Occupational ultraviolet exposure and risk of non-Hodgkin’s lymphomas: a meta-analysis

Non-Hodgkin lymphoma is a heterogeneous group of lympho-proliferative disorders. We performed a meta-analysis to summarize the available evidence from case-control studies and cohort study on the inconsistent association between occupational sun exposure and the risk of non-Hodgkin lymphoma. We searched PubMed, ISI web of science, the Cochrane Library, EMBASE and reference lists for relevant articles. Study specific odds ratios or relative risk and 95% confidence intervals were pooled by using fixed-effects or random-effects models. Ten case-control studies and one cohort study were included in the meta-analysis. Overall, the pooled odds ratios for occupational ultraviolet exposure and non-Hodgkin lymphoma risk was 1.15(95% confidence intervals: 0.99, 1.32; I2 = 44.4%). Occupational sun exposure was positively associated with the risk of NHL 1.14 (95% confidence intervals: 1.05, 1.23; I2=25.4% p for heterogeneity =0.202) in Caucasian population. Common subtypes of non-Hodgkin lymphoma and ultraviolet exposure had the negative results. The pooled odds ratios was 1.16, (95%confidence intervals: 0.90, 1.50) for T-cell non-Hodgkin lymphoma; 0.79, (95%confidence intervals: 0.61, 1.02) for B-cell non-Hodgkin lymphoma; 1.13, (95%confidence intervals: 0.96, 1.34) for chronic lymphocytic leukemia; 1.25, (95%confidence intervals: 0.95, 1.64) for males; 1.49, (95%confidence intervals: 0.99, 2.25) for females. Data suggested that occupational ultraviolet exposure was a risk factor for non-Hodgkin lymphoma in Caucasian population. While, there had no relationship between occupational ultraviolet exposure and risk of non-Hodgkin lymphoma in general population as well as non-Hodgkin lymphoma common subtypes. Besides, gender specific occupational sun exposure also indicated no association on risk of non-Hodgkin lymphoma.


INTRODUCTION
Non-Hodgkin lymphoma (NHL) is a heterogeneous group of lympho-proliferative disorders [1]. As statistic data showed, NHL is the sixth most common cancer in the United States and in the United Kingdom [2,3], and is estimated to be the tenth most common cancer worldwide [3].
As the increasing incidence of NHL mounting studies associated with the relationship between sun exposure and risk of NHL were carried out in westernized countries. But the results have been inconsistent. We therefore undertook a meta-analysis of case-control and cohort studies to quantitatively assess the relationship between occupational sun exposure and risk of NHL and common NHL subtypes.
One study included only women [33] and one study included only men [36], while the rest of the studies did not specify with gender. One study reported the Asian population [35], while the rest of studies focused on Caucasian population. Nine studies reported results for all types of NHL patients, while two studies only included specific types of NHL. The Sweden study included only hairy cell leukemia cases [27]. One European study included only mycosis fungoides cases [37]. The control source of eight studies were population based [27-29, 31-33, 36, 37], two studies were hospital based [30,35], while one study included both population control and hospital based control source [34]. Six studies' data collection method was in-face interview [30,[32][33][34][35]37], while four studies' data collection method was though self-administered questionnaire and followed by a telephone interview [27][28][29]31]. One study used population occupational health service program data [36].
The age of participants were all aged 17 and above. The exposure odds ratios (ORs)/relative risk (RRs) of NHL, the adjustments made for confounding and occupational history assessment were shown in Table 2.

Association between occupational sun exposure and risk of NHL
In the meta-analysis, the summary estimated for occupational sun exposure showed no statistical association between occupational sun exposure and risk of NHL 1.15(95%CI: 0.99, 1.32; I 2 =44.4%). (Figure 2) Among the 11 enrolled studies, three studies showed positive relationship between occupational sun exposure and the risk of NHL [27,29,33]. The ORs differ from 0.75(95%CI: 0.55, 1.03) to 2.30(95%CI: 0.96, 6.20). There was a moderate heterogeneity among these studies (I 2 =44.4%, p for heterogeneity =0.048), therefore we used the random-effect model to calculate the summary OR.
Ten studies reported the Caucasian population. Occupational sun exposure was positively associated with the risk of NHL 1.14 (95%CI: 1.05, 1.23; I 2 =25.4% p for heterogeneity =0.202). No heterogeneity was observed, thus fixed-effect model was provided. (Figure 3) Our results for occupational sun exposure did not display an association in risk for common NHL subtypes. The results of T-cell NHL (ORs: 1.16; 95%CI: 0.90, 1.50) and B-cell NHL (ORs: 0.79; 95%CI: 0.61, 1.02) analyses were presented in Figure 4A and Figure 4B. There are no association in Chronic Lymphocytic Leukemia (CLL) either. The summary ORs for CLL was 1.13 (0.96, 1.34) ( Figure 4C).

Publication bias and sensitivity analyses
In order to evaluate the impact of potential publication bias, we applied the Begg's test (p=0.37) and Egger's test (p=0.37) for the association between occupational sun exposure and the risk of NHL. The results indicated no publication bias among these studies. In addition, no publication bias was detected for the positive association between occupational sun exposure and the risk of NHL in Caucasian population, either (Begg's test: p=0.53; Egger's test: p=0.14). To investigate heterogeneity in our meta-analysis, we evaluated sensitivity analysis within the studies. Eleven studies which included in our meta-analysis were the relationship between occupational sun exposure and risk of NHL, two studies were focused on specific types of NHL [27,37], so we conducted a sensitivity analysis restricted to those nine studies. Results did not change when the aforementioned studies were included or excluded. The pooled OR was 1.09 (95%CI: 0.97, 1.23) with a significant decreased heterogeneity among these nine studies (I 2 =25.1%, p for heterogeneity =0.21 ( Figure 6A). The result was consistently to the overall pooled OR, which suggested our study was reliable.
We applied sensitivity analysis on our positive results among Caucasian population. The pooled estimate did not vary substantially with the exclusion of any single study ( Figure 7).

DISCUSSION
Our meta-analysis of 11 included case-control and cohort studies indicated that occupational ultraviolet exposure was a risk factor for NHL in Caucasian population. However, there is no association between occupational sun exposure and risk of NHL in general population as well as common NHL subtypes. In addition, gender specific study did not show any association either. Most studies included in our meta-analysis showed no association between occupational sun exposure and risk    (continued) exposure increased the risk of NHL. InterLymph organization systematically analyzed ten studies about occupational UV exposure and NHL incidence in 2008, results were consistent with our main meta-analysis [38].
Our main findings showed no association between occupational ultraviolet exposure and NHL risk. When we omitted the study of Asian population our results are positively correlated. Hughes et al reported the pigmentary characteristics and NHL risk. In this study, it categorized the six ethnicity groups including Asian. They have found that the very fair skin compared to brown or olive skin had 44% increased risk of NHL [25]. Interestingly, the Singapore study was not only the Asian population study but also the only low latitudes study [35]. Several studies have yielded the same results. Reduced risk or no effect were found in mid latitudes or low latitudes [28,38], while increased risk was found at higher latitudes [26,[39][40][41]. Grant WB proposes that UVA is a risk factor through impairing the immune system, while UVB is a protection factor through vitamin D production [42]. The ratio of UVA/UVB increases while latitude increases. Thus, the pooled result of our higher latitude Caucasian population studies showed that occupational UV exposure was a risk factor for NHL.
Many of the studies which included in this metaanalysis reported that there is no relationship between occupational UV exposure and NHL risk, however at the same time, reported the daily, casual UV exposure is a protective factor for NHL [28,29,32,34,35]. This Age, smoking, and magnetic field exposure The occupational exposure to sunlight from outdoor work was assessed by an experienced industrial hygienist from the construction industry (N. Hallin). The hygienist classified the sunlight exposure for the job tasks into four categories with exposure scores 0, 1, 2, and 3  result was consistent with InterLymph Organization2008 analysis [43]. One possible explanation is that most studies would have attributed any time duration occupational UV exposure cases into exposure group, while for the individual, it may be exposed to a period of time, rather than continuous exposure [28,35,43]. In addition, farmers often categorized into occupational UV exposure group, but some of these people at the same time contacting potential risk factors NHL such as pesticides, alkylation, etc [30,36]. For example, Smedy et al published that Ever having an outdoor occupation for 1 year or more was associated with a slightly increased risk of non-Hodgkin lymphoma (OR = 1.2, 95% CI = 1.0 to 1.3), but this association was weakened (OR = 1.1, 95% CI = 1.0 to 1.2, ) after additional adjustment for occupational exposure to pesticides [29].  As with any meta-analysis of observational studies, our study has limitations.
First, moderate heterogeneity was found across our main analysis, which can be explained by the multiple differences between studies with regard to the study designs, sample sizes, analysis strategies, participants' baseline characteristics, adjustments for confounders and occupational history assessment methods. For example, the control source of eight studies were population based [27-29, 31-33, 36, 37], and two studies were hospital based [30,35], while one study included both population control and hospital based control source [34]. Sample sizes were different from eleven studies included in this metaanalysis. Two studies had relatively small numbers of participants and specific NHL types [27,37], which raised some concerns regarding the reliability of their results.  Thus, we used the random-effects model to determine the overall estimate of variability.
Secondly, half of the studies in this meta-analysis relied on self-administered questionnaire, while anther half studies gathered information from interview. The participants may have different attitudes and different understanding towards questions under different methods. Besides, the total occupational sun exposure hours are varies across studies and participants. Finally, in a metaanalysis of published studies, the potential publication bias might influence the results, because studies with null results tend not to be published. Nevertheless, our publication bias test showed no possible bias.
In summary, our meta-analysis suggested that occupational UV exposure was a risk factor for NHL in Caucasian population. While, there had no relationship between occupational ultraviolet exposure and risk of NHL as well as NHL common subtypes. Besides, gender specific occupational sun exposure also indicated no association on risk of NHL.

Literature search
We reported this article in accordance with MOOSE (meta-analysis of observational studies in epidemiology) guidelines [44]. We systematically searched four databases: PubMed, ISI web of science, the Cochrane Library and EMBASE for studies published in any languages (up to 2016, August 17th) using the following search items: ultraviolet radiation, ultraviolet ray, ultraviolet light, sunlight exposure or solar ultraviolet exposure combined with non-Hodgkin lymphoma or lymphoid malignancies. The search was restricted to studies of human participants. We also have reviewed the reference lists of all pertinent articles to search for more studies.

Inclusion criteria
To be included in this meta-analysis, studies had to have met the following criteria: (1) NHL cases were medically confirmed by histopathology diagnoses; (2) the study was designed as case-control or cohort study; (3) the occupational sunlight exposure and incidence of NHL were associated; (4) detailed data of odds ratios (ORs) or relative risks (RRs) with 95% confidence interval (CI); (5) All the cases were adult (age≥17 years old). We did not include the studies that only report the death rate of NHL without incidence rate. When there were multiple published reports from the same study population, the most recent or the most informative report was selected for analysis.

Data extraction
We extracted the following information from each study: authors' name, year of publication, study type, study location, sample size (numbers of case patients and control subjects), study period, participation's age, type of control source, assessment of data collection, and statistical covariates adjustment in the analysis. ORs or RRs with corresponding 95%CIs for each study were either extracted directly from the article or calculated from available raw data.

Statistical analysis
To pool the results of individual studies together, we used a general variance-based method in the meta-analysis. The multivariate adjusted ORs and 95% CIs presented in the literature were used. In situations where the incidence is low, the odds ratio approximates the relative risk, therefore, in looking at studies of NHL (a rare condition), it is acceptable to compare OR and RR estimates [45]. The outcomes are presented as a forest plot with 95% CIs.
Statistical heterogeneity among studies was tested with the Q statistic, and statistical inconsistency was quantified with the I 2 statistic [46]. When I 2 was from 0% to 40% along with p >0.10 the heterogeneity might not be important. If the meta-analysis has no heterogeneity, fixedeffects model with the Mantel-Haeszel method would be used to combine the individual studies [47], otherwise, the random-effects method (DerSimonian 1986) was used for pooling [48].
There were two studies focused only on specific types of NHL [27,37]. We further conducted a sensitivity analysis restricted to the rest nine studies to evaluate the stability of the pooled estimates between occupational sun exposure and risk of NHL.
The Egger's test and Begg's test were used to assess for publication bias [49,50]. P<0.05 was considered statistically significant publication bias. All statistical analyses was performed by using STATA (version 11.0; StataCorp, College Station, TX).

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.