630.
Kreps, Saul I. (1963b) Am. Perf: Cosmet. 78, 73-76.
Kreps. S. I. and R. L. Goldernberg (1972) Cosmetics: Science und Technology (Edited by M. S. Balsam
and E. Sagarin), Vol. 1, pp. 241-305. Wiley, New York.
Langner, A. and A. M. Kligman (1972) Arch. Dermatol. 105, 851-855.
Master, K. J., R. M. Sayre. and M. A. Everett (1966) J . Soc. Cosm. Chern. 17, 581-594.
OTC Review Panel on Topical Analgesics (1978) Food and Drug Administration Monograph on
Sunscreens. Federal Register Part 2, 25 August, 1978.
Pathak. M A,. T. B. Fitzpatrick and E. Frenk (1969) N . Engl. J . Med. 280, 1459-1463.
Riegelman, S. and R. P. Penna (1960) J . Soc. Cosm. Chem. 11. 28&291.
Robertson, D. F. and G. A. Groves (1972) M e d . J. Aust. 2, 1445.
Sayre, Robert M., R. L. Olsen and M. A. Everett (1966) J . Inuesr. Dermatol. 46, 24&244.
Sayre. Robert M.. E. Marlowe. P. P. Agin, G. J. Levee and E. W. Rosenberg (1978) Arch. Dermufol.
In press.
Van der leun, J. C. (1966) Ultraviolet Erythema. Doctoral Dissertation. U. of Utrecht, Netherlands,
Willis, I. and A. M. Kligman (1970) Arch. Dermatol. 102, 405-417.