Sun exposure may protect against non-Hodgkin lymphoma: A case-control study



Ultraviolet radiation is a hypothesised risk factor for non-Hodgkin lymphoma (NHL) but no epidemiological study has examined this association using direct measures of sun exposure in individuals. Adults aged 20–74 years living in NSW and ACT, Australia, were the study population. Cases (704 of 829 invited to take part, 85%) were diagnosed January 2000 to August 2001. Controls (694 of 1,136 invited to take part, 61%) were randomly selected from state electoral rolls and frequency-matched to cases by age, sex and state of residence. A self-administered questionnaire and telephone interview measured outdoor hours on working and nonworking days and vacations at 10, 20, 30, 40, 50 and 60 years of age. Logistic regression models of NHL and sun exposure contained the 3 matching variables, ethnicity and sun sensitivity measures as covariates. Contrary to expectations, risk of NHL fell with increasing reported sun exposure hours. Relative to 1.0 for the lowest quarter of total sun exposure hours, the odds ratios (ORs) for successively higher quarters were 0.72 (95% CI 0.53–0.98), 0.66 (0.48–0.91) and 0.65 (0.46–0.91) (ptrend=0.01). The association of sun exposure on nonworking days with NHL was stronger; OR for highest quarter 0.47 (0.34–0.66) (ptrend=0.0001). Risk also fell with sun exposure on vacations; OR for highest quarter 0.60 (0.43–0.85) (ptrend=0.003). These associations appeared strongest in women and in childhood. There was little evident trend in risk with exposure on working day. Our results provide strong statistical evidence for an inverse association between sun exposure and NHL. Increasing evidence that vitamin D may protect against cancer makes UV-mediated synthesis of vitamin D a plausible mechanism whereby sun exposure might protect against NHL. © 2004 Wiley-Liss, Inc.

Recent interest in the possibility that sun exposure might cause non-Hodgkin lymphoma (NHL) has been prompted by findings in ecological studies of parallel geographic patterns and temporal trends in NHL and cutaneous malignancies,1 a negative correlation of NHL incidence with latitude across European countries,2 a positive correlation between incidence of NHL and regional levels of ultraviolet radiation in England and Wales3 and an upward shift in risk of NHL with migration from Europe to Australia.2 There is, though, an inconsistency in these patterns: NHL incidence and mortality have been reported to rise with increasing latitude and fall with increasing estimated ambient ultraviolet irradiance in the United States.4, 5, 6, 7 Positive associations between a personal history of squamous cell carcinoma (SCC) and basal cell carcinoma (BCC), which may indicate high sun exposure, and risk of NHL in cancer registry cohorts8, 9 and weak evidence that sun sensitivity is associated with an increased risk of NHL in individuals,10 however, offer additional support for a causal association.

The known capacity of UV radiation to produce T-cell immune suppression, both locally in the skin and at locations remote from the site of exposure,11 and evidence that risk of NHL is greatly increased in people with T cell immune deficiency associated with HIV infection or treatment after organ transplantation,12, 13 make a causal association between sun exposure and NHL plausible.

While studies that infer individual sun exposure from occupation have been reported, with mixed results,5, 14, 15, 16, 17 no evidence has yet been reported on the relationship between NHL and sun exposure from studies using direct measures of individual sun exposure. We describe here the results of a population-based case-control study conducted in the Australian State of New South Wales (NSW), an environment of high ambient solar radiation, to determine whether high sun exposure is associated with an increased risk of NHL. The study also investigated selected occupational, immunological and infectious risk factors for NHL that will be the subject of future reports.



Eligible cases were patients aged 20–74 and resident in NSW or the adjacent Australian Capital Territory (ACT) who were ascertained following notification to the NSW Cancer Registry or direct notification to the investigators as first diagnosed pathologically with NHL between 1 January 2000 and 31 August 2001. Notification of cancer to the Registry is legally required of all pathology laboratories, hospitals and radiotherapy facilities in NSW. Patients with diagnoses of chronic lymphocytic leukaemia, plasma cell myeloma, precursor B and T lymphoblastic leukaemia, and lymphomatoid granulomatosis grades 1 and 2 were excluded. Notifying clinicians were also asked to exclude patients who had a history of immunosuppression for organ transplantation or HIV infection.

Subjects were considered ineligible if they had poor English language skills or illness or disability that prevented a 60 min telephone interview. Otherwise eligible patients who had died before initial contact or before interview for the study were excluded from data collection. A total of 1,217 eligible patients were ascertained and confirmed as eligible by notifying clinicians (694 males, 523 females). They were an estimated 93.4% of patients diagnosed in the study period and registered as NHL by the Cancer Registry;18, 19 those registered but not ascertained for the study were probably mainly cases registered after ascertainment ended or not diagnosed pathologically. Of those ascertained and initially confirmed as eligible, 197 were subsequently found to be ineligible because of prior immunosuppression or immune deficiency (28), poor English (73), illness (74) or disability preventing interview (22). A further 178 were excluded from the possibility of interview because they had died (144) or could not be contacted (34). Approximately 18–20% of apparently eligible patients were not enrolled because of severity of their illness or death.

Of 842 cases invited to participate, 717 (85%) did so. Thirteen were subsequently excluded after pathology report review because of low confidence in the diagnosis of NHL, leaving 704 cases (410 males, 294 females) for analysis. Most cases (68%) were interviewed within 6 months of diagnosis, 92% within 12 months and almost all (99%) within 18 months (median 147 days).

Pathology reports for consenting cases were reviewed by one of us, an anatomical pathologist with a particular interest in haematopathology (JT), to assess confidence in the diagnosis of NHL and, where possible, to assign a cell phenotype and WHO (ICD-03) code.20 The largest individual classification categories were follicular lymphoma (239 cases) and diffuse large B cell lymphoma (225 cases).

Controls were randomly selected from NSW and ACT electoral rolls (lists of those registered to vote) at intervals throughout the study to match the expected distribution of cases by age, sex and State of residence at diagnosis. Enrolment to vote in State and National elections is compulsory for Australian citizens and an estimated 91% of the population is registered. The eligibility criteria for cases were applied to controls, that is, able to give a 60 min interview and excluded if illness, disability or poor English language skills prevented such an interview. While we did not ask for disclosure of HIV status in controls, the adult prevalence of HIV in Australia (less than 0.1%21) indicated that less than 1 control was likely to be positive. Of 1,687 apparently eligible controls, 145 were subsequently found to be ineligible because of poor English (74), illness (59) or disability preventing interview (12). A further 406 were excluded from the possibility of interview because they had died (5) or could not be contacted (401). In all, 694 of 1,136 eligible and contactable controls were interviewed (61%).

The study was approved by the human research ethics committee at each participating institution. Subjects gave oral informed consent to be interviewed and written informed consent for collection of information about pathology and blood tests from medical records and a blood sample for viral and genetic testing.

Data collection

Procedures were identical for cases and controls. All were mailed a letter that invited their participation in a research project about the effects of the environment on the development of NHL; neither the letter nor enclosures specifically mentioned sun exposure. A computer-assisted telephone interview was the main source of information on sun exposure. Participants were asked to complete a residence and occupation calendar for each year of their life22 and a brief paper questionnaire before the interview; the calendar information was used to customise the interview and assist recall. All interviews were conducted between March 2000 and May 2002 by 4 trained and experienced female interviewers. The interviewers were blinded to the case or control status of subjects and each was allocated equal proportions of cases and controls to interview.

Sun exposure measures

At interview, subjects were asked how many hours they would normally have spent outdoors between 9 am and 5 pm on working or school days, nonworking days or weekends and vacations in the warmer months and the cooler months of the years they turned 10, 20, 30, 40, 50 and 60 (“decade years”). We have previously validated the decade years approach as a predictor of total recalled sun exposure.23 The interviewer explained to the subject that “outdoors” meant hours when they were outside and not under any shade: in this manner, subjects were directed to report hours when they would have been exposed to the sun. The warmer months were defined as being from October to April in the Southern Hemisphere and from March to August in the Northern Hemisphere. We constructed separate variables for working and nonworking days sun exposure by summing the reported hours outdoors across the decade years, and a combined working and nonworking days variable as an indicator of total outdoor exposure. We have previously demonstrated the utility of this method for measuring sun exposure22, 23, 24, 25, 26, 27 and hereafter refer to outdoor exposure simply as sun exposure.

We also asked subjects directly for their hours of occupational sun exposure. For each job recorded in the calendar, data were collected about the number of days worked per week, hours worked per day and hours worked outdoors per day. Occupational hours of exposure were totalled for 50 weeks a year, assuming 2 weeks for vacations and sick leave. For exposure hours on vacations, subjects were asked at each decade year how many vacation weeks they had taken in both the warmer and cooler months and the number of hours they typically spent outdoors during these vacation weeks.

Subjects separately rated their frequency of hat and sunscreen use during the warmer months of the decade years by selecting 1 of 5 categories for each variable at interview: “Always or almost always”, “Not always but more than half the time”, “About half the time”, “Less than half the time” and “Never or hardly ever”.

Variables that indicate skin pigmentation or sun sensitivity (ethnicity, skin colour on the inner, upper arm, eye colour, hair colour and ability to tan) were also measured in our study and have been reported on in Hughes et al.10. Ethnicity was self-assessed in the questionnaire from a list of 12 ethnic groupings and skin colour from a list of 6 categories from very fair to black, while ability to tan on repeated exposure to sunlight was asked in the telephone interview using 4 response categories from “go very brown and deeply tanned” to “get no suntan or get freckled only”.

Three measures of ambient solar irradiance [global solar radiance (GSR; mJ/m2),28, 29 cloud-adjusted solar erythemal UVB (kJ/m2),30 and cloud-adjusted solar UVA (J/cm2)31] were assigned to each residential location for each subject using latitude and longitude coordinates. For a simple latitude measure in Australian and New Zealand born subjects only, residential latitudes in the southern hemisphere were categorised into 3 bands: north of Sydney (<33.5°S), metropolitan Sydney (33.5° to 34.2°S) and south of Sydney (>34.2°S).

Socioeconomic status was assigned to each subject's address at interview using a published index constructed from 1996 Census data on distributions of income, education, unemployment, occupation, non-English speaking background and indigenous origin of households in local government areas.32

Statistical analysis

For lifetime exposure, we analysed working and nonworking days exposure both separately and together in 1 measure, and each of these separately in the warmer and cooler months and for both together. We also examined sun exposure at each individual decade year, without adjustment for sun exposure at other decade years. Other analyses examined the following variables in separate models: sun exposure hours in occupations, sun exposure hours on vacations, sun-related behaviours (hat and sunscreen use), ambient UVA, UVB and GSR, and latitude. Variables for personal sun exposure and ambient solar irradiance were categorised into approximate quarters or thirds of the exposure distribution in controls. It was known a priori22, 33 that men at all ages have higher outdoor hours than women, so separate models of personal sun exposure were constructed for men and women together, men only and women only.

Odds ratios and 95% confidence intervals were calculated in unconditional logistic regression models adjusting for age, sex (where applicable), State of residence (NSW or ACT) and individual characteristics that might be associated with sun avoidance (fair skin and poor ability to tan). Each model also included ethnicity as a covariate because residents born outside Australia were under-represented in the controls in our study relative to expectation based on the 2001 Australian Census, as discussed in Hughes et al.10 Ethnicity was self-assessed in terms of the “region [of the world] from which your family originated” and represented for analysis by the following categories: “British or Irish”; “Asian”; “Western/Northern European”; “Southern European”; “Mixed” and “Other”. Most Australian residents who were born in Australia are of British or Irish origin. To further exclude the possibility that any observed association was due to selection bias in controls, we also estimated them in Australian- and New Zealand-born subjects only.

Tests for linear trend were done by fitting ordered categories of a variable as a single ordinal variable in the logistic regression models; trend p-values were based on the Wald test. The likelihood ratio test was used to calculate p-values for heterogeneity. All analyses were done using SAS software (SAS Institute, Cary NC., 1989).


We observed a moderately strong inverse association between reported sun exposure hours in the warmer and cooler months combined, summed across the decade years, and risk of NHL in men and women together; the odds ratio was 0.65 (95% CI 0.46–0.91) in the highest exposure category (Table I). This inverse association was observed for sun exposure on working and nonworking days together and on nonworking days only, but not on working days only; risk decreased consistently with increasing sun exposure in the former 2 categories. Similar results were obtained when the analysis was confined to Australian- and New Zealand-born subjects (553 cases, 580 controls; data not shown).

Table I. Risk of Non-Hodgkin Lymphoma With Sun Exposure in Warmer and Cooler Months and With Lifetime Occupational Sun Exposure
 Men and womenMenWomen
OR95% CIpOR95% CIpOR95% CIp
  • 1

    Decade years were at 10, 20, 30, 40, 50 and 60 years of age. Men reported outdoors hours of 10–510 hr on working days (summed across 1 week in the warmer months and 1 in the cooler months in each decade year) and 3–164 on nonworking days and women 5.8–366 hr (working days) and 0–202 (nonworking).

  • *

    p-value for contribution of that variable to the model containing age, (sex), state, ethnicity, skin colour and ability to tan.

  • **

    p-value for trend.

Working and nonworking days during the decade years1
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.720.53–0.98 0.690.45–1.05 0.990.63–1.56 
 50–75% exposure quartile0.660.48–0.91 0.900.59–1.38 0.680.41–1.11 
 Highest exposure quartile0.650.46–0.910.03*0.690.44–1.090.20*0.490.29–0.820.02*
   0.01**  0.27**  0.003**
Working days during the decade years
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.980.73–1.33 0.810.54–1.21 1.340.83–2.16 
 50–75% exposure quartile0.910.66–1.25 0.940.61–1.37 1.080.65–1.80 
 Highest exposure quartile0.950.68–1.330.94*0.880.58–1.350.77*0.770.45–1.330.16*
   0.68**  0.69**  0.22**
Nonworking days during the decade years
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.830.61–1.11 1.010.67–1.50 0.790.50–1.24 
 50–75% exposure quartile0.570.42–0.79 0.780.51–1.20 0.520.32–0.85 
 Highest exposure quartile0.470.34–0.660.0001*0.560.35–0.880.04*0.390.23–0.640.001*
   0.0001**  0.008**  0.0001**
Lifetime occupational sun exposure
 None1.00  1.00  1.00  
 Lowest exposure tertile1.030.76–1.40 0.940.63–1.40 1.030.58–1.85 
 Middle exposure tertile1.040.76–1.43 0.880.59–1.31 1.220.70–2.13 
 Highest exposure tertile1.210.87–1.690.70*1.200.81–1.780.47*1.270.73–2.230.79*
   0.30**  0.46**  0.32**

The associations of NHL with working and nonworking day exposure and nonworking day exposure alone appeared to be strongest in women: the OR for working and nonworking day exposure was 0.49 in the highest category of exposure (ptrend = 0.003), and that for nonworking day exposure was 0.39 (ptrend = 0.0001) (Table I).

The numbers of sun exposure hours on nonworking days in the warmer months and the cooler months, when examined separately, were strongly and similarly associated with risk of NHL in both men and in women (data not shown).

When individual decade years were examined, a strong inverse association between sun exposure at age 10 and risk of NHL was observed in men and women together (ORs of 0.87 for the second, 0.67 for the third and 0.54 for the highest exposure quartile, ptrend = 0.0001). This association was also present in males and females when they were analysed separately (ORs 0.65 and 0.47, respectively, for the highest exposure category with ptrend 0.009 and 0.002) and in Australian and New Zealand-born subjects (OR 0.60 for the highest exposure category, ptrend 0.001). For women, but not men, the trend towards falling risk with increasing sun exposure was also apparent and statistically significant for exposure at 30, 50 and 60 years of age. At age 10, everyone reported a weekly pattern of 5 school or working and 2 weekend or nonworking days; half the subjects had more than 4 hours a day of sun exposure on nonworking days in the warmer months and half reported 4 or more hours in the cooler months.

Lifetime occupational sun exposure was not associated with NHL in either sex (Table I). This finding was coherent with that for decade-year exposure on working days, a parallel measure of nonrecreational sun exposure. There was little evidence that either working days or occupational sun exposure modified the association of nonworking days exposure with NHL in men, women or both sexes combined.

Higher levels of sun exposure on vacations during the decade years were associated with a reduced risk of NHL for men and women together, regardless of the time of year (Table II). This inverse association was also evident in men and women separately (Table II), in Australian and New Zealand-born subjects (data not shown) and at all individual decade years for men and women together except age 40 (data not shown).

Table II. Risk of Non-Hodgkin Lymphoma With Vacation Sun Exposure1
Vacation sun exposure during the decade yearsMen and womenMenWomen
OR95% CIpOR95% CIpOR95% CIp
  • 1

    Vacation sun exposure was accumulated across the decade years at 10, 20, 30, 40, 50 and 60 years of age separately for the warmer and cooler months and for the warmer and cooler months combined.

  • *

    p-value for contribution of that variable to the model containing age, (sex), state, ethnicity, skin colour and ability to tan.

  • **

    p-value for trend.

Warmer and cooler months combined
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.980.72–1.32 1.040.70–1.54 0.720.46–1.14 
 50–75% exposure quartile0.820.60–1.12 0.770.51–1.17 0.770.49–1.22 
 Highest exposure quartile0.600.43–0.850.01*0.530.34–0.820.008*0.530.33–0.860.08*
   0.003**  0.002**  0.02**
Warmer months
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.780.57–1.05 0.890.60–1.32 0.550.35–0.88 
 50–75% exposure quartile0.810.59–1.10 0.690.45–1.06 0.620.39–0.98 
 Highest exposure quartile0.650.47–0.910.07*0.720.47–1.100.29*0.550.34–0.880.03*
   0.02**  0.08**  0.02**
Cooler months
 Lowest exposure quartile1.00  1.00  1.00  
 25–50% exposure quartile0.870.64–1.17 0.850.56–1.29 0.980.62–1.57 
 50–75% exposure quartile0.780.58–1.06 0.750.49–1.16 0.810.50–1.33 
 Highest exposure quartile0.640.46–0.880.04*0.620.39–0.970.17*0.670.41–1.100.32*
   0.005**  0.03**  0.07**

Risk of NHL, adjusted for total time outdoors, was not significantly associated with frequency of hat wearing or frequency of sunscreen use in any decade year; neither showed consistent evidence of a trend, either up or down, with increasing use (data not shown). There was no evidence of consistent variation in the effect of sun exposure on risk of NHL among categories of hat use or sunscreen use, or ability to tan.

Risk of NHL showed weak and nonsignificant downtrends with increasing cumulative ambient solar UVB irradiance in the first 10 years of life and over the whole of life (data not shown) and even more weakly with cumulative ambient UVA and GSR and falling residential latitude. Weighting of sun exposure on working and nonworking days or on nonworking days alone by ambient UVB, UVA or global solar radiance in men and women together or in women alone did not materially alter their apparent protection against NHL.

There was no appreciable difference in socioeconomic status between cases and controls (p = 0.755) and its inclusion in models of sun exposure and risk of NHL described above had a negligible effect on the associations observed, increasing by 0.01 only the odds ratios for the third quarters of sun exposure on working and nonworking days and nonworking days only.


We have found an unexpected, significant inverse association between estimated lifetime sun exposure and NHL, which was contrary to our prior hypothesis that sun exposure would increase risk of NHL. Risk fell most strongly and consistently with increasing sun exposure hours on nonworking days and was not associated with sun exposure on working days or an alternative measure of occupational sun exposure. High sun exposure during vacations was also associated with a reduced risk of NHL. None of our estimates of lifetime residential latitude or ambient UV, hat wearing or sunscreen use was associated with risk of NHL. The inverse associations between measures of sun exposure and NHL generally appeared stronger in women than men and there were strong dose-response relationships in women. The apparent protective effect of whole year sun exposure appeared greatest during childhood.

Our case ascertainment was population-based, participation among eligible and contactable cases was high (85%), and final case numbers were large. Cases with more aggressive NHL, however, were probably under-represented in our study population since severity of illness or death prevented the enrolment of some 18–20% of apparently eligible cases. The proportions of follicular lymphomas (34%) and diffuse large B-cell lymphomas (32%) in our cases, however, were as expected in a country like Australia (30–40%).34

The consent rate was comparatively low in eligible and contactable controls (61%) and people born outside Australia were under-represented. Use of the electoral roll as the sampling frame and, possibly, less participation by people of non-English speaking background probably explain this bias with respect to place of origin in controls. Ethnicity assessed by region of family origin was included as a covariate in all models to minimise its effects and excluding cases born outside Australia and New Zealand in key analyses did not materially affect the results, thus suggesting that selection bias on the basis of place of origin in controls did not explain the strong association between sun exposure and NHL. Bias might also have arisen from the differentially low participation rate in all controls. We attempted to minimise selection for or against sun exposure in cases or controls and differential recall of sun exposure by cases and controls by not declaring a specific interest in sun exposure as an environmental factor; in addition, it is doubtful whether the present scientific interest in a role for sun exposure in NHL is at all well known in the community. There was also no appreciable association between socioeconomic status and NHL in our data, which weighs against major bias by socioeconomic status or variables that correlate with it.

We relied on self-reported sun exposure history using methods that we have developed over 15 years and used successfully in a number of case-control studies,23, 24, 35 both in an extensive, face-to-face interview form and the abbreviated telephone interview form of our study. The face-to-face form has been shown to be reliable on repetition at an interval of about 5 years25 and the telephone form has been shown to be reliable in comparison with the face-to-face form.26 Our structured approach to retrospective measurement of sun exposure includes questions about both the pattern and amount of sun exposure and is oriented to personal circumstances or events as recall cues to minimise measurement error. Using the detailed questions, not only is total accumulated lifetime sun exposure hours estimated but, in addition, the components parts can be examined, such as sun exposure on working and nonworking days, or at particular ages. We would limit analyses to accumulated lifetime exposure only when we have a strong hypothesis that it is the biologically relevant variable.

No previous study has examined personal sun exposure as a risk factor for NHL. The evidence in our study suggests that risk of NHL decreased with increasing hours of total lifetime sun exposure. Since this is the first report of such an association, and we had no prior belief about the exposure period of greatest aetiological relevance, we examined the components of total sun exposure. Not only did we observe that the association of total lifetime sun exposure with NHL was apparently stronger in women, but that both men and women had reduced risks of NHL with increasing sun exposure on nonworking days or vacations, and at age 10. The apparently stronger effect of exposure in childhood than at other ages is consistent with our knowledge that sun exposure is highest and similar in boys and girls in childhood in Australian populations.33 The higher exposure hours and their similarity in the sexes at this age may explain why childhood sun exposure appeared to have the strongest association with NHL. In our population, the greatest opportunity for sun exposure is on nonworking days. About half the men and very few women reported spending any time outdoors in occupations in our study.

While the positive correlation of NHL incidence and mortality with latitude and negative correlation with ambient ultraviolet irradiance in the United States4, 5, 6, 7 offer some support for our unexpected finding, they are inconsistent with other reports of an opposite correlation, both within and across countries,2, 3 and contrast with the limited evidence in our study of any association between NHL and residential latitude or measures of ambient solar radiation. This lack of correlation, however, could be due to comparative lack of heterogeneity in residential latitude and ambient UV, being based on a population that spans 9° of latitude and of which more than 60% reside in a single city (Sydney).

There is also some coherence between our findings and previous studies of sun exposure inferred from occupation. We found little evidence that working days or occupational sun exposure influenced risk of NHL, and other studies have generally shown associations close to null among workers with the highest UV exposure,5, 14, 15, 16, 17 with the possible exception of farmers and farm workers, who probably have an increased risk of NHL.5, 36, 37, 38, 39 Classification of exposure in these studies,5, 14, 15, 16, 17 however, was typically based on data from a single time-point rather than a lifetime occupational history and, even with careful assessment by an industrial hygienist, there is real potential for misclassification of sun exposure when inferred from job titles. Thus their evidence for lack of an effect of occupational sun exposure is weak.

If sun exposure protects against NHL, it would be expected that individuals with a history of skin cancer, most of which are strongly related to sun exposure,27 would be at reduced risk and that risk would be less in those with sun sensitive skin, which is more transparent to solar radiation.40 Conversely, sun protective behaviours might be expected to increase risk of NHL. On present evidence, none of these is true. People who have had cutaneous SCC8, 9 and BCC9 consistently have a higher than expected frequency of subsequent NHL in cancer registry cohort studies, and we found that a past history of treatment for skin cancer (unverified) in this population was associated with a slight, nonsignificant increase in risk of NHL,10 a finding consistent with the only other study of self-reported past skin cancer in people with NHL.41 We have also previously reported from this study that fair skin and poor tanning ability are associated with a small increase in risk of NHL.10 We found, in addition, no evidence that sun-protective behaviours (wearing hats or sunscreen) increased the risk of NHL, although this could simply reflect incomplete protection in people who otherwise had high sun exposure.

These substantial inconsistencies aside, our study provides strong statistical evidence for an inverse association between NHL and several measures of what might be termed, perhaps loosely, recreational sun exposure, including nonworking day exposure and vacation exposure. These results could be taken to suggest that if sun exposure does protect against NHL it is an intermittent pattern of sun exposure that is the most protective.

We argued that our original hypothesis that sun exposure increases risk of NHL was supported biologically by evidence that sun exposure causes T-cell immune suppression11 and that risk of NHL is greatly increased in people with T cell immune deficiency associated with HIV infection or immunosuppressive therapy after organ transplantation.12, 13 Is it possible, contrary to this argument, that the effects of sun exposure on immune responses might decrease rather than increase risk of NHL? Available evidence suggests that UV exposure depresses the function of Th1 cells, which facilitate cell-mediated immune responses, while enhancing the activity of Th2 cells, which facilitate humoral immune responses.42 In HIV infection, there is a broadly similar pattern of change,43, 44, although with more extreme depression of Th1 responses, which, as noted above, is associated with an increased not a decreased risk of NHL. However, preliminary analyses of markers of immune function that we have measured in our study show significantly lower risks of NHL in people with histories of hayfever and food allergy (unpublished results), which are not confounded by sun exposure and are consistent with the balance of previous studies of NHL and history of allergy.45, 46, 47, 48 Allergic conditions are associated with a dominant Th2 immune response. Therefore, we cannot reject the possibility that immune function effects of sun exposure could reduce rather than increase risk of NHL.

The role of sun exposure in vitamin D production may also provide a possible explanation for its inverse association with NHL. Experimental studies show that the active form of vitamin D, 1,25D3, has anti-proliferative and pro-differentiation effects on cells that possess vitamin D receptors,49 including lymphocytes50 and lymphoma cell lines.51 There is also evidence that 1,25D3 induces apoptosis and inhibits angiogenesis and tumour spread.49, 52 There is evidence too that plasma concentration of vitamin D is inversely associated with risk of colorectal and prostate cancers,53, 54, 55, 56 and some evidence that sun exposure may protect against these cancers.57, 58

Vitamin D synthesis in humans requires exposure of the skin to solar radiation and body levels of vitamin D are primarily determined by UV-mediated synthesis in skin even in high-latitude countries.59 Vitamin D synthesis due to sun exposure, however, is closely regulated and no more than 15% of the total cutaneous 7-dehydrocholesterol, the substrate for UV action, is converted to provitamin D3 whether skin is exposed to the sun for 30 min or 8 hr at the equator.60 This suggests that vitamin D production would probably be greater when sun exposure is received intermittently than when it is received more continuously, which is coherent with our finding that nonworking day exposure is the measure of sun exposure most strongly related to NHL.

The picture for vitamin D, however, is complicated by its capacity to suppress delayed hypersensitivity (Th1) responses61 and possibly to enhance Th2 cell development.62 Indeed, it has been suggested that some, at least, of the immunosuppressive effects of sun exposure may be mediated by vitamin D production.63 Therefore, more research is required before we can be at all certain that vitamin D synthesis mediates a protective effect of sun exposure against NHL, if there is one.

Our study provides strong statistical evidence of an inverse association between sun exposure and NHL, which, on the face of it, negates the hypothesis that sun exposure causes NHL. It also suggests the hypothesis that sun exposure protects against NHL, and the evidence that vitamin D has antineoplastic properties gives this hypothesis some biological plausibility. This new hypothesis of protection by sun exposure should be tested in other studies.


We gratefully acknowledge all the people who generously participated in the research, the clinicians who gave permission for us to approach their patients, the NSW Central Cancer Registry staff for their assistance with case ascertainment and staff at the Hunter Valley Research Foundation for conducting the telephone interviews. Special thanks to M. Agaliotis and C. Goumas for data collection and data entry and to J. Turner for telephone follow-up. Ann Maree Hughes has a University of Sydney, Faculty of Medicine postgraduate award, Claire Vajdic has a postdoctoral research fellowship from the National Health and Medical Research Council and Bruce Armstrong's research is supported by a University of Sydney Medical Foundation program grant. The National Centre in HIV Epidemiology and Clinical Research is funded by the Commonwealth Department of Health and Ageing. Lin Fritschi is supported by an NHMRC Population Health Career Development Award.