Environmental pesticide exposure and non-Hodgkin lymphoma survival: a population-based study

Non-Hodgkin lymphoma (NHL), a diverse group of malignant neoplasms variously derived from the clonal expansion of B, T, or natural killer cells, is the most common hematologic malignancy, constituting 4.3% of new neoplasms in the United States and 3.3% of all cancer deaths [1]. It is generally associated with a favorable prognosis, with a 5-year relative survival of 73.2% [1]. While the etiology of most NHLs continues to be largely unknown, its pathogenesis likely represents a complex process involving the accumulation of multiple genetic mutations, together with underlying immune dysfunction [2].

Pesticide includes a diverse group of chemicals or biologic agents applied in agricultural, industrial/commercial, and residential settings to control pests [3, 4]. In the United States alone in 2012, over 1.1 billion pounds of pesticide was utilized [5]. Despite its widespread usage, pesticide exposure has been associated with multiple adverse health effects, including cancer development [6].

There is evidence indicating that pesticide exposure is a risk factor for NHL development. Occupational agricultural exposure has been associated with a higher incidence rate of NHL in multiple studies [7‐12] and the International Agency for Research on Cancer determined exposure to three commonly used agricultural organophosphates (glyphosate, malathion, and diazinon) to be carcinogenic and associated with NHL development in 2015 [13]. Genotoxicity and reactive oxygen species generation in healthy cells by pesticide used for agricultural production is likely involved in the transformation of healthy lymphocytes into clonal ones. Prolonged environmental exposure to pesticide could lead to clonal expansion of cells with specific tumorigenic alterations and immunodeficiency, potentially contributing to pathogenesis and resistance [14‐16]. Chronic exposure could also lead to adaptive mechanisms for DNA repair and antioxidant activity, undermining chemotherapy regimens [17]. Furthermore, there is a possibility that carcinogenic effects might not result from direct mutagenic or genotoxic mechanisms from the pesticide itself. For example, endocrine disruption or an indirect initiation of inflammation must be considered as tumor-promoting mechanisms of pesticide [18].

While emphasis has been placed on evaluating pesticide as a causative agent for NHL development, investigation on its association with NHL prognosis also has the potential to provide important insight into NHL etiology and pathogenesis. Only one study thus far has investigated the association between pesticide exposure and NHL treatment outcomes. In a French multicenter retrospective cohort study of 244 patients with diffuse large B-cell lymphoma (DLBCL), occupational exposure to pesticide was found to be associated with increased treatment failure and poorer event-free and overall survival compared to those not exposed to pesticide [19]. These findings have yet to be validated and suggest additional adverse implications of pesticide exposure, rationalizing further study. Therefore, we identified NHL patients from the California Cancer Registry and linked these individuals with the statewide pesticide use reporting database to determine the association of pesticide exposure with lymphoma-specific and overall survival.

We identified 35,808 patients between 2010 and 2016 with a first primary diagnosis of NHL who met our criteria. Baseline demographics are shown in Table 1. Overall, 44.2% of all NHL patients were exposed to pesticide in their census tract of residence. Pesticide exposure was higher in Hispanic/Latino (46.5%) and non-Hispanic white (45.6%) than Asian/Pacific Islander (37.2%) and African American (34.9%) patients with NHL. Almost half (46.8%, 45.8%, 45.4%, 44.4%, and 42.2%) of NHL patients with Burkitt lymphoma, follicular lymphoma, small lymphocytic lymphoma, DLBCL, and T/NK-cell neoplasms had pesticide exposure, respectively.

Forty-four percent of patients with both limited and advanced NHL stage had pesticide exposure. Forty percent of NHL patients in the highest SES quintile had pesticide exposure compared to 44.2% in the lowest SES quintile. Pesticide exposure was found among 67.4% of NHL patients in rural and 40.9% in urban MSSA. Glyphosate, organophosphorus, carbamate, phenoxyherbicide, and 2,4-dimethylamine salt exposure were observed in 34.1%, 26.0%, 10.6%, 14.0%, and 12.8% of patients, respectively.

Median follow-up time was 2.99 years (95% CI: 2.95–3.04) for all patients with a primary diagnosis of NHL. Using CalEnviroScreen 3.0 data, we demonstrated no association between total pesticide exposure and worse lymphoma-specific survival (HR: 0.97; 95% CI: 0.92–1.02) or overall survival (HR: 0.97; 95% CI: 0.93–1.02). In addition, pesticide exposure was not associated with worse lymphoma-specific and overall survival when stratified by NHL subtypes (Table 2) or race/ethnicity (data not shown in Tables).

With the integration of land-use surveys which allowed for analysis of the effects of specific pesticides, glyphosate, organophosphorus, carbamate, phenoxyherbicide and 2,4-dimethylamine salt exposure was not associated with worse lymphoma-specific or overall survival (lymphoma-specific survival: glyphosate HR: 1.01; 95% CI: 0.96–1.06, organophosphorus HR: 0.97; 95% CI: 0.92–1.02, carbamate HR: 0.95; 95% CI: 0.88–1.03, phenoxyherbicide: 1.00; 95% CI: 0.94–1.08, 2,4-dimethylamine salt HR: 0.99; 95% CI: 0.92–1.07; overall survival: glyphosate HR: 1.03; 95% CI: 0.96–1.07, organophosphorus HR: 0.99; 95% CI: 0.95–1.03, carbamate HR: 0.96; 95% CI: 0.90–1.03, phenoxyherbicide: 1.03; 95% CI: 0.97–1.09, 2,4-dimethylamine salt HR: 1.01; 95% CI: 0.96–1.08) (Figs. 1 and 2). When examined by race/ethnicity, no consistent association between varying levels of pesticide exposure and lymphoma-specific or overall survival was found. Furthermore, when examined by pesticide tertile levels, all hazard ratios showed no association between varying levels of pesticide exposure and lymphoma-specific survival or overall survival.

When examined by NHL histology, most pesticide exposure was not associated with worse lymphoma-specific survival, with the exception of carbamate exposure in lymphoblastic lymphoma (HR: 3.87; 95% CI: 1.66–9.01) (Table 3). With further stratification of carbamate exposure into tertile levels among lymphoblastic lymphoma patients, mid (HR: 6.77; 95% CI: 2.20–20.80), but not low (HR: 1.37; 95% CI: 0.18–10.57) or high (HR: 4.19; 95% CI: 0.96–18.26) tertile levels were associated with worse lymphoma-specific survival (Additional file 1: Table S1). In contrast, phenoxyherbicide and 2,4-dimethylamine salt exposure was associated with superior lymphoma-specific survival in mantle cell lymphoma with hazard ratios of 0.68 (95% CI: 0.50–0.92) and 0.71 (95% CI: 0.92–1.02), respectively. The effect persisted with further stratification of phenoxyherbicide and 2,4-dimethylamine salt exposure into tertile levels among mantle cell lymphoma patients with superior lymphoma-specific survival observed with low (HR 0.51; 95% CI: 0.31–0.85), but not mid (HR 0.85; 95% CI: 0.50–1.45) or high (HR 0.77; 95% CI: 0.48–1.24) phenoxyherbicide tertiles and low (HR 0.47; 95% CI: 0.27–0.82), but not mid (HR 1.05; 95% CI: 0.62–1.78) nor high (HR 0.82; 95% CI: 0.49–1.35) 2,4-dimethylamine salt tertiles.

Pesticide exposure was not associated with worse overall survival in any of the NHL histologies examined. In contrast, phenoxyherbicide exposure was associated with superior overall survival in mantle cell lymphoma (HR 0.74; 95% CI: 0.57–0.95). With further stratification of phenoxyherbicide exposure in tertile levels among mantle cell lymphoma patients, low (HR 0.63; 95% CI: 0.43–0.94), but not mid (HR 0.77; 95% CI: 0.47–1.24) nor high tertiles (HR 0.86; 95% CI: 0.58–1.27) were also associated with superior overall survival.

In this large population-based study, we observed pesticide exposure to be prevalent among patients diagnosed with NHL, with high pesticide exposure particularly observed in Hispanics/Latinos and non-Hispanic whites. Using multiple measures of agricultural pesticide exposure, we did not observe pesticide exposure to be consistently associated with worse lymphoma-specific or overall survival.

Our study observed pesticide exposure among 46.5% of Hispanics/Latinos and 45.6% of non-Hispanic white NHL patients. Given the racial/ethnic distribution of hired agricultural workers in the United States (57% Hispanic/Latino and 32% non-Hispanic white), we would expect Hispanics/Latinos to have the highest exposure level. However, our neighborhood-level pesticide exposure does not capture occupational exposures outside the neighborhood or exposure to non-agricultural pesticides within the neighborhood, limiting our ability to fully capture pesticide exposure levels. Indeed, our analysis showed no consistent associations between pesticide exposure and worse lymphoma-specific and overall survival among each race/ethnicity.

Our findings are supported by an analysis of 232 NHL patients from the Danish Diet, Cancer and Health cohort which found no association between organochlorine pesticide concentrations in adipose tissue and survival [42]. In contrast, a French multicenter retrospective cohort study of 244 DLBCL patients treated with anthracycline-based immunochemotherapy and with complete occupational histories observed increased treatment failure and poorer event-free and overall survival with occupational pesticide exposure [19]. The survival disadvantage was especially notable in patients with agricultural occupations compared with other exposed patients. In addition, these outcome differences persisted even though patients without pesticide exposure were more likely to present with disseminated disease and patients with and without exposure had no differences in demographic characteristics, initial presentation, international prognostic index, and treatment protocol. Our differences in findings may relate to the type of study design or our inability to classify groups with high personal pesticide exposure; our study included an unselected population of nearly all lymphoma patients in California, whereas the French study included a much smaller sample of patients from 6 medical centers. Given potential associations between pesticide exposure and NHL prognosis, more detailed studies are warranted.

Pesticides exert their toxic effect through various mechanisms such as reactive oxidative damage, chromosomal damage, and alteration of DNA repair mechanisms. Cytogenetic analysis of peripheral blood lymphocytes in 120 Serbian people exposed to pesticide demonstrated higher levels of chromosomal aberrations when compared to those without pesticide exposure [43]. Similarly, evaluation of bone marrow samples from commercial, family, and organic farmers showed higher levels of chromosomal abnormalities in commercial and family farmers compared to organic farmers [44]. In addition, low concentrations of an organophosphorus, malathion, although not associated with cell death, was shown to upregulate oncogenes, downregulate tumor suppressor genes and induce gene expression of human B and T lymphocytes and immunoglobulin production which are involved in the initiation, progression, and pathogenesis of cancer [45]. These results highlight the need for further correlative and molecular studies, including DNA damage signature analysis and RNA sequencing, to identify biomarkers associated with pesticide exposure.

There are limitations to our study. Studies pertaining to the effects of pesticide are investigated worldwide due to its widespread use; pesticide usage patterns likely vary with different locations and the relative importance of residential pesticide exposure is limited to areas with prevalent pesticide use. Our study also utilized census tract as a proxy for pesticide exposure because of a lack of more detailed patient-reported data, such as direct vs. indirect exposure to pesticide (e.g., occupational exposure vs. exposure via aerosol or food contamination), differences in routine activity among patients in the same census tract, or duration of patient residency in a specified census tract. These factors are important in understanding the association between pesticide exposure and NHL. Furthermore, although the PUR database is regarded as one of the most extensive worldwide, guidelines for pesticide use reporting according to the California Department of Pesticide Regulation may limit its utility, as it does not capture details of residential exposure, which accounts for up to 10% of overall pesticide exposure in the United States [5]. Another limitation includes our short follow-up time (~3 years) for lymphoma-specific and overall survival which may underestimate the potential association between pesticide exposure and NHL prognosis.

Strengths of this study include the large population-based cohort of 35,808 patients with NHL which allowed for a robust multivariate analysis of outcomes. In addition, this study used pesticide usage data available from state resources which is an objective measurement of pesticide usage. Lastly, this is one of the few studies that specifically investigate the association between pesticide exposure and NHL treatment outcomes and survival, as previous studies primarily focused on the association between pesticide exposure and NHL incidence.

In summary, pesticide exposure is common among patients with NHL, particularly among Hispanics/Latinos and non-Hispanic whites. Although we found no consistent associations between agricultural pesticide exposure at the neighborhood level and worse survival, additional investigations with more details of personal pesticide exposure and longer follow-up time are warranted based on potential associations between pesticide exposure and NHL prognosis.