The study was conceived and designed by M.T.v.L. and C.M.V. The data were obtained from data custodians by M.T.v.L. and C.M.V. All authors contributed to the analytical plans. The data were analyzed by M.T.v.L. and M.O.F. M.T.v.L. and C.M.V. drafted the manuscript. All authors reviewed, revised and approved the final draft.
Correspondence to: Marina T. van Leeuwen, NHMRC Postdoctoral Research Fellow, Adult Cancer Program, Lowy Cancer Research Centre, Prince of Wales Clinical School, The University of New South Wales, Sydney NSW 2052, Australia, Tel.: 612–9385 8638, Fax: 612–9385-1430, E-mail: firstname.lastname@example.org
Given the uncertainty surrounding solar ultraviolet radiation (UVR) exposure and risk of lymphoid neoplasms, we performed an ecological analysis of national Australian data for incident cases diagnosed between 2002 and 2006. Subtype-specific incidence was examined by latitude band (<29°S, 29–36°S, ≥37°S), a proxy for ambient UVR exposure, using multiple Poisson regression, adjusted for sex, age-group and calendar year. Incidence increased with distance from the equator for several mature B-cell non-Hodgkin lymphomas, including diffuse large B-cell [incidence rate ratio (IRR) = 1.37; 95% confidence interval (CI): 1.16–1.61 for latitude ≥37°S relative to <29°S], lymphoplasmacytic (IRR = 1.34; 95% CI: 1.12–1.61), mucosa-associated lymphoid tissue (IRR = 1.32; 95% CI: 0.97–1.80) and mantle cell lymphoma (IRR = 1.29; 95% CI: 1.05–1.58), as well as plasmacytoma (IRR = 1.52; 95% CI: 1.09–2.11) and plasma cell myeloma (IRR = 1.15; 95% CI: 1.03–1.27). A similar pattern was observed for several mature cutaneous T-cell neoplasms, including primary cutaneous anaplastic large cell lymphoma (IRR = 4.26; 95% CI: 1.85–9.84), mycosis fungoides/Sézary syndrome (IRR = 1.72; 95% CI: 1.20–2.46), and peripheral T-cell lymphoma not otherwise specified (NOS) (IRR = 1.53; 95% CI: 1.17–2.00). Incidence of mixed cellularity/lymphocyte-depleted (IRR = 1.60; 95% CI: 1.16–2.20) and nodular sclerosis Hodgkin lymphoma (IRR = 1.57; 95% CI: 1.33–1.85) also increased with distance from the equator. Many of these subtypes have a known association with infection or immune dysregulation. Our findings support a possible protective effect of UVR exposure on the risk of several lymphoid neoplasms, possibly through vitamin D-related immune modulation critical in lymphomagenesis.
It was initially suggested that exposure to solar ultraviolet radiation (UVR) might increase the risk of non-Hodgkin lymphoma (NHL). This theory was based on the known systemic immunosuppressive effects of UVR and the increased risk of NHL and UVR-related cutaneous neoplasms, non-melanoma skin cancer (NMSC) and cutaneous melanoma, in immunosuppressed individuals.1992 In addition, geographical and temporal similarities in the incidence of NHL, NMSC and melanoma,1994 and the increased risk of NHL in patients with these skin cancers,1995 led to the hypothesis of UVR exposure as a shared etiological determinant.
A causative effect of UVR exposure on NHL risk has not been substantiated. Ecological evidence relating NHL incidence and mortality rates to residential ambient UVR, often approximated by latitude, has been conflicting; a positive association was reported by early studies from the United Kingdom1996 and Europe,1996 while a negative association was later reported by studies from the United States.1995, 1996, 2006, 2004, 2010 More recently, observational studies have also indicated a protective effect of UVR exposure on NHL risk based on a variety of measures. An inverse association between NHL risk and level of residential ambient UVR has been reported by case–control2010, 1997 and some,2011 although not all,1999, 2011 cohort studies and an inverse association between NHL risk and recalled recreational sun exposure was reported by a pooled analysis of case–control studies from Australia, Europe and the United States.2008
The inconsistent evidence for a causative or protective effect of UVR exposure on NHL risk might be related to etiological heterogeneity by NHL subtype, although subtype-specific data are few. In addition, there has been limited ecological assessment of the relationship between Hodgkin lymphoma (HL) and ambient UVR,2006, 2006 and cohort evidence is inconclusive.2011 To better elucidate the association between UVR exposure and risk of both NHL and HL, we used population-based registry data from Australia between 2002 and 2006 to examine subtype-specific incidence of lymphoid neoplasms by residential latitude at diagnosis.
Material and Methods
National data for all incident lymphoid neoplasms (ICD-O-3 9590–9734, 9760–9764, 9820–9837, 9940, 9948), as notified by statutory obligation to Australian cancer registries, were obtained for the period 2002–2006 (n = 34,712). Counts of incident cases were stratified by calendar year of diagnosis, sex and five-year age-group, as well as by latitude band (<29°S, 29–36°S, ≥37°S), where latitude could be assigned based on residential postal area at diagnosis (n = 34,676; 99.9%). Latitude is an accepted proxy for ambient UVR exposure in Australia; these latitude bands broadly correlate with annual average measures of biologically effective UVR, i.e., <29°S represents the highest level of UVR and ≥37°S the lowest level of UVR.2004 They also demarcate Australia's three most populous cities, Brisbane, Sydney and Melbourne, and have been used for previous Australian studies of cutaneous melanoma2012 and NMSC.2006 Corresponding population estimates were obtained from the Australian Bureau of Statistics.
Data were grouped into lymphoid neoplasm subtypes as defined by the InterLymph nested hierarchical classification for epidemiological research,2010 based on the 2008 WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues. All subtypes of NHL and HL were included, and were grouped according to the highest hierarchical level permitted by the data with the exception of NHL not otherwise specified (NOS), for which lineage was not available. In accordance with the InterLymph nested hierarchical classification, and to enable comparison with other epidemiologic studies, plasma cell neoplasms and lymphoid leukemias were included under the “NHL” rubric.2010 Being a heterogeneous group, lymphoid neoplasms NOS (n = 1,017; 3%) were not examined.
For each subtype with at least 50 cases, Poisson regression was used to compute adjusted incidence rate ratios (IRR) with 95% confidence intervals (CI) by latitude band, adjusting for sex and age-group. Age-groups were defined based on the age-profile of lymphoid neoplasm subtypes: 0–44, 45–64, ≥65 for mature B- and NK-/T-cell NHL subtypes, and 0–14, 15–24, 25–44, 45–64, ≥65 for precursor neoplasms and HL subtypes. IRRs were also adjusted for calendar year of diagnosis. In Australia, overall NHL incidence appears to have stabilized in recent years, although HL incidence may be increasing.2010 To assess the potential effect of any differential trends in incidence by latitude band, tests for statistical interaction between latitude band and calendar year were conducted. To assess the potential effect of differential UVR exposure between males and females, tests for statistical interaction between latitude band and sex were conducted. Similarly, to assess the potential effect of differential UVR exposure between children and adults, tests for statistical interaction between latitude band and age-group were conducted; this investigation was restricted to HL subtypes only due to insufficient cases in children for the mature NHL subtypes.
In all models, the log of the population size was used as an offset and a scale parameter was included to adjust for over-dispersion. Latitude band was modeled as a categorical variable, and as a continuous variable to determine linear trend. Modeling was performed using STATA Version 11.1 (StataCorp, TX). The level of statistical significance was set at p < 0.05.
The study was conducted in accordance with institutional ethics committee approval.
A total of 31,234 cases of NHL was registered during the study period, 25,167 (81%) of which were mature B-cell NHL, 1,422 (5%) were mature NK-/T-cell NHL, and 3,001 (10%) were NHL NOS.
Of the 12 mature B-cell NHL subtypes examined, only chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL) showed a significant inverse association with increasing latitude band or with increasing distance from the equator (p-trend = 0.011; Fig. 1). Six mature B-cell NHL subtypes showed a significant positive association with latitude band, including diffuse large B-cell lymphoma (DLBCL, p-trend < 0.001), lymphoplasmacytic lymphoma (LPL, p-trend = 0.002), mantle cell lymphoma (MCL, p-trend = 0.005), marginal zone lymphoma of mucosa-associated lymphoid tissue (MALT lymphoma, p-trend = 0.037) and the plasma cell neoplasms, plasmacytoma (p-trend = 0.015), and plasma cell myeloma (PCM, p-trend = 0.022). For these subtypes the difference in incidence between the <29°S and ≥37°S latitude bands was typically 30%, but ranged from 15% to 52%, i.e., IRRs ranged from 1.15 to 1.52. It should be noted that although a statistically significant test for trend was detected for MALT lymphoma, CIs were overlapping and this finding could be considered borderline. There was no significant association with latitude band for follicular lymphoma (p-trend = 0.861), splenic marginal zone lymphoma (SMZL, p-trend = 0.840), nodal marginal zone lymphoma (NMZL, p-trend = 0.135, hairy cell leukemia (HCL, p-trend = 0.747), or Burkitt lymphoma (p-trend = 0.164).
There was a significant positive association with increasing latitude band or with increasing distance from the equator for three of the four mature NK-/T-cell NHL subtypes examined (Fig. 2), including primary cutaneous anaplastic large cell lymphoma (C-ALCL, p-trend < 0.001), mycosis fungoides/Sézary syndrome (MF/SS, p-trend = 0.018), and peripheral T-cell lymphoma NOS (PTCL NOS, p-trend = 0.001), although not other types of PTCL (p = 0.965). The difference in incidence between the <29°S and ≥37°S latitude bands ranged from 53% to 426%. Extranodal NK-/T-cell lymphoma (ENNKTL) showed a significant inverse association with latitude band (p-trend = 0.008).
No significant association with latitude band or with distance from the equator was observed for precursor NHL (ICD-O-3 9727-29, 9835-37; n = 1,644; p-trend = 0.190); relative to <29°S, the IRRs were 1.12 (95% CI: 0.98–1.28) for 29–36°S and 1.12 (95% CI: 0.95–1.31) for ≥37°S (data not shown in figure). Similarly, no significant association with latitude band was observed for NHL NOS (ICD-O-3 9591, 9675, 9684, 9832; n = 3,001; p = 0.171); relative to <29°S, the IRRs were 1.67 (95% CI: 1.46–1.93) for 29–36°S and 0.88 (95% CI: 0.74–1.05) for ≥37°S for NHL NOS (data not shown in figure).
There were 2,418 cases of HL registered during the study period, 2,300 (95%) of which were classical HL and 118 (5%) of which were nodular lymphocyte predominant HL (NLPHL). Two classical HL subtypes, mixed cellularity/lymphocyte-depleted (MC/LD, p-trend = 0.009) and nodular sclerosis (p-trend < 0.001), showed a significant positive association with increasing latitude band or with increasing distance from the equator (Fig. 3). For both subtypes, the difference in incidence between the <29°S and ≥37°S latitude bands was around 60%. The remaining HL subtypes, lymphocyte-rich (p-trend = 0.912), classical HL NOS (p-trend = 0.533) and NLPHL (p-trend = 0.394), showed no significant association with latitude band.
No statistically significant interactions were observed between latitude bands in incidence trends over the five-year time period studied with the exception of nodular sclerosis HL, for which an increasing incidence trend was observed for the latitude band ≥37°S, but not for the other latitude bands (p-interaction = 0.035). No statistically significant interactions were observed between latitude and sex or age-group.
In this ecological analysis of incident lymphoid neoplasms registered in Australia between 2002 and 2006, incidence increased with increasing latitude band, or distance from the equator, for many subtypes of mature B-cell and NK-/T-cell NHL, and the major subtypes of classical HL. The gradient was strongest for classical HL, for which incidence was in the order of 60% higher in the most southerly latitude band. These findings provide support for a protective effect of ambient UVR exposure on the risk of several lymphoid neoplasms. The mixed association with UVR exposure by lymphoid neoplasm subtype may reflect etiological heterogeneity.
Inverse associations between ambient UVR and overall NHL incidence and mortality rates have been noted by ecological studies from the United States,1995, 1996, 2006, 2004, 2010 but not the United Kingdom1996 or Europe,1996 from which positive associations have been reported. Observational data have also been diverse.2010, 1997, 2011, 1999, 2011, 2008, 2010 Case–control studies have generally reported inverse associations between overall NHL risk and either residential ambient UVR exposure2010, 1997 or measures of personal sun exposure.2008 For example, the InterLymph Consortium pooled analysis of Australian, European and United States' case–control studies (n = 8,243), reported an inverse association between NHL risk and recalled recreational, although not occupational, sun exposure.2008 An inverse association between residential ambient UVR exposure and NHL risk was reported by the Californian Teachers' Study cohort (n = 629), but not by the United States Nurses' Health Study cohort (n = 1,064), nor by a large, nationwide Swedish cohort based on residential latitude (55°N to 69°N; n = 10,381).1999 No association with NHL risk was reported either with annual number of sunburns, past bathing vacations, or use of artificial tanning devices by a Swedish-Norwegian cohort, although there was a relatively small number of incident cases (n = 158).2010
No prior ecological studies, and few observational studies,2011, 1999, 2011, 2008 have performed NHL subtype-specific analyses. In agreement with our data, an inverse association was reported with residential ambient UVR exposure for DLBCL (n = 155) and multiple myeloma (n = 119) by the Californian Teachers' Study cohort.2011 An inverse association was also reported with recalled recreational sun exposure for DLBCL (n = 2,176), mantle cell (n = 315) and some NK-/T-cell lymphomas (n = 217) by the InterLymph Consortium pooled analysis.2008 In contrast with our data, a positive association with residential ambient UVR exposure was reported for CLL/SLL by the Californian Teachers' Study cohort (n = 125),2011 and no association was reported by the Swedish cohort (n = 4,018).1999 The InterLymph Consortium pooled analysis2008 reported an inverse association with recalled recreational sun exposure for follicular lymphoma (n = 1,642). No significant subtype-specific differences were reported by the Nurses' Health Study cohort despite similar subtype-specific case numbers as seen in the Californian Teachers' Study.2011
Knowledge of the relationship between HL risk and UVR exposure is scarce, and there are no published data by HL subtype. Two prior ecological studies in the United States reported a significant inverse association between HL incidence and mortality rates and ambient UVR.2006, 2006 Consistent with our data, this association was stronger for HL than for NHL, but no association between HL risk and residential ambient UVR exposure was reported by the Californian Teachers' Study cohort,2011 although there were few cases (n = 37). Case–control studies based on personal measures of sun exposure have produced varied results. An inverse association was reported with history of solarium use and sunburns 5–10 years before diagnosis by the Scandinavian lymphoma etiology (SCALE) study (n = 618).2005 An inverse association was also reported with a history of vacations in sunny climates and use of tanning devices in a population-based German case–control study (n = 115)2007 but no association was reported with either childhood or adulthood recreational or occupational exposure by a later multicentre European study (n = 268).2008
The latitude gradient we observed for several subtypes of NHL and HL is opposite that observed in Australia for cutaneous melanoma2012 and NMSC,2006 arguing against a shared etiological role for UVR. One postulated mechanism by which UVR exposure might inhibit, rather than promote, lymphomagenesis is through UVB-induced cutaneous synthesis of vitamin D.2007 Vitamin D has been shown to inhibit proliferation and increase differentiation of a number of normal and cancer cell lines, including B-NHL cell lines, but only at concentrations higher than observed physiologically.1993 Low serum 25-hydroxyvitamin D [25(OH)D] levels have also been associated with poorer event-free and overall survival for patients with DLBCL and T-cell NHL,2010 and with CLL progression.2011 In addition, HL prognosis was found to be better for those diagnosed during the autumn versus winter months, seasonality being strongly correlated with serum 25(OH)D levels.2005
In addition to its potential anti-cancer properties, a broader role for vitamin D is suggested in relation to immune regulation, including modulation of both the adaptive and innate immune response.2011 Interestingly, many B-cell NHL subtypes for which UVR exposure appears to be protective in this study are post germinal-centre, with plasmacytic, plasmablastic or immunoblastic differentiation, suggesting that only those lymphoma types which arise from lymphocytes which have been exposed to antigens may occur at decreased rates. This includes MALT lymphoma, LPL, plasmacytoma, PCM and about 50% of DLBCL; the exception is MCL in which most cases have a pre-germinal centre phenotype.
In addition, the above subtypes and several others for which UVR exposure appears protective, have a known or suspected infectious etiology, or an association with immune dysregulation. For instance, some classical HL subtypes, especially MC/LD and to a lesser extent NS, are causally associated with infection by Epstein-Barr virus (EBV), as is some DLBCL.2009 MALT lymphoma is associated with infection by Helicobacter pylori as well as autoimmune diseases including Sjögren disease and Hashimoto thyroiditis and several mature B-cell lymphomas, including LPL, are associated with infection by hepatitis C virus.2009 A moderate association with Borrelia burgdorferi infection was reported for MCL by the SCALE study (n = 148),2008 although this putative association has not been confirmed. A personal history of infection, inflammatory and autoimmune diseases has been associated with an increased risk of PCM.2011 Chronic antigenic stimulation by an infectious agent and the maintenance of cutaneous inflammation has been suggested in the etiology of MF/SS, though no specific agents have been identified. The role of immune dysfunction is also emphasized by the response to immunosuppressive therapies, including phototherapy.2004 We observed no association with latitude for follicular lymphoma, for which there is no apparent association with infection or immune dysregulation.2010 Nor was a gradient evident for HCL, a neoplasm for which no infectious cause has been identified, and for which BRAF V600E mutations have been almost universally detected.2012
Unlike for the other EBV-related neoplasms, incidence decreased with distance from the equator for the NK-/T-cell NHL subtype, ENNKTL.2009 EBV preferentially infects B-cells, and infection is normally kept in check by an EBV-specific, cytotoxic T-cell response. Vitamin D has been shown to increase the number and proportion of cytotoxic T-cells. On one hand, vitamin D might confer a protective effect on the risk of EBV-related B-cell neoplasms through improved EBV-specific immunity.2012 On the other hand, this could perhaps lead to increased opportunistic potential for NK-/T-cell infection by EBV. Support for this theory comes from evidence showing that while EBV is generally B-lymphotrophic, it can also opportunistically infect other cell types, including epithelial, NK- and T-cells.2011 Race is another factor worthy of consideration. Incidence of ENNKTL is rare in Western countries, and is seen more frequently in East Asia, and Central and South America.2011 However, the latitude gradient observed for ENNKTL in this study does not reflect the geographical distribution of the East Asian population in Australia, arguing against substantial confounding by race.
Incidence also decreased with distance from the equator for CLL/SLL, for which infectious agents, including the Merkel cell polyomavirus,2011 have been implicated. Bacterial infection, evidenced by antibiotic use, might also be important in CLL/SLL pathogenesis.2012 Some under-notification of CLL laboratory diagnoses has been identified in New South Wales,2011 a large Australian state falling within the central latitude band (29–36°S) in our study. Any systematic geographical variation in reporting by pathology laboratories will have influenced our findings for this neoplasm.
Vitamin D-mediated effects on immune regulation have been offered in explanation of the similar, albeit stronger, latitude gradients observed for multiple sclerosis and other autoimmune diseases, including Type 1 diabetes, in Australia and other Western countries.2005 For instance, vitamin D-related modulation of EBV-specific immunity has been suggested in the etiology of multiple sclerosis, although this remains unresolved.2011 Interestingly, inconsistency between countries in the latitude gradients for NHL are also seen for multiple sclerosis; for both, increases in incidence with distance from the equator are seen for mid-latitude countries like Australia, but not for higher latitude countries such as those of Scandinavia and Northern Europe,2007, 2011 although this has not been examined systematically. Comparably higher dietary vitamin D intake by Scandinavian populations is one possible explanation for this discrepancy.2011
Despite biological plausibility, epidemiological evidence of a protective effect of vitamin D on the risk of lymphoid neoplasms is lacking. Recently pooled cohort data failed to detect an association between pre-diagnostic serum 25(OH)D concentrations and overall NHL risk (n = 1,353), or risk of CLL/SLL (n = 401), DLBCL (n = 344) or follicular lymphoma (n = 206)2010 but reliance on a single measure as a surrogate for long-term vitamin D status might be inadequate,2010 and the median age at blood draw (61–62 years) might not have been the etiologically relevant time period. Findings from observational studies measuring dietary vitamin D intake have been negative; no association was reported by the Californian Teachers' Study2011 or the Nurses' Health Study.2011 It is also possible that modulation of the immune system by other UVR-related mechanisms, independent of vitamin D, might be important.2011 Melatonin suppression for instance, also correlated with UVR, has been questioned in the etiology of some cancers, particularly breast cancer.2010
This large, national, population-based ecological study systematically assessed the relationship between ambient UVR exposure, approximated by residential latitude band, and subtype-specific risk of lymphoid neoplasms. It used a sophisticated modeling approach to control for important demographic characteristics and changes in incidence over time. Latitude is a reliable surrogate of biologically effective UVR in Australia,2004 and of serum 25(OH)D concentration in the Australian adult population,2012 but misclassification of exposure is unavoidable and is expected to bias our findings towards the null. Residential latitude at diagnosis may not reflect that of lifetime residence. In addition, it is possible that our findings reflect uncontrolled confounding by demographic, environmental or behavioral factors that vary with latitude and predict the risk of some lymphoid neoplasms. For example, race, strongly associated with both NHL and HL incidence, and also a modifier of vitamin D status, is not uniformly collected by Australian cancer registries and thus could not be included as a covariate in our analyses.
Finally, despite using the most recent hierarchical classification for subtype-specific analyses, specificity was limited for some subtypes. For instance, neoplasms could not be stratified further by immunophenotype, such as by tumor-EBV status, as this characteristic is not routinely collected by the cancer registries. We were therefore unable to examine diversity within subtypes, such as for DLBCL. And despite significant widespread improvements in diagnosis and classification, more than 10% of cases were either NHL NOS or LN NOS.
In conclusion, these data provide support for a possible protective effect of UVR exposure on the risk of several subtypes of lymphoid neoplasm. The observation that this effect was confined mostly to subtypes with an infectious etiology or an association with immune dysregulation is novel and warrants replication. A role for vitamin D-mediated control in the pathogenesis of these lymphoid neoplasms is suggested.
The authors thank the state and territory cancer registries for the use of their data, and the Australian Institute of Health and Welfare for cancer data extraction and mapping. They thank the Australian Bureau of Statistics for the provision of population census data.