the prevalence and determinants of solar keratoses at a subtropical latitude (queensland, australia)

The prevalence and determinants of solar keratoses at asubtropical latitude (Queensland, Australia)

C.A.FROST, A.C.GREEN* AND G.M.WILLIAMS†Merck, Sharp and Dohme, P.O. Box 79, Granville, NSW 2142, Australia*Queensland Institute of Medical Research, Brisbane, Qld 4029, Australia†Australian Centre for International & Tropical Health & Nutrition, Brisbane, Qld 4029, Australia

Accepted for publication 17 July 1998

Summary We report the association between skin pigmentation and individual sun exposure, and theoccurrence of solar keratoses (SKs) in an unselected population, quantified for the first time. SKswere examined in a representative sample of 197 residents of the community of Nambour inQueensland, Australia. Estimates of sun exposure were combined with a measure of ultraviolet (UV)flux to estimate actual UV exposure, both occupational and recreational, during childhood and adultlife. The number of episodes of painful sunburn was used as a measure of intermittent, intense UVexposure. Eighty-three participants (43%) had at least one SK, while 35 (18%) had more than 10 SKsdiagnosed. The age- and sex-adjusted odds ratios (ORs) for the development of SKs were higher inindividuals with fair (OR ¼ 14·1) or medium skin (OR ¼ 6·5), compared with olive-skinnedindividuals. Individuals with poor ability to develop a suntan were similarly at increased riskcompared with others. High levels of occupational UV exposure during adult life were confirmed asbeing strongly associated with prevalent SKs (OR ¼ 2·4 for heavy/maximal adult exposure), with aneven stronger association seen in those individuals with multiple SKs (OR ¼ 4·3 for maximal adultexposure). Although no clear association was demonstrated between SK prevalence and accumu-lated childhood sun exposure, a history of even one episode of sunburn in childhood was stronglyassociated with SK prevalence (peak OR of 5·9 for one sunburn).

Solar keratoses (SKs) are epidermal dysplasias whichaffect a high percentage of pale-skinned individualsworld-wide, and together with basal cell carcinoma(BCC) and squamous cell carcinoma (SCC) of the skin,they constitute a growing public health problem inthese populations.1 SKs are among the strongest pre-dictors of skin cancer,1,2 and are generally believed to bethe precursors of SCC.3 Chronic exposure to solarultraviolet (UV) radiation, in particular the UVB com-ponent, and sun-sensitive cutaneous phenotypes, arebelieved to be the principal risk factors in their develop-ment.4–9 For example, people with the highest UVexposure based on their occupational and/or recrea-tional patterns are known to be at highest risk ofdeveloping SKs;9–13 and within white populationsliving in a sunny climate, SK prevalence rates of 40%or more are not uncommon.14 However, among the fewexisting studies there are methodological limitations,which make interpretation of their results difficult.14

Regarding the role of sun exposure, there are few dataabout the total dose of UV radiation required, or theinfluence of the time and pattern of exposure in the

development of SKs. In a review of previous studies ofsun exposure and skin cancer,6 the difficulties involvedin direct measurement of sun exposure were acknowl-edged. Most studies relied on simple recall measures,with only three studies attempting a quantitative esti-mate of total sun exposure. The most detailed determi-nation of UV exposure was conducted among Marylandwatermen;15 however, even in this study, no accountwas taken of either recreational or childhood exposure.The importance of childhood exposure has been studiedby comparing the rates of SKs between Australian-bornsubjects and British immigrants, with higher ratesbeing reported in those who were born in Australia,or who had migrated at an early age.7,11,16 However,age at arrival in Australia, and duration of exposure, areinextricably linked, and in these studies it is impossibleto ascertain how much of the effect of early age ofmigration is in fact due to long duration of exposure.17

The role of sunburn in the development of SKs hasalso been relatively neglected, especially with respect tothe age at which the sunburns occurred. In an Aus-tralian skin cancer prevalence survey,9 an association

British Journal of Dermatology 1998; 139: 1033–1039.

1033q 1998 British Association of Dermatologists

between sunburn and prevalent SKs was detected inparticipants who reported having six or more episodesof painful sunburn, but no consistent association wasdetected when fewer than six sunburns were reported.

In those individuals who are unable to avoid sunexposure, behavioural modifications to reduce sunexposure, such as the wearing of hats and long sleeves,and the application of a high protection sunscreen, arecurrently advocated. Reduction of SKs and of skincancer by these measures is also claimed, althoughonly limited evidence exists to show their effectivenessin preventing actinic lesions.18,19

To obtain more detailed information about the role ofsun exposure in the development of SKs and in order toclarify how this relationship is modified by skin type andthe wearing of protective clothing, a follow-up study(the Nambour SK study) was commenced in 1992 toinvestigate the natural history and determinants of SKs.In this report, the association between cutaneous pig-mentary factors and individual UV exposure and theoccurrence of SKs in an unselected population has beenquantified for the first time.

Materials and methods

Subjects

The Nambour SK study was undertaken in a subset ofparticipants in an Australian field trial of prevention ofskin cancer (the Nambour trial).20 All participants inthe Nambour trial had been randomly selected in 1986from residents of Nambour, a subtropical townshipwhich lies 100 km north of Brisbane, Queensland atlatitude 268S.1 The Nambour trial commenced in 1992,when 1626 adults aged between 25 and 75 years wererandomized by factorial design to receive (or not)sunscreen, and beta-carotene or placebo.20

Trial participants were eligible for inclusion in theNambour SK study if they lived within driving distanceof Brisbane. In mid-1992 an age-stratified randomsample of 201 subjects was selected for the SK studyand 200 agreed to take part. Of these, three subjectswith keratoses who had a diagnosis of disseminatedsuperficial actinic porokeratosis were excluded.

Data collection

At interview detailed information was gathered aboutspecific factors thought to be related to the developmentof SKs, including ethnic background, skin colour, abilityto tan, and number of episodes of painful sunburn

experienced both before and after the age of 20 years.A structured questionnaire was used to obtain informa-tion about the participants’ place of residence (town/city and country), and main occupations and activitiesundertaken during designated periods of life: ages 5–12, 13–19, and in subsequent decades. Using thisinformation as a cue, participants were then asked toestimate, for each period, the number of hours spent inthe sun between 10·00 and 14·00 h (the period of theday which accounts for , 66% of daily solar UVBradiation).21 Each person was also asked to estimatethe percentages of time they wore a hat, long sleeves, orused sunscreen while outdoors between 10·00 and14·00 h, before and after the age of 20 years.

All skin examinations of the head, neck, hands andforearms were conducted by a single medically quali-fied investigator (C.A.F.) who had received additionalclinical training in the diagnosis of SKs and skincancer. Subjects were usually visited at home duringdaylight hours to ensure adequate natural lighting forskin examinations (a Philips PL-C 10w/82 compactfluorescent globe was used to supplement naturallighting when necessary).

An SK was defined as a discrete, irregularly scaly(keratotic) lesion with or without erythema occurringon a background of solar-damaged skin. To assess thevalidity of clinical diagnoses of SKs in this study ahistological validation study was undertaken among22 affected study participants.

Data analysis

The number of hours spent in the sun between 10·00and 14·00 h on weekdays (or other routine or workingdays) was used to calculate: (i) the ‘everyday’ level ofexposure in early life (by combining the exposurereceived between the ages of 5 and 19 years); (ii) adultoccupational exposure (representing the exposurereceived from age 20 years onwards); and (iii) lifetimeoccupational exposure, being the summation of theprevious two variables. Similarly, the numbers ofhours spent in the sun between 10·00 and 14·00 hduring weekends and holidays were used to calculatechildhood, adult and lifetime recreational exposure.Total lifetime sun exposure was estimated by summingchildhood and adult sun exposure across weekdays,weekends and holidays.

To take into account the geographical differences inUV flux, latitudes of residence were determined from theresidential information, and a weighting variable cre-ated based on the world distribution of UV at wavelength

1034 C.A.FROST et al.

q 1998 British Association of Dermatologists, British Journal of Dermatology, 139, 1033–1039

307·5 nm.22 Estimates of occupational and recreationalUV exposure were calculated by combining the weight-ing variable for each period of residence with theestimated hours of daily sun exposure during thattime. Childhood UV exposure estimates were categorizedinto tertiles and adult and lifetime estimates into quar-tiles for analysis, with the category reflecting minimalexposure selected as the reference group.

The number of SKs diagnosed in the subjects atbaseline had a distribution that was highly skewed,and attempts to normalize the data by transformationwere unsuccessful. The outcome variable (number ofprevalent SKs at baseline) was therefore categorized.Analysis using ordinal logistic regression was found tobe inappropriate because a linear relationship did notexist between the outcome groups. Comparisons were

therefore made between subjects with no SKs and thosewith at least one SK (‘SK prevalence’); and those withno or few SKs (0–10) and those with many SKs (11 þ:‘SK burden’) in a series of logistic regression analyses.Maximum likelihood estimates of the odds ratios (ORs)and their 95% confidence intervals were calculatedusing EGRET.23 All risk factors were adjusted for ageand sex, and UV exposure variables were also adjustedfor tanning ability. To examine dose–response, tests fortrend were obtained by treating the independent vari-able of interest as a continuous variable.

Results

Validation study

Of 22 randomly chosen clinically diagnosed SKs (one oneach of 22 participants), 20 were confirmed afterindependent histological diagnosis by a single dermato-pathologist. Of the two lesions incorrectly clinicallydiagnosed as SKs, one was histologically identified as asuperficial BCC, and the other as an intraepidermalcarcinoma.

Prevalence of solar keratoses

Of 197 eligible subjects, 193 (98%) were interviewedand examined at baseline. SKs were diagnosed clinicallyin 83 study participants (43%), with the prevalence inmen being 52·4% and in women 36·0% (Fig. 1).Prevalence rose with age in both sexes, ranging from10·7% in women and 20·0% in men aged 30–39 years,

DETERMINANTS OF SOLAR KERATOSES 1035


100

80

60

40

20

030–39 40–49 50–59 60–69

Age (years)

Females Males

Pre

vale

nce (

%) 77.3

64.356.3

46.454.2

22.220.010.7

Figure 1. Age-specific prevalence of solar keratoses in Nambour, June1992.

Table 1. Association between cutaneous pigmentation and solar keratosis (SK) prevalence and SK burden in Nambour residents, June 1992

SK prevalence SK prevalence SK multiplicity

Risk factor Subjects Adj. ORa 95% CIb Adj. ORa 95% CIb Adj. ORa 95% CIb

Skin colour 193Olive 17 1·0Medium 50 6·5 1·2–34·7 1·0c 1·0d

Fair 126 14·1 2·9–69·6 3·3 1·6–6·6 10·3 2·8–38·3P for trend <0·001Tanning ability 193Only tan 24 1·0Burn then tan 129 3·1 1·0–9·4 1·0c 1·0d

Always burn 40 8·0 2·2–29·2 3·0 1·3–7·0 7·2 2·7–18·8P for trend <0·001

aAdjusted odds ratios (Adj. OR) adjusted for age group and sex; bCI, confidence interval; cReference category taken as the first two categoriescombined to allow comparison with subjects with SK multiplicity; dReference category taken as the first two categories combined due to smallnumber of subjects with baseline characteristic.

to 64·3% in women and 77·3% in men aged 60–69 years. Thirty-five individuals (18%) were diagnosedas having 11 or more prevalent SKs.

The categories representing the lowest degree of skincolour or tanning ability showed the strongest associa-tions with SK prevalence (Table 1). In addition, thetrends of increasing odds of SK prevalence with decreas-ing skin pigment were highly significant for both skincolour and tanning ability (P <0·001 in both cases).With respect to SK burden, no individual either witholive skin or who tanned without burning in response tothe sun was diagnosed with multiple SKs; therefore,reference groups were formed by combining these

groups with those with medium cutaneous pigmenta-tion. Similar patterns of association were seen, with thecalculated ORs being some two to three times higherthan those calculated for SK prevalence using the samereference categories. Ethnicity could not be consideredin the analyses as only five participants were of south-ern European ancestry (none of whom had SKs diag-nosed); all remaining participants were of British ornorthern European ancestry.

Sun exposure

Increasing levels of everyday exposure in childhood



SK prevalence SK multiplicitySubjects

Risk factor (n ¼ 193) Adj. ORa 95% CIb Adj. ORa 95% CIb

Occupational exposureChildhood

Minimal 55 1·0 1·0Moderate 76 0·73 0·32–1·64 0·81 0·28–2·4Heavy 62 1·14 0·49–2·7 0·60 0·19–1·9P for trend 0·75 0·37

AdultMinimal 48 1·0 1·0Moderate 46 1·42 0·49–4·1 2·1 0·42–11·0Heavy 50 2·4 0·87–6·9 2·4 0·45–12·7Maximal 49 2·4 0·73–7·8 4·3 0·77–24·3P for trend 0·08 0·09

LifetimeMinimal 49 1·0 1·0Moderate 48 0·95 0·34–2·7 2·0 0·42–9·9Heavy 47 2·8 1·02–7·7 3·2 0·63–16·3Maximal 49 1·84 0·60–5·7 4·4 0·84–22·7P for trend 0·08 0·06

Recreational exposureChildhood

Minimal 53 1·0 1·0Moderate 70 0·61 0·26–1·41 1·48 0·46–4·8Heavy 70 0·82 0·35–1·94 1·72 0·54–5·5P for trend 0·71 0·36

AdultMinimal 47 1·0 1·0Moderate 49 0·38 0·13–1·07 1·44 0·27–7·6Heavy 48 0·76 0·28–2·1 4·7 0·98–22·9Maximal 49 1·87 0·63–5·5 4·9 1·01–23·5P for trend 0·22 0·02

LifetimeMinimal 48 1·0 1·0Moderate 47 0·54 0·19–1·54 1·72 0·31–9·7Heavy 48 1·73 0·65–4·6 6·4 1·33–30·3Maximal 50 0·95 0·34–2·7 4·7 1·00–22·0P for trend 0·55 0·02

aAdjusted odds ratios (Adj. OR) adjusted for age group, sex and tanning ability; bCI, confidenceinterval.

Table 2. Association between ultravioletexposure variables and solar keratosis (SK)prevalence and SK burden in Nambourresidents, June 1992

showed no consistent association with SK prevalence(Table 2). In contrast, increasing levels of adult exposurewere positively associated with SK prevalence, withmaximum ORs of 2·4 in individuals classified ashaving heavy or maximal UV exposure. Combiningexposures received in childhood and adulthood tocreate a lifetime estimate did not add to the predictivevalue of adult exposure alone. With respect to SKburden, both adult and lifetime occupational exposuredemonstrated a stronger association with the diagnosisof multiple SKs than with the presence of any SKs, butagain with no statistically significant linear trend (Table2). However, unlike the plateau in ORs seen in the SKprevalence model, the ORs continued to rise withincreasing exposure for SK burden.

There was no consistent pattern of associationbetween recreational exposure at any age and SKprevalence (Table 2). In contrast, there was a signifi-cant trend of increasing risk of SK burden with increas-ing levels of both adult and lifetime exposure, with theORs peaking at 4·9 and 6·4, respectively. The numberof episodes of painful sunburn received prior to age 20was strongly associated with SK prevalence and SKburden (Table 3). Participants reporting even onepainful sunburn were approximately six times morelikely to have a prevalent SK than those reporting nosunburns. For sunburns received after age 20 theassociation was much weaker, more than five sun-burns being required to elevate the OR above 2·0.Adjustment for total childhood or lifetime exposureeither had little effect, or increased the risk (notshown), while adjustment for tanning ability reduced

the strength of the associations slightly, but did notalter their statistical significance.

Sun protection

No association was found between protective habitsreported before age 20 years and SK prevalence or SKburden, while after age 20 all sun protection practiceswere positively associated with the occurrence of SKs(not shown). For example, hat-wearing was stronglyassociated with SKs, with ORs ranging from 5·7 inparticipants wearing hats less than 50% of the time,to 10·0 and 8·3 in those wearing them more than 50%of the time, and all the time, respectively.

Discussion

The findings reported here demonstrate that both skincolour and tanning ability are important phenotypicdeterminants of SK prevalence. The very strong associa-tion detected between cutaneous phenotype and SKburden was not unexpected, reflecting the heightenedsusceptibility of this group to the effects of solar UVradiation.

One of the main aims of this study was to quantifythe association between sun exposure and the timingof that sun exposure and SKs. Calculated high levels ofUV exposure received during adult life or over anindividual’s entire lifetime were found to be stronglyassociated with SK prevalence. This supports thefindings of the study conducted by Vitasa et al.15

among commercial watermen in Maryland, in which



Table 3. Association between number of episodes of sunburn and solar keratosis (SK) prevalence and SK burden in Nambour residents, June 1992

SK prevalence SK multiplicity

Risk factor Subjects ORa ORb 95% CIc ORa ORb 95% CIc

Sunburns <20 years 1930 51 1·0 1·0 1·0 1·01 42 5·9 5·9 2·1–16·8 6·6 5·9 1·45–24·12–5 69 4·1 3·7 1·41–9·7 4·5 3·1 0·80–11·9>5 31 7·0 5·3 1·67–16·7 9·6 4·4 0·98–19·9P for trend <0·001 0·004 0·002 0·08

Sunburns $ 20 years 1930 89 1·0 1·0 1·0 1·01 29 1·15 1·10 0·41–2·9 1·2 1·22 0·34–4·42–5 58 1·84 1·56 0·68–3·6 1·87 1·08 0·35–3·3>5 17 3·9 3·6 1·03–13·0 2·8 2·3 0·55–9·2P for trend 0·02 0·05 0·08 0·4

a Odds ratios (OR) adjusted for age group and sex; b OR adjusted for age group, sex and tanning ability; cCI, confidence interval.

the risk of developing SKs was 1·5 times higher insubjects whose cumulative adult UVB exposureexceeded the median. In contrast, exposure in child-hood (with the exception of sunburns) showed aninconsistent association with both SK prevalenceand SK burden, contrary to a previous suggestionthat sun exposure in childhood is particularly impor-tant in determining the subsequent development ofSKs.16 This was based on the lower incidence of SKsseen in British people who migrated to Australia afterthe age of 20 compared with those who migratedbefore age 20. However, this could equally beexplained by the duration of exposure rather thanexposure occurring at an early age.

A secondary aim of the study was to determinewhether UV exposure influenced the number of SKswhich developed. A much stronger association wasfound between UV exposure and SK burden than SKprevalence, and a different dose–response effect wasgenerally demonstrated, with a levelling off in the riskfor SK prevalence after moderate exposure, but apersistent increase in risk for SK burden. These find-ings suggest that persons with large numbers of SKsare more susceptible to the mutagenic properties of UVradiation. It is also probable that individuals with SKsmay have reduced their recent sun exposure as aresult of medical advice, leaving those with the highestexposure at apparently lower risk. This trend was notevident in the very susceptible group with multipleSKs, however.

Patterns of exposure were investigated by separatingexposure accumulated during occupational and recrea-tional pursuits. Adult occupational exposure exhibited astrong association with SK prevalence, while both adultand lifetime occupational exposure were even morestrongly associated with SK burden. This is consistentwith the findings of other skin cancer surveys usingmore crude measures of UV exposure.6 Recreationalexposure per se was neither strongly nor consistentlyassociated with SK prevalence; however, adult and totallifetime recreational exposure did demonstrate a strongpositive association with SK burden.

Of great interest is the observation that while cumu-lative sun exposure in childhood was not clearly asso-ciated with SK occurrence, a history of any painfulsunburns in childhood was strongly associated withthe development of SKs. Similar findings have beenreported with respect to both BCC and SCC.24,25 Epi-sodes of sunburn in adult life were also consistently, butless strongly, associated with the occurrence of SKs, butlacked a clear association with SK burden. This may

relate to the adoption of sun-avoidance behaviour byindividuals with multiple SKs. Whether the higher risksassociated with sunburn in childhood are an indicationof an increased susceptibility of children to the effects ofintense sun exposure is not yet clear. However, it is notmerely a reflection of a higher total childhood or lifetimeexposure in these individuals, as adjustment for thesevariables either had little effect, or increased the risk.

Sunscreens, the wearing of hats, and other protectivemeasures are promoted widely with the rationale thatthey will reduce the amount of UV radiation reachingthe target cells in the skin, thereby reducing the devel-opment of actinic lesions. However, in the current study,most of the sun-protection practices under scrutinywere found to be associated with an increase in SKoccurrence rather than demonstrating the expectedprotective effect. Some of the excess risk demonstratedfor sunscreen use may be explained by the use of ‘suntanning lotions’ rather than effective sunscreens, whichcame into general use only during the last 20–30 years.The most likely explanation for these findings, however,is residual confounding by skin phenotype and sunexposure factors.26

These results should be interpreted bearing in mindthat everyday sun exposure in Queensland is ubiqui-tous, and therefore ‘forgettable’. In addition, the methodof measurement of sun exposure, although more com-plex than previously undertaken,6 is still a relativelycrude estimate of actual sun exposure. A degree of non-differential error would be expected, particularly assome participants were asked to recall exposures occur-ring up to 60 years previously. In addition, the Nambourcommunity has been taking part in skin cancerresearch for over a decade,9 and as skin cancer aware-ness campaigns are a regular feature in the Australianmedia, it is possible that those individuals with SKs mayhave recalled their past exposure differently from thoseindividuals without such lesions. However, the role ofoccupational vs. recreational, and childhood vs. adultexposure in the development of SKs is not well known,so it would be expected that these exposures would berecalled similarly. The finding of a positive associationfor adult occupational exposure, and the lack of anassociation between childhood or recreational expo-sures and SK prevalence, thus militate against differen-tial recall bias being a major limitation in interpretingthese data.

Information bias is another potential source of error,given that all diagnoses of SKs were clinical, but isprobably of less importance in those individuals withmultiple SKs, as it is highly unlikely that a multiplicity of



these lesions could be incorrectly diagnosed. This biaswas minimized as far as possible by adhering to strictdiagnostic criteria, and is reflected in the high diagnos-tic accuracy reported in the validation study. Regardingour use of self-reported (C.A.F. observed) skin colour,both the dermatologists’ and participants’ gradings ofskin colour have been shown to be similar in theNambour population.27 This, together with the size ofthe risk estimates, and the concurrence of these findingswith those of numerous other descriptive and analyticalskin cancer studies, makes it unlikely that observationbias could materially affect these results.

In summary, this report substantially strengthens theevidence from other studies which has implicated a lackof cutaneous pigment and UV exposure in the develop-ment of SKs.9,15 The stronger association detected forcutaneous phenotype in those individuals with multipleprevalent SKs suggests the existence of a group ofindividuals highly susceptible to the mutagenic effectsof UV radiation, possibly via an inherited genetic muta-tion, e.g. affecting DNA repair, or at a locus for a tumoursuppressor gene or an oncogene. Tissue samples havebeen collected for genetic analyses to help clarify thisissue. While these findings require confirmation, theyhighlight the risks associated with sunburn in childhoodand high levels of UV exposure in the development of SKs,and provide support for the current approach of publichealth authorities in recommending sun-avoidancebehaviour, especially in people with sun-sensitive skins.

Acknowledgments

This work was supported by the National Health andMedical Research Council of Australia, grant number922608.

References1 Green A, Battistutta D. Incidence and determinants of skin cancer

in a high-risk Australian population. Int J Cancer 1990; 46: 356–61.

2 Kricker A, English DR, Randell PL et al. Skin cancer in Geraldton,Western Australia: a survey of incidence and prevalence. Med JAust 1990; 152: 399–407.

3 Bendl BJ, Graham JH. New concepts on the origin of squamous cellcarcinomas of the skin: solar (senile) keratosis with squamous cellcarcinoma—a clinicopathologic and histochemical study. ProcNatl Cancer Conf 1970; 6: 471–88.

4 Blum HF. Carcinogenesis by Ultraviolet Light. Princeton: PrincetonUniversity Press, 1959.

5 International Agency for Research on Cancer. Monographs on theEvaluation of Carcinogenic Risks to Humans, Vol. 55: Solar andUltraviolet Radiation. Lyon: International Agency for Research onCancer, 1992.

6 Kricker A, Armstrong BK, English DR. Sun exposure and non-melanocytic skin cancer. Cancer Causes Control 1994; 5: 367–92.

7 Marks R, Ponsford MW, Selwood TS et al. Non-melanotic skincancer and solar keratoses in Victoria. Med J Aust 1983; 2: 619–22.

8 Marks R. Premalignant disease of the epidermis. The Parkes Weberlecture 1985. J R Coll Phys London 1986; 20: 116–21.

9 Green A, Beardmore G, Hart V et al. Skin cancer in a Queenslandpopulation. J Am Acad Dermatol 1988; 19: 1045–52.

10 Engel A, Johnson ML, Haynes SG. Health effects of sunlightexposure in the United States. Results from the first NationalHealth and Nutrition Examination Survey, 1971–74. Arch Der-matol 1988; 124: 72–9.

11 Goodman GJ, Marks R, Selwood TS et al. Non-melanotic skincancer and solar keratoses in Victoria—clinical studies II. Aus-tralas J Dermatol 1984; 25: 103–6.

12 Silverstone H, Searle JHA. The epidemiology of skin cancer inQueensland: the influence of phenotype and environment. Br JCancer 1970; 24: 235–52.

13 Silverstone H, Campbell CB, Hosking CS et al. Regional studies inskin cancer. First report: north-western Queensland. Med J Aust1963; 2: 312–15.

14 Frost CA, Green AC. Epidemiology of solar keratoses. Br J Dermatol1994; 131: 455–64.

15 Vitasa BC, Taylor HR, Strickland PT et al. Association of non-melanoma skin cancer and actinic keratosis with cumulative solarultraviolet exposure in Maryland watermen. Cancer 1990; 65:2811–17.

16 Marks R, Jolley D, Lectsas S, Foley P. The role of childhoodexposure to sunlight in the development of solar keratoses andnon-melanocytic skin cancer. Med J Aust 1990; 152: 62–6.

17 Khlat M, Vail A, Parkin M, Green A. Mortality from melanoma inmigrants to Australia: variation by age at arrival and duration ofstay. Am J Epidemiol 1992; 135: 1103–13.

18 Thompson SC, Jolley D, Marks R. Reduction of solar keratoses byregular sunscreen use. N Engl J Med 1993; 329: 1147–51.

19 Naylor MF, Boyd A, Smith DW et al. High sun protection factorsunscreens in the suppression of actinic neoplasia. Arch Dermatol1995; 131: 170–5.

20 Green A, Battistutta D, Hart V et al. The Nambour skin cancer andactinic eye disease prevention trial: design and baseline character-istics of participants. Control Clin Trials 1994; 15: 512–22.

21 Groves GA. Ultraviolet radiation and the skin. In: Malignant SkinTumours (Emmett AJJ, O’Rourke MGE, eds). Melbourne: ChurchillLivingstone, 1991; 23–35.

22 Schulze R. Measurement of sunlight: the ultraviolet ‘B’ radiationin sunlight. In: International Conference on Ultraviolet Carcinogen-esis. NCI monograph 50 1978, 213–15.

23 EGRET. Seattle: Statistics and Epidemiology Research Corporation,1990.

24 Gallagher RP, Hill GB, Bajdik CD et al. Sunlight exposure, pigmen-tation factors, and risk of nonmelanocytic skin cancer. II. Squa-mous cell carcinoma. Arch Dermatol 1995; 131: 164–9.

25 Gallagher RP, Hill GB, Bajdik CD et al. Sunlight exposure, pigmen-tary factors, and risk of nonmelanocytic skin cancer. I. Basal cellcarcinoma. Arch Dermatol 1995; 131: 157–63.

26 Holman CD, Evans PR, Lumsden GJ, Armstrong BK. The determi-nants of actinic skin damage: problems of confounding amongenvironmental and constitutional variables. Am J Epidemiol 1984;120: 414–22.

27 Green A, Martin NG. Measurement and perception of skin colourin a skin cancer survey. Br J Dermatol 1990; 123: 77–84.



the prevalence and determinants of solar keratoses at a subtropical latitude (queensland, australia)

Documents