Epidemiology of Follicular Lymphoma

INTRODUCTION

Follicular lymphoma (FL) is one of the most common indolent non-Hodgkin lymphoma (NHL) subtype in Western countries. While understanding of the epidemiology of NHL subtypes has been hampered by changes in classification over time, this situation has greatly improved since the implementation of the World Health Organization (WHO) classification for Hematologic and Lymphoma tissues in 2001¹ and the nested classification of lymphoid neoplasms for epidemiologic research from the International Lymphoma Epidemiology Consortium (InterLymph) in 2007.² The diagnosis of FL has high concordance between referral and expert review (85.5%),³ and unlike many NHL subtypes, FL has good agreement between the InterLymph classification and the historical Working Formulation (88.9%),² allowing for a longer time range of epidemiologic studies that can address FL time trends and risk factors. Nevertheless, a historical focus on NHL overall and a lack of NHL subtype information in epidemiologic studies prior to the 1980s along with limited sample sizes has hampered progress. A comprehensive review of the epidemiology of NHL was recently published,⁴ and this review focuses on FL descriptive epidemiology as well as analytic epidemiology based on larger and more definitive studies, especially meta- and pooled analyses.

DESCRIPTIVE EPIDEMIOLOGY

Incidence

In 2016, an estimated 13,960 cases were diagnosed in the US, representing 12.4% of mature NHLs.⁵ For the latter part of the 20^th century, NHL rates were rapidly increasing in Western countries and then stabilized around 2000.⁴ FL incidence in the US increased in the US from 1992-2001.⁶ Using data from the US Surveillance, Epidemiology and End Results (SEER) program,⁷ the age-adjusted incidence rate (using the US year 2000 as the standard population) for FL from 2000-2016 was 3.5 per 100,000, and was 1.2 times higher in men (3.9) than women (3.3). The incidence of FL increases sharply with age (Figure 1). Rates were highest in non-Hispanic whites (4.1), followed by Hispanics of all races (2.9), then non-Hispanic blacks (2.4), American-Indian/Alaska Natives (1.7), and Asian or Pacific Islanders (1.7). As shown in Figure 2, from 2000-2016 the incidence rates have been steady for all ethnicity/race and gender groups with the exception of statistically significant (P<0.05) decreases for non-Hispanic white females (annual percent change, −0.98%) and increases for Hispanic females (+1.01%).

Figure 1.

Incidence rates (per 100,000) by age group for follicular lymphoma, SEER18, United States, 2000-2016.

Figure 2.

Age-adjusted incidence rates (per 100,000) for follicular lymphoma by gender and race/ethnicity (*Non-Hispanic), SEER18, United States, 2000-2016.

There are fewer data to make international comparisons, but in a large international pathology study of the relative frequencies of NHL subtypes from 24 countries, FL accounted for a higher percentage in developed countries (25.5%) compared to developing countries (15.3%).⁸ Compared to age-standardized incidence rates in the US for FL (3.5), rates were lower in Australia in 1997-2006 (3.1),⁹ Europe for 2000-2002 (2.2),¹⁰ the UK for 2004-2012 (2.8),¹¹ Singapore for 2008-2012 (1.0),¹² and Japan for 2008 (1.1),¹³ acknowledging differences in cancer registration, timeframes and standard populations. Age-adjusted FL rates were also slightly higher for females than males in Europe and the UK, but not in the US, Australia, Singapore or Japan. Differences in age-standardized rates for FL by ethnicity were also observed in the UK for 2001-2007, with the highest rates in whites (1.6) then South Asians (1.2), Blacks (0.6), and Chinese (0.6).¹⁴ FL incidence rates stabilized in France for 2000-2009,¹⁵ while they were still increasing in Australia for 1997-2008,⁹ Singapore for 1998-2012¹² and Japan for 1993-2008.¹³ The world age-standardized rates (per 100,000) of FL in Chinese migrants from Hong Kong to British Columbia (1.21) was closer to that of Hong Kong (1.66) than that of non-Chinese in British Columbia (2.97),¹⁶ suggesting potential genetic differences in susceptibility not impacted by migration to a new environment with higher FL rates.

Survival and mortality

The 5-year relative survival rate for FL, which accounts for competing causes of mortality, is summarized in Figure 3 from the SEER Program (18 registries) for cases diagnosed 2000-2016 by race/ethnicity and gender.⁷ Rates ranged from 80-90%, were similar by gender, and were highest for male American Indians/Alaska Natives (90.2%) and lowest for male Hispanics (80.2%). The 5-year relative survival rate for FL in the EUROCARE-5 study¹⁷ of 20 countries increased from 2000-2002 (64.1%) to 2003-2005 (69.0%) to 2006-2008 (74.3%), while in the UK¹¹ it was 86.5% for cases diagnosed 2004-2012. In Singapore,¹² 5-year age-standardized relative survival for FL increased from 1998-2002 (43.8%) to 2003-2007 (64.9%) to 2008-2012 (82.3%). In the SEER Program, overall survival rates for FL improved in all age and sex groups for the period 2001-2009 compared with 1992-2000.¹⁸ Similarly, overall and relative survival improved from 2000-2010 in Sweden in all age and sex groups, particularly elderly women, and correlated with the adoption of first-line rituximab use.¹⁹

Figure 3.

Five-year relative survival rates by gender and race/ethnicity (*Non-Hispanic), SEER18, United States, 2000-2016.

Using a landmark analysis, FL patients who had a progression, relapse, or retreatment within 12 months of diagnosis had poor subsequent overall survival compared to an age and sex-matched background population (standardized mortality ratio (SMR)=3.72, 95%CI 2.78-4.88), while patients who did not have any of these events in the 12 months after diagnosis had no added mortality beyond the background population; for immunochemotherapy-treated patients, this landmark occurred at 24 months after diagnosis.²⁰ These results emphasize the clinical importance of early failures, while patients achieving event-free status at 12/24 months can expect a normal life expectancy.

While relative survival has been increasing since 2000, lymphoma remains the leading cause of death in the first decade after diagnosis in FL patients treated in the rituximab era, with a cumulative risk of mortality of 10.3% at 10 years (and 13.3% when combined with treatment-related mortality).²¹ In contrast, the cumulative risk of mortality as a result of non-lymphoma related causes was 5.1% at 10 years. Notable, the 10-year cumulative risk of lymphoma or treatment-related mortality for patients with an event within 24 months of diagnosis was 36.1% (compared to 6.7% for patients achieving EFS24) and after transformation was 45.9% (compared to 8.1% for patients who did not transform), emphasizing the importance of these events in driving poor outcomes in FL.

Mortality rates for FL are not available, as NHL subtype information is rarely recorded on death certificates. Using SEER and other population data, Howlander and colleagues²² showed that overall NHL mortality rates increased from 1975-1997, and then decreased from 1998-2011 and that incidence-based mortality rates (an approach to link NHL subtype incidence data with mortality data) for FL began to decline 5.3% per year after 1997. In 2011, an estimated 11% of NHL deaths in the US were due to FL.²²

PRECURSORS

The t(14;18) translocation, giving rise to a BCL2-IGH fusion, is a genetic hallmark of FL (>85% of FL have the translocation). It can be been detected at low levels in 50-70% of presumably normal adults, and higher levels of circulating t(14;18) cells (above 1 in 10,000 blood cells) is associated with a 23-fold higher risk of developing FL (95%CI 9.98-67.3).²³ Even with this strong association, the high prevalence of the biomarker relative to the low incidence of FL suggests the t(14;18) event is necessary but not sufficient for follicular lymphomagenesis.

FAMILY HISTORY AND GENETIC SUSCEPTIBILITY

In the InterLymph Subtypes Project, a family history of NHL was associated with a 1.99-fold increased risk of FL (95%CI 1.55-2.54) and this association was not attenuated by adjustment for potential confounding factors.²⁴ This risk estimate aligns with data from a pooled analysis of five Nordic registries from 1955-2010 (standardized incidence ratio (SIR)=1.6, 95%CI 1.4-2.0). Risk of FL was also elevated in persons with a first-degree relative with FL (SIR=2.1, 95%CI 1.3-3.4), DLBCL (SIR=2.6, 95%CI 1.7-3.6), SLL (SIR=3.6, 95%CI 1.0-9.1), MCL (SIR=2.6, 95%CI 0.9-6.1),²⁵ HL (OR=1.47, 95%CI 0.90-2.40),²⁴ multiple myeloma (OR=1.93, 95%CI 1.06-3.51),²⁴ and CLL (OR= 1.6, 95%CI 0.87-2.8).²⁶ While familial aggregation represents both shared genetic and environmental factors, the findings to date suggest that the associations with family history are not strongly confounded by non-genetic risk factors.²⁷ Furthermore, there appears to be shared etiology across multiple subtypes, with many subtypes showing a stronger association with a family history for that specific subtype, suggesting a role for subtype-specific genetic factors.²⁷

Genetic epidemiology studies are used to identify genetic loci. Linkage studies in families are most commonly used to identify genes with major susceptibility effects (e.g., Mendelian), but there are no linkage studies of FL or most major NHL subtypes.²⁷ The lack of major genes in FL and other common NHL subtypes may be due to underpowered studies and the complexity of accurate phenotyping of NHL subtypes over time, but alternatively raises the hypothesis that multiple, low to moderate risk variants that are common in the population (i.e., >5% minor allele frequency) may be more relevant in lymphoma etiology than single, highly penetrant variants that are very rare.²⁸ To identify common genetic variants, the case-control study design is most commonly used, and either evaluates candidate genes (hypothesis driven) or the entire genome simultaneously via an agnostic (hypothesis-free) genome-wide association study (GWAS) approach.²⁷ Prior to GWAS, there were a large number of studies evaluating candidate genes with risk of FL, but most of the studies failed to replicate for a variety of reasons.²⁹ In the largest GWAS of over 2100 FL cases, the HLA region showed a striking association with FL, with the top single nucleotide polymorphism (SNP) at 6p21.32 (rs12195582) reaching P=5.36 × 10⁻¹⁰⁰.³⁰ After imputing HLA alleles and amino acids (AA), the strongest signal mapped to four linked DRβ1 multiallelic AAs at positions 11, 13, 28 and 30, which reside in the peptide binding cleft and impact several binding pockets in DRβ1. As these are key positions that impact allelic binding, these results suggest an important role for DRβ1 peptide presentation in FL etiology. After accounting for this locus, two additional independent, genome-wide signals were identified in HLA class II (rs17203612) and class I (rs3130437, near HLA-C). Outside of the HLA region, loci were identified at 11q23.3 (near CRCX5), 11q24.3 (near ETS1), 3q28 (in LPP), 18q21.33 (near BCL2) and 8q24.21 (near PVT1). These genes all link B-cell biology to follicular lymphomagenesis. In the context of GWAS of other lymphoma subtypes, FL shares loci (although specific SNP results vary) at 6q21.32-33 with DLBCL, CLL, HL, MZL, and NKTCL; 8q24 with DLBCL, CLL, and HL; 3q28 with CLL; and 18q21.33 with CLL.²⁸ These results parallel the family study data, supporting a genetic architecture of risk across subtypes as well as for specific subtypes.

MEDICAL HISTORY RISK FACTORS

Infections

The International Agency for Research on Cancer (IARC) has classified Epstein-Barr virus (EBV), Hepatitis C virus (HCV), Human immunodeficiency virus, type 1 (HIV-1), Kaposi’s sarcoma herpes virus (KSHV), Human T-cell lymphotrophic virus, type 1 (HTLV-1), and Helicobacter pylori as carcinogenic to humans (Group 1) with sufficient evidence that they cause certain types of lymphoma in humans, but only HCV has shown an association with FL.³¹ HCV infection was associated with FL (OR=2.73; 95%CI 2.20-3.38) in a meta-analysis of 542 FL cases and 4041 controls,³² and this association was also observed in a large SEER-Medicare case-control study (OR=1.88; 95%CI 1.17-3.02).³³ However, in an InterLymph pooled analysis of 1181 FL cases and 6269 controls, HCV seropositivity was not associated with FL (OR=1.02; 95%CI 0.65-1.60).³⁴ Treatment of HCV can induce remission of FL.³⁵

In a 2009 evaluation, IARC judged that there was only limited evidence of a causal link of Hepatitis B virus (HBV) with NHL,³⁶ although in a more recent meta-analysis of 17 case-control and 5 cohort studies (over 40,000 NHL cases), HBV infection was associated with an increased risk of NHL overall (OR=2.24, 95%CI 1.80-2.78), but for FL risk was only increased in studies from high HBV prevalent countries (OR=1.66, 95%CI 1.02-2.70).³⁷

Transplantation

Solid organ transplantation includes an early, intense induction phase of immunosuppression followed by a later maintenance phase and long-term chronic immune dysfunction. In a large cohort, solid organ transplant recipients had a 6-fold overall excess risk of NHL compared to the general population (SIR=6.2; 95%CI 5.9-6.5).³⁸ The excess risk was associated with a distinct spectrum of NHL subtypes (e.g., DLBCL, Burkitt, and several T-cell lymphomas) that did not include indolent subtypes including FL (SIR=0.9; 95%CI 0.7-1.3).

Autoimmunity and Atopy

In an InterLymph pooling project,³⁹ risk of FL was associated with Sjögren Syndrome (SS; OR=3.91, 95%CI 1.39-11.0), particularly secondary SS (OR=7.55, 95%CI 1.75-32.7), but not other autoimmune disorders including ulcerative colitis, type 1 diabetes, celiac disease or systemic lupus erythematosus. SS was also strongly associated with MZL (OR=30.6) and DLBCL (OR=8.92) suggesting potential shared pathogenesis. In the SEER-Medicare NHL case-control study, FL was associated with rheumatoid arthritis (OR=1.3, 95%CI 1.1-1.5), autoimmune hemolytic anemia (OR=3.4, 95%CI 1.4-8.2) and aplastic anemia (OR=2.4, 95%CI 1.1-5.2); was only weakly associated with SS (OR=1.3, 95%CI 0.7-2.2); and was not associated with 22 other autoimmune conditions.⁴⁰

A history of atopic disorders was inversely associated with FL (OR=0.87, 95% 0.80-0.94), which was not confounded by other FL risk factors.²⁴ Inverse associations were observed for allergy (excluding drug allergy), food allergy, asthma, and hay fever (ORs 0.79-0.88), but not eczema. However, the association of allergies and FL was null in the Multiethnic cohort study (HR=0.97, 95%CI 0.64-1.47), raising some concern echoed more broadly in the NHL literature of reverse causality underlying this association.⁴¹

Immune Markers

Biomarkers of immune function, particularly circulating cytokine and chemokines, provide insight into lymphomagenesis. While most studies have not reported subtype-specific NHL results, a meta-analysis of four general population, HIV-negative cohort studies found positive associations of sCD30 (OR=3.78 for greater than reference level, 95%CI 2.50-5.72) and sCD27 (OR=1.77, 95%CI 1.04-2.09) with FL risk.⁴² These soluble forms of tumor necrosis factor receptors are markers of B cell activation, supporting a mechanistic role for chronic B-cell activation even at the subclinical level.

Hormonal and Reproductive

Unlike most NHL subtypes with a higher male sex ratio, in FL this is much smaller or even reversed, raising etiologic hypotheses related to hormonal and reproductive factors. In an InterLymph pooled analysis, FL risk was inversely associated with number of pregnancies (trend OR=0.88, 95%CI 0.81-0.96) and was positively associated with hormonal contraception use (OR=1.30, 95%CI 1.04-1.63), with higher risks for the latter when use started after age 21 years, use was <5 years, or use stopped >20 years before diagnosis.⁴³ There were no associations with menstrual history or other details of reproductive history. Postmenopausal hormone therapy (HT) was inversely associated with FL risk (OR=0.82; 95%CI 0.66-1.01) in an InterLymph pooled analysis, particularly for current use or later age at initiation.⁴⁴ Type of HT was not available, although when stratified by history of hysterectomy, the inverse association for FL was only observed in women with an intact uterus. In contrast to case-control studies, cohort studies have generally found no or increased FL risk with HT, and in the 13-year followup of the randomized portion of the Women’s Health Initiative, unopposed estrogen or estrogen plus progestin was not associated with FL risk.⁴⁵

Other medical exposures

Type 2 diabetes mellitus was not associated with FL in a meta-analysis of 4 studies.⁴⁶ While risk of B-cell NHL increases after use of chemotherapy or radiotherapy for cancer treatment, most cases are aggressive subtypes and there are no risk estimates for FL.⁴⁷ Radiation after solid organ cancer was not associated with risk of FL, with the possible exception of radiotherapy for thyroid cancer (RR=4.38, 95%CI 1.40-13.7).⁴⁸ History of blood transfusion was not associated with FL risk in a meta-analysis of case-control and cohort studies.⁴⁹ In an InterLymph pooled analysis of 13 case-control studies, blood transfusion was inversely associated with FL in both men (OR=0.70, 95%CI 0.56-0.87) and women (OR=0.77, 95%CI 0.64-0.92)⁵⁰ and was not confounded by FL-specific risk factors.²⁴ The unexpected inverse association lacks a compelling causal explanation.⁵⁰

OCCUPATIONAL AND ENVIRONMENTAL RISK FACTORS

In an InterLymph pooled analysis of ten case-control studies (including 2140 FL cases), none of the 25 a priori defined occupational groups was associated with FL risk, although there were increased risks associated with specific occupations including “spray-painter (except construction)” (OR=2.67, 95%CI 1.36-5.25) and >10 years of employment as a “medical doctor” (OR=2.23, 95%CI 1.17-4.26) and decreased risks associated with employment as “bakers/millers” (OR=0.54, 95%CI 0.30-0.99) and “university and higher education teachers” (OR=0.62, 95%CI 0.44-0.89).⁵¹ The few studies of pesticide exposure have either not reported FL specific results, or have had very small numbers. In a pooled analysis of 4 case-control studies on lymphoma risk and organophosphate and carbamate insecticide use, only malathion use was associated with FL (OR=1.58, 95%CI 1.11-2.27),⁵² but this was not seen in a large meta-analysis⁵³ or pooled analysis of cohort studies.⁵⁴ In a systematic review and metaanalysis, only a DDT and FL association was identified (OR=1.5, 95%CI 1.0-2.4),⁵³ while in a pooled analysis of cohort studies of over 300,000 agricultural workers (and 214 incident FL cases), none of the 14 pesticide chemical groups or 33 individual active ingredients were associated with FL.⁵⁴

In an InterLymph pooling project, any use of hair dye was associated with increased risk of FL among women (OR=1.3, 95%CI 1.0-1.6), particularly for women who began using hair dye before 1980 (OR=1.4; 95%CI 1.1-1.9), where there were associations with duration of use for permanent, dark and light color dyes as well as with cumulative use.⁵⁵ For women who started using hair dyes in 1980 or later, duration of use was only associated with use of permanent and dark-colored dyes. It is not clear if these patterns reflect long-term usage or a change in dye formulations in the early 1980s.⁴ Work as hairdresser has not been associated with FL risk.⁵¹

FL risk was inversely associated with recreational sun exposure (OR=0.73 for highest vs. lowest quartile, 95%CI 0.62-0.86) in an InterLymph pooled analysis,⁵⁶ and this association was not confounded by other FL risk factors.²⁴ A meta-analysis that also included cohort studies reported a slightly weaker association (RR=0.78; 95%CI 0.70-0.88), but no association with dietary vitamin D intake or serum/plasma 25-hydroxyvitamin D,⁵⁷ leaving immunomodulation by sun exposure as a more probable putative mechanism.⁴

LIFESTYLE AND BEHAVORIAL RISK FACTORS

Smoking

Cigarette smoking was not associated with FL overall in two large meta-analyses,^58,59 although in one meta-analysis⁵⁸ elevated risk of FL in former female smokers was observed but only in case-control studies (OR=1.33, 95%CI 1.07-1.66). In an InterLymph pooled analysis, cigarette smoking was associated with FL risk in females (OR=1.22, 95%CI 1.09-1.37) but not males (OR=0.98, 95%CI 0.87-1.10; P-heterogeneity=0.004), and this was limited to current smokers and was most strongly related to duration than frequency of cigarettes smoked.²⁴ A large cohort study not in the meta-analyses also found that female current smokers had an elevated risk of FL (RR=2.13, 95%CI 1.20–3.77) that was not apparent among former female smokers or males.⁶⁰

Alcohol

Alcohol drinkers had a lower risk of FL (OR=0.80, 95%CI 0.69-0.92) in a meta-analysis of 21 case-control and 8 cohort studies, which was observed for NHL overall (OR=0.85, 95%CI 0.79-0.91) and for most major subtypes except CLL/SLL.⁶¹ An updated meta-analysis of only cohort studies also found an inverse association of alcohol use with FL (OR=0.85, 95%CI 0.78-0.93) and no strong association with type of alcohol.⁶² In the InterLymph Subtypes Project, history of alcohol use was inversely associated with FL among women (OR=0.79, 95%CI 0.68-0.91) but not men (OR=0.95, 95%CI 0.80-1.12), and there was no clear pattern with type of alcohol, duration, number of drinks or cumulative alcohol consumption.²⁴

Anthropometrics and physical activity

In a meta-analysis of cohort studies, body mass index (BMI), abdominal fatness, and recreational physical activity were not associated with risk of FL, while height was positively associated with risk in women (RR=1.22 for highest vs. lowest quartile, 95%CI 1.01-1.48).⁶³ In the InterLymph Subtypes Project, FL was not associated with usual adult BMI, but was positively associated with BMI as a young adult (OR=1.21 per 5 kg/m², 95%CI 1.09-1.35).²⁴ There was also an association with height (OR=1.15 for highest vs. lowest quartile, 95%CI 1.02-1.30) restricted to men and no association with physical activity. The BMI association did not appear to be confounded by other FL risk factors, and similar associations have been observed in the two largest cohort studies published to date.^64,65

Diet

Our knowledge of dietary associations with FL has been limited due reliance on case-control studies (with attendant potential biases) and relatively few prospective cohort studies with subtype data or sufficient numbers of FL cases.⁴ The most mature results are from a meta-analysis of case-control studies and cohort studies which found vegetable (OR=0.70 for high versus low intake, 95%CI 0.53-0.92) but not fruit (OR=0.96, 95%CI 0.72-1.28) intake was associated with FL.⁶⁶

INSIGHTS FROM NHL SUBTYPE RISK FACTOR PATTERNS

The InterLymph Subtypes Project, which included a variety of risk factor data on 11 NHL subtypes, formally evaluated whether risk factors vary by NHL subtype as well as whether subtypes cluster based on risk factor profliles.⁶⁷ With respect to etiologic heterogeneity (i.e., do risk factors vary by NHL subtype), Table 1 summarizes risk factors that were associated with 1 or more NHL subtypes, and for those with P<0.01 a formal heterogeneity test was conducted. For example, family history of NHL was strongly associated with 1 or more NHL subtypes (P=1.7 × 10⁻¹³) but there was no evidence for subtype heterogeneity (P=0.52) across subtypes: 8 of 11 subtypes showed ORs>1.5 (5 at P<0.05), including FL. Based on this analysis, there was evidence for subtype heterogeneity of associations (P<0.05) for most autoimmune diseases, HCV, height, smoking, and occupation as a teacher (FL associations shown in the table). In contrast, homogenous risks across most subtypes were observed for family history of NHL, allergy/hay fever, blood transfusion, young adult BMI, alcohol use, recreational sun exposure, socioeconomic status, and selected occupations, for which FL showed associations similar to NHL overall and to many NHL subtypes.

Table 1.

Heterogeneity of risk factor associations for follicular lymphoma in the context of other major NHL subtypes, ^* InterLymph Subtypes Project

Risk Factor	P ASSET§	NHL OR (95% CI)¶	P Homogeneity¥	FL OR (95% CI)¶	Other statistically significant subtypes (direction of OR)
Family history of NHL	1.7 × 10−13	1.79 (1.51-2.13)	0.52	1.99 (1.55-2.59)	MZL(↑), DLBCL(↑), CLL/SLL(↑), MCL(↑)
B-cell activating autoimmune disease	3.8 × 10−22	1.96 (1.60-2.40)	9.8 × 10−10	1.27 (0.90-1.79)	MZL(↑), LPL/WM(↑), DLBCL(↑)
-Systemic lupus erythematosis	1.9 × 10−8	2.83 (1.82-4.41)	0.18	1.81 (0.91-3.59)	MF/SS(↑), PTCL(↑), MZL(↑), LPL/WM(↑), DLBCL(↑)
- Sjögren Syndrome	6.3 × 10−18	7.52 (3.68-15.4)	7.3 × 10−9	3.23 (1.19-8.80)	MZL(↑), LPL/WM(↑), DLBCL(↑)
T-cell activating autoimmune disease	0.0053	1.07 (0.95-1.21)	0.012	0.90 (0.72-1.11)	MF/SS (↑), PTCL(↑)
-Celiac disease	5.2 × 10−11	1.77 (1.05-2.99)	5.1 × 10−8	1.27 (0.53-3.01)	PTCL(↑), DLBCL(↑)
-Systemic sclerosis/scleroderma	0.0051	1.03 (0.41-2.58)	0.065	1.08 (0.23-5.00)	MF/SS(↑), BL(↑), HCL(↑)
Hepatitis C Virus (seropositive)	2.3 × 10−8	1.81 (1.39-2.37)	0.0021	0.57 (0.30-1.10)	MZL(↑), LPL/WM(↑), DLBCL(↑), CLL/SLL(↑)
Allergy	5.9 × 10−8	0.86 (0.81-0.92)	0.24	0.88 (0.79-0.98)	PTCL(↓), DLBCL (↓), CLL/SLL(↓), MCL(↓)
Hay fever	9.1 × 10−9	0.82 (0.77-0.88)	0.12	0.82 (0.73-0.91)	PTCL(↓), BL(↓), LPL/WM(↓), DLBCL(↓), MCL(↓)
Blood transfusion (ever)	8.8 × 10−5	0.83 (0.77-0.91)	0.050	0.78 (0.68-0.89)	DLBCL(↓), CLL/SLL(↓), HCL(↓)
Young adult Body Mass¶	4.2 × 10−9	1.95 (1.51-2.53)	0.28	2.13 (1.44-3.14)	DLBCL(↑)
Height¶	0.0017	1.20 (1.08-1.32)	0.024	1.20 (1.02-1.40)	BL(↑), DLBCL(↑), CLL/SLL(↑), HCL(↑)
Alcohol consumption (≥1 drink per month)	8.9 × 10−8	0.87 (0.81-0.93)	0.062	0.86 (0.77-0.96)	MZL(↓), BL(↓), DLBCL(↓), PTCL(↓)
Smoking (duration)¶	2.2 × 10−9	1.06 (0.99-1.14)	3.2 × 10−9	1.19 (1.06-1.33)	PTCL(↑), MZL(↑), LPL/WM(↑), CLL/SLL(↓),
Recreational sun exposure¶	2.7 × 10−9	0.74 (0.66-0.83)	0.79	0.70 (0.58-0.84)	PTCL(↓), MZL(↓), DLBCL(↓), CLL/SLL(↓)
Socioeconomic status¶	3.4 × 10−5	0.88 (0.83-0.93)	0.061	0.94 (0.85-1.03)	BL(↓), DLBCL(↓), HCL(↑)
General farm worker	0.0082	1.28 (1.10-1.50)	0.34	1.18 (0.88-1.57)	MF/SS(↑), CLL/SLL(↑),
Painter	0.0048	1.22 (0.99-1.51)	0.085	1.31 (0.93-1.84)	MF/SS(↑)
Teacher	5.6 × 10−4	0.86 (0.77-0.95)	0.0062	0.92 (0.79-1.07)	MZL(↓). BL(↓), DLBCL(↓),

^*Include the following non-Hodgkin lymphoma (NHL) subtypes: diffuse large B-cell lymphoma (DLBCL); Burkitt lymphoma (BL); peripheral T-cell lymphoma (PTCL); marginal zone lymphoma (MZL); chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL/SLL); follicular lymphoma (FL); mantle cell lymphoma (MCL); lymphoplasmacytic lymphoma/Waldenstrom Macroglobulinemia (LPL/WM); mycosis fungoides/Sezary syndrome (MF/SS); hairy cell leukemia (HCL).

^§P-ASSET is a subset-based statistical approach used to test whether an exposure was associated with 1 or more NHL subtypes.

^¥P-Homogeneity is a test of homogeneity of ORs across NHL subtypes and is derived from the random effects meta-analysis Q statistic.

^¶Odds Ratio (OR) and 95% confidence intervals (CI) derived from fixed effects logistic regression, adjusted for age, race/ethnicity, sex and study. OR represents risk per ordinal increase in the following categories: body-mass index as a young adult (<18.5, 18.5-22.4, 22.5-24.9, 25.0-29.9, ≥30 kg/m²), height (sex-specific quartiles, males: <172.0, 172.0-177.7, 177.8-181.9, ≥182.0 cm; females: <159.0, 159.0-162.9, 163.0-167.9, ≥168.0 cm), duration of cigarette smoking (0, 1-19, 20-29, 30-39, ≥40 years), recreational sun exposure (hours per week, study-specific quartiles), and socioeconomic status (low, medium, high; measured by years of education for studies in North America or by dividing measures of education or socioeconomic status into tertiles for studies in Europe or Australia).

To identify whether certain subtypes shared risk factor profiles, hierarchical clustering was used based on an analysis of all risk factor data simultaneously. Key findings were that B-versus T-cell lymphomas first divided into two clusters, and then for B-cell lymphomas, MZL and BL formed a cluster, followed by a FL and MCL cluster, a LPL/WM and DLBCL cluster, and finally a CLL/SLL cluster. Overall, no FL-specific risk factors were identified, and with the exception of autoimmune diseases and HCV, FL largely had a similar risk factor profile that paralleled NHL overall, and most closely matched the risk factor profile for MCL.

OPPORTUNITIES FOR PREVENTION

The descriptive epidemiology of FL, which shows strong international variation and the highest incidence in Western countries, suggests the potential for prevention. The most robust risk factors, including family history, selected autoimmune diseases, and allergy/atopy provide biologic insights but no current direct routes to primary prevention. Definitive genetic loci have been identified, but they are limited and not yet sufficiently robust for screening in families or the population. While there are suggestive leads for modifiable lifestyle and behavioral risk factors that lend themselves to public health interventions (e.g., reducing hair dye use, young adult BMI, and smoking), others FL risk factors present more complicated prevention approaches (e.g., protective effects of recreational sun exposure and moderate alcohol use), and none of these risk factors are established. New leads related to certain infections, precursor conditions and immune biomarkers with FL risk may provide future routes to prevention.

CONCLUSIONS AND FUTURE DIRECTIONS

Progress on identifying FL-specific risk factors has accelerated with the implementation of the InterLymph nested classification as well as with the availability of large pooled analyses. However, beyond family history of NHL, genetic loci, and Sjögren Syndrome, there are no established risk factors for FL although there are promising leads related to atopy, certain infections, hair dye, recreational sun exposure, smoking, alcohol use and anthropometrics, but these require further research including better exposure assessment and stronger study designs. We remain early in our understanding of genetic susceptibility to FL, and there are likely to be additional loci identified by GWAS and next generation sequencing. Deeper understanding of the striking role that HLA plays in FL risk through the antigen processing and presentation is needed. Progress on all of these fronts will also allow more robust evaluation of gene-environment interactions. Advances in the understanding of FL biology, both through study of precursors and the genomic and epigenomic architecture of the FL tumor genome, should also help identify novel links to specific risk factors. Finally, most studies have been conducted in Western populations of European ancestry, and future studies need to address racial/ethnic differences and populations of contrasting risk.

KEY POINTS.

• An estimated 14,000 cases of FL were diagnosed in the US in 2016; FL incidence is slightly higher in men, rises steeply with age, and is highest in non-Hispanic whites.

• FL incidence shows geographic variation, has been stable in the US and France since 2000 but has been rising in other Western and Asian countries.

• Five-year relative survival rates in the US range from 80-90% across sex and major racial/ethnic groups; relative survival rates have been increasing since 2000 in US and other Western and Asian countries.

• Beyond family history of NHL, genetic loci, and Sjögren syndrome, there are no established risk factors for FL although there are promising leads related to atopy, certain infections, anthropometrics, hair dye, recreational sun exposure, smoking, and alcohol use.

SYNOPSIS.

Follicular lymphoma (FL) is a common indolent lymphoma subtype in Western countries. FL incidence increases with age, and shows considerable variation by race/ethnicity and geography. In the US and France, FL incidence has been stable since 2000, while in other Western and Asian countries it has been increasing. Five-year relative survival rates survival rates have been increasing in Western and Asian countries. Progress on identifying FL-specific risk factors has accelerated with the implementation of the InterLymph nested classification and the availability of larger epidemiologic studies and pooled analyses of these studies. However, beyond family history of NHL, genetic loci, and Sjögren syndrome, there are no established risk factors for FL although there are promising leads related to atopy, certain infections, anthropometrics, hair dye, recreational sun exposure, smoking, and alcohol use, but these require further research including better exposure assessment and more robust studies.