Both chronic HBV infection and naturally acquired HBV immunity confer increased risks of B-cell non-Hodgkin lymphoma

We performed a retrospective case-control study on the association between HBV infection and NHL. Newly diagnosed NHL patients who had received a histological diagnosis at Union Hospital (affiliated to Tongji Medical School, Huazhong University of Science and Technology) between 2010 and 2017 were included. Hospital records were reviewed for patient age, sex, lymphoma diagnosis, year of diagnosis, and laboratory results at the first testing (including five HBV serum markers, human immunodeficiency virus (HIV), hepatitis C virus (HCV) and treponema pallidum (TP) antibody). The original study included 4044 cases between the ages of 1 to 97 years old. A sufficient amount of serum or plasma and the precise pathologically verified lymphoma diagnoses were available for 3502 cases. Amongst these cases, we had already excluded patients who tested positive for the HIV, HCV or TP antibody to eliminate possible interaction among those pathogens and impact of those pathogenic microorganisms on the development of NHL. The data analysis was performed for NHL (n = 3502), B-cell NHL (B-NHL, n = 2535), and T-cell NHL (T-NHL, n = 967); the more frequent B-cell-specific entities, including DLBCL (n = 1224), Burkitt’s lymphoma (BL, n = 90), follicular lymphoma (FL, n = 253), small lymphocytic lymphoma/chronic lymphocytic leukaemia (SLL/CLL, n = 341), mantle cell lymphoma (MCL, n = 105), marginal zone lymphoma (MALT, n = 314), lymphoplasmacytic lymphoma (LPL, n = 30), precursor B lymphoblastic lymphoma (PBLL, n = 37), and other B-cell lymphomas (n = 141); and the more frequent T-cell-specific entities, including peripheral T-cell lymphoma-unclassified (PTCL-unclassified, n = 161), NK/T-cell lymphoma (NK/T, n = 436), angioimmunoblastic T-cell lymphoma (AITL, n = 122), anaplastic large cell lymphoma (ALCL, n = 73), precursor T lymphoblastic lymphoma (PTLL, n = 119), and other T-cell lymphomas (n = 56).

For each case, two controls were selected, which were balanced for age (< 30, 30–39, 40–49, 50–59, 60–69, ≥70 years), sex and year of diagnosis. The controls were admitted for a wide spectrum of acute conditions, which included fractures and traffic accident injuries, facial plastic and eye surgery. Further exclusions for the controls were as follows: (1) diseases associated with HBV infection; (2) a recorded history of malignant disease; (3) diabetes or autoimmune disease. In total, 7004 controls were screened.

The blood test results were retrospectively collected from the medical records. An enzyme-linked immunosorbent assay was applied to test these serum samples from the cases and controls for five markers of HBV, including HBsAg, hepatitis B surface antibody (anti-HBs), hepatitis B e antigen (HBeAg), hepatitis B e antibody (anti-HBe), and hepatitis B core antibody (anti-HBc). The same samples were additionally tested for the HIV, HCV, and TP antibody. These assays were conducted in the Central Laboratory of Union Hospital.

According to the Hans algorithm,^[3]we used immunohistochemistry (IHC) jalgorithms based on the expression of markers, including CD10, BCL6, and MUM-1, to divide the DLBCL patients into two categories as follows: GCB type (n = 356) and non-GCB type (n = 663).

We further subdivided the B-NHL patients into common HBsAg-positive groups and HBsAg negative groups including: HBsAg positive/HBeAg positive/anti-HBc positive patients (indicating acute or HBeAg positive chronic HBV infection), the HBsAg positive/anti-HBe positive/anti-HBc positive patients (indicating inactive HBsAg carrier state or HBeAg negative chronic HBV infection), the HBsAg negative/anti-HBs positive/anti-HBc positive group (implying immunity due t;o natural infection), the HBsAg negative/anti-HBs positive/anti-HBc negative group (implying immunity due to hepatitis B vaccination), the HBsAg negative/anti-HBs negative/anti-HBc positive group (several interpretations are possible, including the lack of HBV immunity, resolved infection, occult HBV infection or without a confirmable history of HBV infection) and the HBsAg negative/anti-HBs negative/anti-HBc negative group (meaning the patient was susceptible to HBV infection) [11].

We further conducted an updated meta-analysis, and the detailed process was shown in Additional file 1: Supplementary Method. Eligible case-control studies and cohort studies were included, the combined effect was reflected by odds ratio (OR), the calculations and graphs of the meta-analysis were performed in R 3.4.0.

There were 3502 patients diagnosed with NHL and 7004 control patients eventually incorporated. The two groups were well matched in terms of the age distribution, sex ratio and year of diagnosis, (P = 1.0, P = 1.0 and P = 1.0, respectively), and the case-control ratio was 1:2 (Additional file 2: Table S1).

A total of 523 patients (14.9%) with NHL were HBsAg positive, which was significantly higher than the number of positive controls (8.8%). Specifically, the proportion of HBsAg-positive cases was 17.0% amongst B-NHL patients and 9.4% amongst T-NHL patients. A subgroup analysis of the association between NHL and the presence of HBsAg was performed for the main histological subtypes (Table 1). HBsAg positivity was associated with a significantly increased risk of B-NHL in both the univariate (OR (95% CI): 2.14 (1.87–2.44)) and multivariate analyses (AOR (95% CI): 2.14 (1. 88–2.45)) and for DLBCL in particular (OR = 2.42, 95% CI: 2.05–2.86; AOR = 2.45, 95% CI: 2.07–2.89). Notably, FL and SLL/CLL showed a significantly high carrier rate in both the univariate and multivariate logistic regression analyses. Except that HBsAg positivity was associated with a significantly increased risk of AITL in both the univariate (OR (95% CI): 1.80 (1.09–2.99)) and multivariate analyses (AOR (95% CI): 1.84 (1.10–3.06)), there was no significant association between HBsAg positive and other T-NHL subtypes including NK/T, PTCL-unclassified, PTLL, ALCL.

Compared to the HBsAg−/anti-HBs−/anti-HBc- subgroup, all the three HBsAg positive subgroups showed a significant positive association with B-NHL in the univariate and multivariate logistic regressions (AOR (95% CI): 6.23 (3.95–9.82), 4.37 (3.55–5.37), 1.51 (1.22–1.86), respectively), and the HBsAg negative/anti-HBs positive/anti-HBc positive subgroup was also statistically associated with an increased risk of B-NHL (AOR (95% CI): 2.25 (1.96–2.57)). Compared with the HBsAg negative/anti-HBs positive/anti-HBc positive subgroup, the other three HBsAg negative subgroups were negatively associated with B-NHL (AOR (95% CI): 0.45 (0.39–0.51), 0.46 (0.40–0.54), 0.84 (0.71–0.99), respectively), as shown in Table 2.

There were 9 cohort studies [12‐20] and 25 case-control studies [8, 9, 21‐43] plus our study selected (Additional file 3: Figure S1). A total of 494,112 NHL cases were enrolled in the meta-analysis. The main characteristics of the included studies are listed in Additional files (Additional file 4: Table S2 and Additional file 5: Table S3). Using a random effects model, an increased OR of NHL in the patients with HBV was observed (OR = 2.09; 95% CI 1.76–2.50, P ≤ 0.01), as shown in Fig. 1. We further analysed the association between the NHL subtypes and HBsAg in different HBV prevalence populations (Table 3). In high HBV prevalence countries, an increased OR of DLBCL with HBsAg positive patients was observed (OR = 2.69; 95% CI 1.88–3.86, P < 0.01); the OR of FL in the HBV patients was 2.42 (95% CI 1.36–4.32, P < 0.01), and the OR of ALCL in the patients with HBV was 3.23 (95% CI 1.27–8.18, P = 0.04). In sensitivity analysis, when excluding the one study that did not test the HBV status with seropositivity [21] or the six studies [24, 25, 31, 38, 39, 43] with NOS scores [44] less than seven, the overall results were similar.