Open Access

Research Article

Methods Top

Prevalence data source. The 1995 Australian National Health Survey (NHS) was conducted by the Australian Bureau of Statistics (ABS) during the 12-month period from January 1995 to January 1996. Approximately 54,000 people from all states and territories and across all age groups provided information about their own health status. Residents of a stratified, multistage area, random sample of 23,800 private households and households in nonprivate dwellings were interviewed in person by ABS interviewers; persons within each state/territory had a known, and in the main equal, chance of selection. Residents of hospitals and other institutions were excluded, however, and the Northern Territory (NT) sample was predominantly urban. Responses were received from 91.5% of households and 97% of people from these households fully completed questionnaires. Quarterly population estimates were used by ABS for standardization, and prevalence rate estimates were adjusted for household size (ABS 1995a

The survey interviewers recorded information on recent illness and/or long-term conditions, as reported by respondents. Specific questioning about conditions, including diabetes, arthritis, and asthma, was followed by action-based questions on recent visits to hospitals or health professionals. Information on medication usage, including insulin and medications for asthma, arthritis, and allergies, was also elicited (ABS 1995b). Disease conditions were classified by the International Classification of Diseases, 9th Revision (cited by ABS 1995a), categories. Information recorded in the survey was not medically verified, however.

For this study we used summary-data prevalence estimates for four immune-related disorders from the published survey results (ABS 1995c). We compared the age- and sex-standardized prevalence rates of type 1 diabetes, RA, eczema/dermatitis, and asthma, per 1,000 population, over the eight major state and territory regions of Australia, as listed in Table 1.

Latitude and ultraviolet radiation. The latitude of the regional capital city (decimal degrees south; Geoscience Australia mapping agency) was used for each region (Table 1). Three UVR measures were examined, each relating to the Australian UV index (Lemus-Deschamps et al. 1999) and expressed in milliwatts per square meter (where 1 UV index unit = 25 mW/m²). The measures were, for each regional capital, an arithmetic mean of 12 monthly UV index averages, each monthly average calculated from single, daily (at local solar noon), erythemally weighted clear-sky UV index values derived from ozone data over the period 1979-1993 (Lemus-Deschamps L. Personal communication), a midwinter minimum (June) solar-noon UVR value, and a midsummer maximum (January) solar-noon UVR value (Table 1

Statistical analysis. To account for the NHS sampling variability between regions, we calculated approximate standard errors (SEs) for each regional prevalence estimate, as outlined in Appendix H and "Technical Note: Sampling Variability" in the respective ABS publications (ABS 1995a, 1995c). The approximate SEs were calculated from SEs for the corresponding numerator estimates shown in Table 2, as listed by ABS for each region (ABS 1995a, 1995c). Age-standardization factors for the different regions were also applied, as recommended in the same Technical Note, and the final SEs for the prevalence rates calculated (Table 2). The reciprocal of the variance of each prevalence rate was then used to weight the relationships to account for differing sample sizes in the regions.

Associations between the environmental variables and the immune disorder prevalence rates were examined by variance-weighted least-squares regression of the prevalence rate estimates versus, first, the latitude values of the regional capital cities, and second, the three UVR measures for each of the regional capitals. The magnitude of change in disease prevalence rates over the north-south range was compared by substituting latitude values for Darwin and Hobart into the regression equations. The statistical program Stata 7.0 (Stata Corporation, College Station, Texas, USA) was used for regression analyses.

Because the use of capital city latitudes may not have been representative of the regional population distribution, particularly for larger regions with the capital at either a northern or southern extreme of the region rather than medially placed, a sensitivity analysis was carried out to test the effect of using an alternative latitude value midway between the two main population centers in these regions. Results were then compared with those obtained from the regional capitals.

Results Top

Latitude and immune disorders. The relationships between latitude of the regional capitals and immune disorder prevalence are shown in Figure 1 as regression lines fitted to the prevalence estimates; 95% confidence intervals (CIs) are also shown, and these indicate, reciprocally, the relative weighting applied to each estimate in the regression analysis.

Figure 1.

Associations between latitude and prevalence per 1,000 (open circle) of type 1 diabetes, asthma, eczema/dermatitis, and RA. 95% CIs of prevalence estimates shown as error bars; solid lines denote statistically significant association, p < 0.05; dotted lines denote nonsignificant trend.

Prevalence of type 1 diabetes was positively correlated with latitude (Pearson r = 0.77; p = 0.026), with the prevalence increasing 2.97-fold over the north-south latitude gradient. Asthma prevalence, on the other hand, was inversely correlated with latitude (Pearson r = -0.72; p = 0.046), with the prevalence rate decreasing 0.7-fold, i.e., by approximately one-third, over the same latitude range. Although both eczema/dermatitis and, to a much lesser extent, RA showed trends of increasing prevalence with increasing latitude, these were not statistically significant (Figure 1

Sensitivity analysis. In four of the eight regions, Western Australia (WA), South Australia (SA), Victoria (VIC), and Australian Capital Territory (ACT), at least 70% of the state or territory populations resided in the capital metropolitan area (ABS census data 1991/1996, not shown). Among the remaining regions in which the population was more dispersed outside of the capital, only NT and Queensland (QLD) were also both of wide latitudinal range and had capitals located noncentrally in terms of the regional latitude range, thus potentially biasing the associations markedly.

The effect on the initial associations of using alternative latitude values midway between the two main population centers, Darwin and Alice Springs, for NT (i.e., latitude 18.08 degrees south), and Brisbane and Townsville for QLD (i.e., latitude 23.36 degrees south), was to strengthen the statistical significance of the associations for type 1 diabetes, asthma, and eczema/dermatitis when one or both midway latitudes were substituted. For example, for type 1 diabetes and asthma, the statistical significance of the associations was raised maximally [from p = 0.026 to p = 0.010 for diabetes (r = 0.77 to 0.84), and from p = 0.046 to p = 0.013 (r = -0.72 to -0.82) for asthma] when midway latitudes were used for both regions. Both approaches unavoidably entail some exposure misclassification for dispersed regional populations. Whereas the latter (midway latitude) approach may lead to better estimation of the correlation, by using the capital latitude values for type 1 diabetes and asthma as we have done (Table 1), the estimate was somewhat biased toward the null hypothesis of no association and thus toward underestimation of r. Similarly, for eczema/dermatitis the statistical significance was raised maximally [from p = 0.206 to p = 0.174 (r = 0.50 to 0.54)] when only the QLD latitude was altered, but this was still nonsignificant at the 5% level. For RA, on the other hand, using both midway latitudes gave maximal change, but the statistical significance of the observed trend was lowered [from p = 0.727 to p = 0.795 (r = 0.15 to 0.11)]. In summary, the use of the regional-capital latitude values rather than latitude values of the midway points did not alter the overall conclusions of whether the observed associations were statistically significant at the 5% level.

Ultraviolet radiation and immune disorders. In light of the sensitivity analysis, UVR values for the regional capitals were used. A solar-noon, clear-sky measure of ambient UVR was chosen because midday, or noontime, exposure to the maximal UVR level in sunlight in Australia suppresses the systemic immune response in humans (Hersey et al. 1983). Type 1 diabetes prevalence was inversely correlated with regional average solar-noon UVR (Pearson r = -0.80; p = 0.018), whereas asthma prevalence was positively correlated with regional average solar-noon UVR (Pearson r = 0.73; p = 0.040) (Figure 2). Eczema/dermatitis and RA prevalence showed inverse trends with average solar-noon UVR, but these were not significant (Table 3). The prevalence of asthma was positively correlated with either average or midwinter solar-noon UVR but not midsummer UVR. Although type 1 diabetes prevalence was inversely correlated with all three ambient UVR measures, it was more closely correlated with average or midwinter solar-noon UVR than with midsummer UVR (Table 3

Figure 2.

Associations between regional average ambient solar-noon UVR and prevalence per 1,000 (open circle) of type 1 diabetes and asthma. 95% CIs of prevalence estimates shown as error bars.

Table 1.

Table 2.

Table 3.

Discussion. Using summary age- and sex-standardized data on self-reported prevalence of immune disorders from the ABS 1995 NHS, strong gradients in type 1 diabetes prevalence with latitude and also, inversely, with UVR were observed. The magnitude of change for type 1 diabetes in Australia was an approximately 3-fold increase in prevalence from the northernmost region, NT, to the southernmost region, Tasmania (TAS). These ecologic data provide support for a previously proposed beneficial effect of UVR on autoimmune disorders such as type 1 diabetes (McMichael and Hall 1997). In contrast to type 1 diabetes, asthma correlated negatively with latitude and positively with regional annual or midwinter UVR, but the magnitude of change for asthma prevalence from the north to south of Australia was only 0.7-fold. For both RA and eczema/dermatitis, there were no statistically significant associations between latitude/UVR and disease prevalence.

A strength of this ecologic analysis derives from the relative genetic homogeneity of the Australian population (McLeod et al. 1994) and relative uniformity also in the standard of national health care. A further advantage is the wide latitude range and the resulting wide range of ambient UVR levels over the Australian continent. A high total response rate to the survey was achieved by the ABS over all regions.

One potential source of (probably random) error, however, lies in the self-reported, nonverified nature of the data, which could have resulted in some misclassification of disease conditions. This is particularly so for eczema/dermatitis, the classification of which appears to have been based largely on medication usage; there were no direct disease-specific questions on eczema (ABS 1995b). The classification eczema is often used as an umbrella term encompassing various dry, itchy skin conditions (McNally et al. 2000); dermatitis similarly can include contact dermatitis. Also the eczema/dermatitis category in the NHS included heat eczema and sunburn (ABS 1995a), leading to possible regional differences in prevalence. In addition, although arthritis was allocated four specific questions, misclassification between rheumatoid arthritis, osteoarthritis, and general arthritic or rheumatism conditions could have occurred. Type 1 diabetes, on the other hand, was classified on the basis of 24 direct questions for diabetes and insulin usage, including duration of use, both past and future expected use, and age at first use (ABS 1995b). The relative validity of classification of type 1 diabetes, a serious disease requiring specific treatment, should therefore have been more regionally consistent.

Exposure misclassification at the ecologic (i.e., state or territory) level compared with the individual level could have occurred because actual personal UVR exposure depends on both behavior in relation to the sun (i.e., whether sun avoidant or taking sun protection measures) and on regional ambient UVR. For example, regional UVR would not have been a good measure of personal sun exposure among sun-avoidant people, and if the proportion of such people in each local population varied markedly, then population-level exposure assessment would be biased. In fact, the proportion was small and did not vary substantially by region [percentage recording "don't go out in sun" was 3.4% for NT, 2.0% for QLD, and 2.3% for TAS (ABS 1995c)]. For most individuals, however, personal UVR exposure varies between 5 and 15% of daily total ambient UVR (Gies et al. 1999). Ambient UVR levels thus may provide a reasonable measure of average personal sun exposure at the population level.

Another possible source of measurement error--using capital city latitudes and UVR values as representative of each region--was shown by sensitivity analysis not to significantly bias the results even for the two regions having the most noncentral capitals and the widest latitude range, QLD and NT. However, we considered only current residence and could not take into account the contribution of prior residence areas, or the timing of the critical UVR exposure, which may have occurred earlier in life, or even possible migration after disease initiation, as prevalence rather than incidence data were used. In addition, we could not control for latitude when examining the relationships between the immune disorders and UVR because of collinearity between UVR and latitude.

Potential ecologic confounders include other possible causal factors for immune-related disorders that vary with latitude or UVR; for example, regional infection patterns that may be associated with climatic differences. Infectious agents have been linked to the etiology of immune disorders, including diabetes (Dahlquist 1998; EURODIAB 2000; Kamradt and Mitchison 2001; Notkins and Lernmark 2001), RA (Albert and Inman 1999; Ebringer and Wilson 2000; Wilson et al. 2000), and asthma (Blasi et al. 2001). A lack of infections in early life may also adversely affect immune development--the so-called hygiene hypothesis (Bach 2002). A second important ecologic factor to consider for asthma is that there are marked regional differences in allergen levels. For example, an 11-fold higher level of mean house dust mite allergen (Der p1) concentration in homes in Sydney, New South Wales (low latitude, warm and more humid) compared with TAS (higher latitude, cool and dry) (Couper et al. 1998) may contribute to the inverse latitude gradient for asthma. Other confounding factors could include environmental temperature and dietary differences (Dahlquist 1998). These potential confounders warrant more detailed consideration in future research. There may also be interaction between some of these environmental exposures, for example, infection and nutritional factors (Haverkos 1997). However, it is not possible to determine such interactions in ecologic analyses, as there is a lack of data on joint environmental exposures at the individual level (Ponsonby et al. 2002

UVR has an inverse, or lack of, association with the immune-related disorders analyzed except for asthma, which showed a positive association between prevalence and UVR. We are uncertain of the significance of this finding, particularly as the prevalence increase over the UVR range was only low. UVR-induced Th2 upregulation is one possibility, but UVR can also suppress Th2 responses (Garssen et al. 1999). In addition, the contribution of regional allergen levels to asthma prevalence could be important. Furthermore, much of asthma at the population level may be nonallergic (Pearce et al. 1999), with Th2 mechanisms not involved.

The stronger association found between latitude and type 1 diabetes prevalence in Australia is consistent with similar incidence gradients found in Western Europe and North America (DERIG 1988; Karges et al. 1995; Karvonen et al. 1993) and also within China (Yang et al. 1998). The corresponding inverse association between UVR levels in Australia and type 1 diabetes prevalence is also consistent with previous photoimmunologic work showing that ultraviolet B irradiation has systemic as well as local immunosuppressive effects in humans and animals (Clydesdale et al. 2001; Duthie et al. 1999; Garssen et al. 1999; Sleijffers et al. 2001), as does also ultraviolet A irradiation (Nghiem et al. 2001). UVR exposure may be protective against Th1-mediated disorders such as type 1 diabetes by downregulating Th1 autoimmune responses by several different immunoregulatory mechanisms (Clydesdale et al. 2001; Duthie et al. 1999; Ponsonby et al. 2002

One of these possible mechanisms for downregulation involves UVR-induced vitamin D. Thus, the proposed protective role of UVR for autoimmune type 1 diabetes may act through its important role in vitamin D synthesis in the skin. Some 90% of plasma vitamin D in humans is produced endogenously via skin exposure to UVR in sunshine (Norris 2001) and particularly so in Australia where foods are not generally fortified with vitamin D (Mason and Diamond 2001; Pasco et al. 2001). There are also recent reports of vitamin D deficiency in some Australian populations (Grover and Morley 2001), particularly in winter (McGrath et al. 2001; Pasco et al. 2001). For type 1 diabetes, this possible mechanism is consistent with recent reports of decreased risk of this disorder in offspring of mothers supplemented with cod liver oil in pregnancy (Stene et al. 2000), and decreased incidence of type 1 diabetes in children supplemented with vitamin D in infancy (EURODIAB 1999; Hypponen et al. 2001). The Finnish birth cohort study reported a rate ratio of 0.12 (95% CI, 0.03-0.51) for diagnosis of type 1 diabetes by adulthood comparing regular versus no vitamin D supplementation in the first year of life (Hypponen et al. 2001). These studies suggest that vitamin D may prevent initiation of type 1 diabetes, and thus accord with the vitamin D-protective hypothesis proposed for this and other autoimmune disorders such as MS (Cantorna 2000; Hayes et al. 1997). For MS this hypothesis is consistent with the recent finding of a strong inverse gradient between ambient UVR levels and MS prevalence in another Australian ecologic study (van der Mei et al. 2001). Our analogous finding for ambient UVR and type 1 diabetes in Australia supports the specific prediction that autoimmune diseases other than MS, such as type 1 diabetes, should show latitude and/or UVR gradients similar to those seen for MS if these immune disorders are influenced by UVR exposure (McMichael and Hall 1997

In conclusion, our ecologic analysis of data from the 1995 NHS has demonstrated a regional gradient of type 1 diabetes prevalence within Australia that is inversely associated with regional UVR levels measured at local solar noon. The inverse association with UVR is consistent with that found for another autoimmune disease, MS, in Australia and consistent with type 1 diabetes latitudinal gradients seen in the Northern Hemisphere. The finding is also consistent with photoimmunologic evidence of UVR-induced immunosuppression and suggests a beneficial effect of UVR in preventing both these autoimmune conditions. Analytic epidemiology studies investigating risks of type 1 diabetes and other immune disorders in relation to personal UVR exposure in humans are now required (McMichael and Hall 2001).

References Top

1. ABS. 1995a. National Health Survey: Users' Guide, Australia (ABS Catalogue no. 4363.0). Canberra, Australian Capital Territory:Australian Bureau of Statistics.
2. ------. 1995b. National Health Survey: Data Reference Package (ABS Catalogue no. 4363.0.001). Canberra, Australian Capital Territory:Australian Bureau of Statistics.
3. ------. 1995c. National Health Survey: Summary Results, Australian States and Territories (ABS Catalogue no. 4368.0). Canberra, ACT:Australian Bureau of Statistics.
4. Albert LJ, Inman RD. 1999. Molecular mimicry and autoimmunity. N Engl J Med 341:2068–2074. Find this article online
5. Bach JF. 2002. The effect of infections on susceptibility to autoimmune and allergic diseases. N Engl J Med 347:911–920. Find this article online
6. Blasi F, Cosentini R, Tarsia P, Capone P, Allegra L.. 2001. Atypical pathogens and asthma: can they influence the natural history of the disease? Monaldi Arch Chest Dis 56:276–280. Find this article online
7. Cantorna MT. 2000. Vitamin D and autoimmunity: is vitamin D status an environmental factor affecting autoimmune disease prevalence? Proc Soc Exp Biol Med 223:230–233. Find this article online
8. Clydesdale GJ, Dandie GW, Muller HK. 2001. Ultraviolet light induced injury: immunological and inflammatory effects. Immunol Cell Biol 79:547–568. Find this article online
9. Couper D, Ponsonby AL, Dwyer T. 1998. Determinants of dust mite allergen concentrations in infant bedrooms in Tasmania. Clin Exp Allergy 28:715–723. Find this article online
10. Dahlquist G.. 1998. The aetiology of type 1 diabetes: an epidemiological perspective. Acta Paediatr Suppl 425:5–10. Find this article online
11. Damian DL, Halliday GM, Taylor CA, Barnetson RS. 1998. Ultraviolet radiation induced suppression of Mantoux reactions in humans. J Invest Dermatol 110:824–827. Find this article online
12. Davidson A, Diamond B.. 2001. Autoimmune diseases. N Engl J Med 345:340–350. Find this article online
13. DERIG 1988. Geographic patterns of childhood insulin-dependent diabetes mellitus. Diabetes Epidemiology Research International Group. Diabetes 37:1113–1119. Find this article online
14. Duthie MS, Kimber I, Norval M. 1999. The effects of ultraviolet radiation on the human immune system. Br J Dermatol 140:995–1009. Find this article online
15. Ebringer A, Wilson C.. 2000. HLA molecules, bacteria and autoimmunity. J Med Microbiol 49:305–311. Find this article online
16. EURODIAB 1999. Vitamin D supplement in early childhood and risk for Type I (insulin-dependent) diabetes mellitus. The EURODIAB Substudy 2 Study Group. Diabetologia 42:51–54. Find this article online
17. ------ 2000. Infections and vaccinations as risk factors for childhood type I (insulin-dependent) diabetes mellitus: a multicentre case-control investigation. EURODIAB Substudy 2 Study Group. Diabetologia 43:47–53. Find this article online
18. Garssen J, Vandebriel RJ, De Gruijl FR, Wolvers DA, Van Dijk M, Fluitman A, et al. 1999. UVB exposure-induced systemic modulation of Th1- and Th2-mediated immune responses. Immunology 97:506–514. Find this article online
19. Gies HP, Roy C, McLennan A, Tomlinson D. 1999. Trends in ultraviolet radiation. In: Trends in Sun Protection Seminar. Melbourne, Australia:Anti-Cancer Council of Victoria, 1-19.
20. Grover SR, Morley R. 2001. Vitamin D deficiency in veiled or dark-skinned pregnant women. Med J Aust 175:251–252. Find this article online
21. Hammond SR, English DR, McLeod JG. 2000. The age-range of risk of developing multiple sclerosis: evidence from a migrant population in Australia. Brain 123(pt 5):968–974. Find this article online
22. Haverkos HW. 1997. Could the aetiology of IDDM be multifactorial? Diabetologia 40:1235–1240. Find this article online
23. Hayes CE, Cantorna MT, DeLuca HF. 1997. Vitamin D and multiple sclerosis. Proc Soc Exp Biol Med 216:21–27. Find this article online
24. Hersey P, Haran G, Hasic E, Edwards A.. 1983. Alteration of T cell subsets and induction of suppressor T cell activity in normal subjects after exposure to sunlight. J Immunol 131:171–174. Find this article online
25. Hypponen E, Laara E, Reunanen A, Jarvelin MR, Virtanen SM. 2001. Intake of vitamin D and risk of type 1 diabetes: a birth-cohort study. Lancet 358:1500–1503. Find this article online
26. Kamradt T, Mitchison NA. 2001. Tolerance and autoimmunity. N Engl J Med 344:655–664. Find this article online
27. Karges WJ, Ilonen J, Robinson BH, Dosch HM. 1995. Self and non-self antigen in diabetic autoimmunity: molecules and mechanisms. Mol Aspects Med 16:79–213. Find this article online
28. Karvonen M, Tuomilehto J, Libman I, LaPorte R.. 1993. A review of the recent epidemiological data on the worldwide incidence of type 1 (insulin-dependent) diabetes mellitus. World Health Organization DIAMOND Project Group. Diabetologia 36:883–892. Find this article online
29. Kelly DA, Walker SL, McGregor JM, Young AR. 1998. A single exposure of solar simulated radiation suppresses contact hypersensitivity responses both locally and systemically in humans: quantitative studies with high-frequency ultrasound. J Photochem Photobiol B 44:130–142. Find this article online
30. Kero J, Gissler M, Hemminki E, Isolauri E.. 2001. Could TH1 and TH2 diseases coexist? Evaluation of asthma incidence in children with coeliac disease, type 1 diabetes, or rheumatoid arthritis: a register study. J Allergy Clin Immunol 108:781–783. Find this article online
31. Lemus-Deschamps L, Rikus L, Gies HP. 1999. The operational Australian ultraviolet index forecast 1997. Meteorol Appl 6:241–251. Find this article online
32. Mackay IR. 2000. Science, medicine, and the future: tolerance and autoimmunity. Bmj 321:93–96. Find this article online
33. Mason RS, Diamond TH. 2001. Vitamin D deficiency and multicultural Australia. Med J Aust 175:236–237. Find this article online
34. McGrath JJ, Kimlin MG, Saha S, Eyles DW, Parisi AV. 2001. Vitamin D insufficiency in south-east Queensland. Med J Aust 174:150–151. Find this article online
35. McLeod JG, Hammond SR, Hallpike JF. 1994. Epidemiology of multiple sclerosis in Australia. With NSW and SA survey results. Med J Aust 160:117–122. Find this article online
36. McMichael AJ, Hall AJ. 1997. Does immunosuppressive ultraviolet radiation explain the latitude gradient for multiple sclerosis? Epidemiology 8:642–645. Find this article online
37. ------ 2001. Multiple sclerosis and ultraviolet radiation: time to shed more light. Neuroepidemiology 20:165–167. Find this article online
38. McNally NJ, Williams HC, Phillips DR, Strachan DP. 2000. Is there a geographical variation in eczema prevalence in the UK? Evidence from the 1958 British Birth Cohort Study. Br J Dermatol 142:712–720. Find this article online
39. Miller DH, Hammond SR, McLeod JG, Purdie G, Skegg DC. 1990. Multiple sclerosis in Australia and New Zealand: are the determinants genetic or environmental? J Neurol Neurosurg Psychiatry 53:903–905. Find this article online
40. Nghiem DX, Kazimi N, Clydesdale G, Ananthaswamy HN, Kripke ML, Ullrich SE. 2001. Ultraviolet A radiation suppresses an established immune response: implications for sunscreen design. J Invest Dermatol 117:1193–1199. Find this article online
41. Norris JM. 2001. Can the sunshine vitamin shed light on type 1 diabetes? Lancet 358:1476–1478. Find this article online
42. Notkins AL, Lernmark A. 2001. Autoimmune type 1 diabetes: resolved and unresolved issues. J Clin Invest 108:1247–1252. Find this article online
43. Pasco JA, Henry MJ, Nicholson GC, Sanders KM, Kotowicz MA. 2001. Vitamin D status of women in the Geelong Osteoporosis Study: association with diet and casual exposure to sunlight. Med J Aust 175:401–405. Find this article online
44. Pearce N, Pekkanen J, Beasley R.. 1999. How much asthma is really attributable to atopy? Thorax 54:268–272. Find this article online
45. Platts-Mills T, Vaughan J, Squillace S, Woodfolk J, Sporik R.. 2001. Sensitisation, asthma, and a modified Th2 response in children exposed to cat allergen: a population-based cross-sectional study. Lancet 357:752–756. Find this article online
46. Platts-Mills TA. 2002. The role of allergens in the induction of asthma. Curr Allergy Asthma Rep 2:175–180. Find this article online
47. Ponsonby AL, McMichael A, van der Mei I. 2002. Ultraviolet radiation and autoimmune disease: insights from epidemiological research. Toxicology 181-182:71–78. Find this article online
48. Reynolds NJ, Franklin V, Gray JC, Diffey BL, Farr PM. 2001. Narrow-band ultraviolet B and broad-band ultraviolet A phototherapy in adult atopic eczema: a randomised controlled trial. Lancet 357:2012–2016. Find this article online
49. Selgrade MK, Repacholi MH, Koren HS. 1997. Ultraviolet radiation-induced immune modulation: potential consequences for infectious, allergic, and autoimmune disease. Environ Health Perspect 105:332–334. Find this article online
50. Sleijffers A, Garssen J, de Gruijl FR, Boland GJ, van Hattum J, van Vloten WA, et al. 2001. Influence of ultraviolet B exposure on immune responses following hepatitis B vaccination in human volunteers. J Invest Dermatol 117:1144–1150. Find this article online
51. Stene LC, Ulriksen J, Magnus P, Joner G. 2000. Use of cod liver oil during pregnancy associated with lower risk of Type I diabetes in the offspring. Diabetologia 43:1093–1098. Find this article online
52. van der Mei IA, Ponsonby AL, Blizzard L, Dwyer T. 2001. Regional variation in multiple sclerosis prevalence in Australia and its association with ambient ultraviolet radiation. Neuroepidemiology 20:168–174. Find this article online
53. Weinshenker BG. 1996. Epidemiology of multiple sclerosis. Neurol Clin 14:291–308. Find this article online
54. Wilson C, Tiwana H, Ebringer A.. 2000. Molecular mimicry between HLA-DR alleles associated with rheumatoid arthritis and Proteus mirabilis as the aetiological basis for autoimmunity. Microbes Infect 2:1489–1496. Find this article online
55. Yang Z, Wang K, Li T, Sun W, Li Y, Chang YF, et al. 1998. Childhood diabetes in China. Enormous variation by place and ethnic group. Diabetes Care 21:525–529. Find this article online