Clinical Prediction of Failure of Lamivudine Prophylaxis for Hepatitis B Virus-Infected Patients Undergoing Cytotoxic Chemotherapy for Malignancy

Abstract

Although lamivudine (LAM) prophylaxis is recommended for patients infected with hepatitis B virus (HBV) undergoing chemotherapy for malignant disease, HBV reactivation sometimes occurs during or after LAM administration. The aim of this study was to determine predictors of LAM prophylactic failure in patients with malignancies. Patients with malignancies were routinely screened for serum hepatitis B surface antigen (HBsAg) from June 2002 to August 2008. All consecutive, HBsAg-positive patients received LAM prophylaxis during and after completion of chemotherapy. We assessed risk factors for virologic breakthrough and withdrawal hepatitis. Death without HBV reactivation was regarded as a competing risk event, which was adjusted by Fine and Gray's model. A total of 110 patients were included in this study. They received LAM prophylaxis for a median of 9.2 months. Virologic breakthrough occurred in 15 patients at a median of 10.9 months from the initiation of LAM prophylaxis. Withdrawal hepatitis occurred in 15 patients at a median of 2.4 months after cessation of LAM prophylaxis. Multivariable analysis showed that high baseline HBV DNA titer (≥2,000 IU/ml) (hazard ratio [HR], 9.94; P = 0.0063) and the use of rituximab (HR, 3.19; P = 0.027) were significant predictors of virologic breakthrough and that high baseline HBV DNA titer (HR, 5.90; P = 0.007), liver cirrhosis (HR, 10.4; P = 0.002), and distant metastasis (HR, 5.14; P = 0.008) were independent risk factors for withdrawal hepatitis. Patients with high viremia, liver cirrhosis, rituximab treatment, and distant metastasis are at high risk of prophylactic failure and need antiviral agents with a greater barrier to resistance.

INTRODUCTION

Patients with hepatitis B virus (HBV) infection who undergo chemotherapy for a malignancy are at risk of an interruption of chemotherapy as well as liver-related morbidity and mortality due to HBV reactivation (1, 29). The incidence of HBV reactivation in hepatitis B surface antigen (HBsAg)-positive carriers receiving cytotoxic chemotherapy has been estimated to be 48 to 52.7% (18). In particular, well-established risk factors for HBV reactivation are young age, male gender, lymphoma, and the use of anthracycline, rituximab, and steroids as part of anticancer therapy (5, 27, 31).

Lamivudine (LAM), a nucleoside analogue, shows antiviral efficacy in the treatment of chronic hepatitis B (CHB) (4, 13) and, as reported recently, in the prevention of chemotherapy-induced reactivation of HBV (9, 12, 17, 20, 27). Several prospective studies demonstrated that the incidence of HBV reactivation among patients who received LAM prophylaxis is less than 20%, compared with 20 to 78% in historical, untreated controls (9, 16, 17, 20, 27). Therefore, LAM is routinely recommended with initiation of cytotoxic or immunosuppressive therapy in HBsAg-positive patients (19).

Although antiviral prophylaxis effectively prevents HBV reactivation, prophylactic failure occasionally results from virologic breakthrough or withdrawal flare. In spite of the clear utility of LAM for prophylaxis in HBsAg-positive patients, recent studies have brought to light the emergence of LAM-resistant strains of HBV as a result of extended LAM therapy (9, 11, 17). However, to date, there have been insufficient data on the emergence rate of the tyrosine-methionine-aspartate-aspartate (YMDD) motif mutation and on the clinical impact of these mutants in immunosuppressed subjects undergoing chemotherapy.

With respect to the problems associated with short-term (withdrawal hepatitis) and long-term LAM therapy (the emergence of LAM-resistant mutants), the selection of appropriate antiviral agents and the optimal duration of therapy may reduce the potential for additional complications or prophylactic failure in high-risk patients. Therefore, the aims of the present study were to assess the relative risk of antiviral prophylactic failure and thus to determine the optimal strategy for antiviral prophylaxis in HBsAg-positive patients with oncologic and hematologic malignancies undergoing chemotherapy. (This article was presented as a poster at the 44th Annual Meeting of the European Association for the Study of the Liver [EASL] in Copenhagen, Denmark, 22 to 26 April 2009, and the 51st Annual Meeting of the American Society of Hematology [ASH] in New Orleans, LA, 5 to 8 December 2009.)

MATERIALS AND METHODS

Patients.

HBsAg-positive patients (≥18 years of age) with oncologic and hematologic malignancies who received prophylactic LAM (Zeffix; Glaxo Wellcome, Greenford, United Kingdom) therapy were retrospectively reviewed between June 2002 and August 2008 at Seoul National University Hospital. The following patients were excluded from this study: (i) those who had previous exposure to antiviral therapy, including LAM for therapeutic purposes against HBV infection; (ii) those who were started on antiviral agents other than LAM as antiviral prophylaxis; (iii) those with other causes of chronic liver disease besides HBV (i.e., seropositive for anti-hepatitis C virus antibody or with excessive alcohol consumption [>20 g/day]); (iv) those who had decompensated liver states, such as jaundice, ascites, variceal bleeding, or hepatic encephalopathy; and (v) those who received LAM as deferred treatment of hepatitis flare after initiation of chemotherapy. The study protocol was reviewed and approved by the Institutional Review Board of Seoul National University Hospital.

Study measurements.

The following data were recorded for each patient: age and gender; serologic status for HBV as indicated by the presence of HBsAg, hepatitis B e antigen (HBeAg), and anti-hepatitis B e antibody (anti-HBe) concentrations of HBV DNA in serum at baseline; serum aspartate aminotransferase (AST) and alanine aminotransferase (ALT) levels; underlying liver status (the presence of liver cirrhosis) at baseline; the specific type of malignancy (oncologic versus hematologic); primary liver cancer (hepatocellular carcinoma and cholangiocarcinoma); the presence of distant metastasis; and the type of chemotherapy (the use of anthracycline agents, rituximab, and/or steroids as a chemotherapeutic or antiemetic). All eligible patients were followed regularly (every 3 months) to monitor the HBV status, including serum HBeAg, anti-HBe, and HBV DNA concentrations and biochemical parameters, such as serum bilirubin, AST, and ALT levels. The presence of liver cirrhosis was diagnosed if there were either radiological findings of cirrhosis or clinical signs of portal hypertension, such as thrombocytopenia and splenomegaly.

HBV DNA determinations and genotyping.

Serum HBeAg and anti-HBe (radioimmunoassay [RIA] enzyme-linked immunosorbent assay [ELISA] rapid kit; Shin Jin Medics, Inc., Seoul, South Korea) and HBV DNA titers (lower limit of quantification, <20 IU/ml) (Cobas Amplicor HBV monitor test; Roche Molecular Systems, Pleasanton, CA) were measured every 3 months during the study period. Quantification of HBV DNA and mutational analysis of virologic breakthrough in patients were conducted in parallel while the patients were undergoing chemotherapy and prophylactic LAM. Genotypic analysis for LAM resistance was performed as previously described (2). Genotypic analysis for resistance to adefovir (ADV) (Hepsera; Gilead Sciences, Inc., Foster City, CA) and entecavir (ETV) (Baraclude; Bristol-Myers Squibb, Princeton, NJ) for patients who developed virologic breakthrough during antiviral treatments was performed using restriction fragment mass polymorphism analysis as previously described (3). No mutation associated with drug resistance was able to be detected if serum HBV DNA titers were lower than 357 IU/ml.

Definition of hepatic events.

Based on the American Association for the Study of Liver Diseases practice guidelines (19), virologic breakthrough during antiviral treatments was defined as an elevation in serum HBV DNA titer of 10-fold or more above the nadir regardless of serum ALT levels. Withdrawal hepatitis was defined as the elevation of serum ALT levels above the upper limit of normal (40 IU/ml) accompanied by a serum HBV DNA level of ≥20,000 IU/ml in HBeAg-positive patients or the elevation of serum ALT levels accompanied by a serum HBV DNA level of ≥2,000 IU/ml in HBeAg-negative patients (10). Rescue therapy for patients with virologic breakthrough included nucleoside or nucleotide analogues with potent antiviral activity against HBV variants, such as daily therapy with 10 mg of ADV or 1 mg of ETV. Time to virologic breakthrough was measured from initiation of LAM, while time to withdrawal hepatitis was measured from discontinuation of LAM. The primary outcome measures were virologic breakthrough during LAM treatment and withdrawal hepatitis after discontinuation of LAM, with death due to progression of malignant disease without occurrence of hepatic events as a competing risk event.

Statistical analysis.

Univariable analysis was performed to determine the significant risk factors for virologic breakthrough and withdrawal hepatitis by the Gray's test or Fine and Gray's model. Factors with P ≤ 0.20 from univariable analysis were included in a consecutive multivariable analysis. Fine and Gray's regression model was used in multivariable analysis of competing risks by using a stepwise forward method (6). For a withdrawal hepatitis event, the Cox proportional hazards model was applied because there was no competing risk event preceding the withdrawal hepatitis event (23). The proportional hazard assumptions in both the Fine and Gray's model and the Cox model were tested by the graphical exploration of the Schoenfeld's residuals (7, 25). The 95% confidence interval (CI) for the hazard ratio (HR) was also calculated, and statistical significance was defined as P < 0.05. For those analyses, R 2.8.1 software was used (R Foundation, Vienna, Austria). The association between total duration of LAM therapy or duration of additional LAM therapy after completion of chemotherapy and the occurrence of withdrawal hepatitis was evaluated by Student's t test. This was evaluated by statistical software package SPSS 17.0 for Windows (SPSS, Inc., Chicago, IL). To determine the optimal duration of additional LAM therapy, we analyzed data using the conditional inference tree structure model (8).

RESULTS

Baseline characteristics of study subjects.

A total of 128 HBsAg-positive patients with oncologic and hematologic malignancies were reviewed consecutively for this retrospective cohort study. Among these, 18 patients were excluded due to the following reasons: 9 patients had already received LAM for therapeutic purposes more than 6 months prior to study entry; 7 were started on other antiviral agents, such as ADV, ETV, or clevudine (Levovir; Bukwang Pharmaceutical, Seoul, South Korea) instead of LAM; and 1 patient acquired HBsAg during chemotherapy and another died of disease progression (anaplastic thyroid cancer) just 1 week after initiation of LAM.

The baseline characteristics of the 110 eligible subjects are shown in Table 1. The median age of the subjects was 54 years (range, 26 to 80 years), and the median serum ALT level at baseline was 25 IU/liter (range, 6 to 76 IU/liter) prior to starting chemotherapy. Twenty-nine patients (26.4%) had hematologic malignancies, and the other 81 (73.6%) had oncologic malignancies. The type of underlying malignancies and the chemotherapy regimens used are detailed in Table 2. The median follow-up period was 22 months (range, 1.2 to 83.6 months), and the median duration of prophylactic LAM therapy was 9.2 months (range, 0.8 to 46.4 months) (see Table S1 in the supplemental material).

Table 1.

Baseline characteristics of the 110 patients who started chemotherapy and lamivudine and the 61 patients who had finished chemotherapy and prophylactic lamivudine therapy^a

Characteristic	Result for evaluation of:
	VB (n = 110)	WH (n = 61)b
Male/female ratio	61:49	28:33
Age, yr (%)
≥50	72 (65.5)	41 (67.2)
<50	38 (34.5)	20 (32.8)
HBeAg, no. (%)
Positive	15 (13.6)	4 (6.6)
Negative	84 (76.4)	51 (83.6)
NAc	11 (10.0)	6 (9.8)
HBV DNA titer, IU/ml (%)
≥2,000	40 (36.4)	17 (27.9)
<2,000	61 (55.5)	37 (60.7)
NA	9 (8.1)	7 (11.4)
Median ALT, IU/ml (range)	25 (6–76)	26 (4–76)
Liver cirrhosis, no. (%)
Yes	19 (17.3)	8 (13.1)
No	91 (82.7)	53 (86.9)
Type of malignancy, no. (%)
Oncologic	81 (73.6)	48 (78.7)
Hematologic	29 (26.4)	13 (21.3)
Type of chemotherapy, no. (%)
Steroid-containing regimen	83 (75.5)	43 (70.5)
Anthracycline-containing regimen	47 (42.7)	28 (45.9)
Rituximab-containing regimen	15 (13.6)	7 (11.5)
Primary liver cancer, no. (%)d
Yes	8 (7.3)	0 (0)
No	102 (92.7)	61 (100)
Distant metastasis, no. (%)
Yes	49 (44.5)	17 (27.9)
No	52 (47.3)	40 (65.6)
Unclassifiede	9 (8.2)	4 (6.5)

^aALT, alanine aminotransferase; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; VB, virologic breakthrough; WH, withdrawal hepatitis.

^bPatients who had been followed for more than 3 months after discontinuation of prophylactic lamivudine therapy.

^cNA, laboratory data not available at baseline.

^dPrimary liver cancer includes hepatocellular carcinoma and cholangiocarcinoma but not metastatic liver cancer.

^eEight patients with acute or chronic leukemia and one with multiple myeloma were not able to be classified into either side with regard to distant metastasis.

Table 2.

Types of underlying malignancies and chemotherapy regimens used

Type of malignancy	Diagnosis	n	Chemotherapy regimen
Oncologic	Breast cancer	24	Anthracycline, cyclophosphamide, docetaxel
	Colorectal cancer	17	Fluorouracil, oxaliplatin, irinotecan
	Stomach cancer	16	Fluorouracil, cisplatin, oxaliplatin, docetaxel, etoposide, irinotecan, paclitaxel
	Non-small cell lung cancer	11	Genexol, cisplatin, carboplatin, gemcitabine, vinorelbine, paclitaxel, fluorouracil
	Hepatocellular carcinoma	5	Fluorouracil, mitomycin, cisplatin, sorafenib
	Head and neck cancer	3	Docetaxel, fluorouracil, cisplatin, tegafur
	Cholangiocarcinoma	3	Gemcitabine, fluorouracil, irinotecan, cisplatin
	Brain tumor	1	Bleomycin, etoposide, cisplatin, temozolomide
	Osteosarcoma	1	Methotrexate, adriamycin, cisplatin
Hematologic	Non-Hodgkin lymphoma	20	Rituximab, cyclophosphamide, vincristine, prednisolone, cytarabine, vinorelbine, carboplatin, bleomycin, procarbazine, adriamycin, mitoxantrone, melphalan, methotrexate, busulfan etoposide, daunorubicin, cisplatin
	Leukemia	8	Cytarabine, idarubicin, melphalan, methotrexate, cyclophosphamide, mitoxantron, fludarabine, daunorubicin, vincristine, prednisolone, busulfan
	Multiple myeloma	1	Vincristine, adriamycin, dexamethasone, thalidomide

Virologic breakthrough during lamivudine prophylaxis.

Of the 110 patients, 15 developed virologic breakthrough during prophylactic LAM therapy. The overall actuarial rates for virologic breakthrough during LAM treatment at 6, 12, and 18 months were 3.1%, 16.7%, and 34.9%, respectively (Fig. 1A). The median duration from initiation of LAM to virologic breakthrough was 10.9 months (range, 0.4 to 23.5 months) (see Table S1 in the supplemental material).

Fig 1.

Cumulative incidence function for hepatic events. The solid line shows the cumulative incidence rate for virologic breakthrough (VB) (A) and withdrawal hepatitis (WH) (B) using the Kaplan-Meier method. The dotted line shows the cumulative incidence rate for the competing risk event (i.e., death occurring prior to hepatic events).

Table 3 summarizes the main baseline characteristics and clinical outcomes of the 15 patients with virologic breakthrough. Among the 15 patients with virologic breakthrough, LAM resistance-associated mutations were documented in 7 patients at a median of 16.7 months (range, 9.6 to 23.5 months) after initiation of LAM prophylaxis: four patients substituted ETV for LAM, and 3 patients added ADV to LAM. One patient who developed the M204I mutation in reverse transcriptase died of progressive hepatic failure despite ADV add-on therapy. ETV rescue therapy selected for an ETV resistance mutation (S202G) in 1 patient along with emerging LAM resistance-associated mutations (L180M and M204V). Five of 7 patients who developed LAM resistance mutations showed biochemical flare, while only 2 of 8 patients who did not develop LAM resistance mutations had biochemical flare during antiviral prophylaxis.

Table 3.

Main baseline characteristics and clinical outcomes of the 15 patients who developed virologic breakthrough^a

Patient no.	Age (yr)	Gender	HBeAg/anti-HBe	HBV DNA at baseline (IU/ml)	Liver cirrhosis	ALT (IU/liter)		Underlying malignancy	HBV DNA at VB (IU/ml)	Mutational pattern of genotypic resistance	Duration of LAM (days)	Time to VB after last chemotherapy (days)	Rescue therapy	Clinical course
						Baseline	Peak
1	57	M	+/+	6.81 × 105	Yes	22	48	Stomach cancer	2.60 × 106	M204I	513	113	LAM + ADV	Improvement
2	43	M	−/+	1.39 × 102	No	72	924	NHL	1.64 × 104	L180M	312	32	Maintenance of LAM	Death due to disease progression
3	68	M	+/−	2.36 × 107	No	38	537	Colorectal cancer	5.61 × 106	L180M M204V	704	35	Switch to ETV	Improvement
4	63	M	+/−	3.44 × 104	No	45	162	Colorectal cancer	1.75 × 107	L180M M204V	536	222	Withdrawal of LAM	Improvement
5	40	M	−/+	7.02 × 103	No	39	257	Leukemia	1.89 × 105	V173L M204I	502	343	Switch to ETV	Improvement
6	60	M	−/+	1.55 × 107	No	28	92	NHL	>1.1 × 108	M204I	288	182	LAM + ADV	Death due to hepatic failure
7	52	F	+/−	>1.1 × 108	No	67	40	Stomach cancer	7.69 × 106	(i) L180M M204V (ii) L180M M204V S202G	435	258	(i) Switch to ETV, (ii) LAM + ADV	Improvement
8	45	M	−/+	3.75 × 104	Yes	52	30	HCC	2.83 × 103	NIb	350	78	Maintenance of LAM	Death due to disease progression
9	55	M	+/−	1.09 × 108	No	29	38	NHL	>1.1 × 108	NI	462	343	Maintenance of LAM	Improvement
10	68	M	−/+	8.78 × 103	Yes	25	107	Stomach cancer	2.18 × 104	NI	12	21	Switch to ADV	Improvement
11	66	F	−/+	2.10 × 10	No	7	26	Multiple myeloma	1.82 × 103	NI	301	7	Maintenance of LAM	Improvement
12	64	M	−/+	1.20 × 106	Yes	26	17	Stomach cancer	8.90 × 103	NI	161	39	Maintenance of LAM	Death due to disease progression
13	35	M	+/−	>1.1 × 108	No	54	179	NHL	>1.1 × 108	NI	1,156	16	Maintenance of LAM	Improvement
14	49	F	+/−	>1.1 × 108	No	12	17	NHL	7.12 × 106	NI	328	186	Withdrawal of LAM	Improvement
15	27	F	+/−	>1.1 × 108	No	11	29	NHL	3.75 × 106	NI	75	27	Switch to ETV	Improvement

^aF, female; M, male; ADV, adefovir; anti-HBe, hepatitis B e antibody; ETV, entecavir; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; LAM, lamivudine; NHL, non-Hodgkin's lymphoma; VB, virologic breakthrough.

^bNI, not identifiable. The sensitivity of the assay for HBV mutations was 357 IU/ml, and no mutation associated with drug resistance was found using DNA genotyping.

The significant predictors of virologic breakthrough included male gender, positive HBeAg, initial HBV DNA titer of ≥2,000 IU/ml, underlying hematologic malignancy, and the use of rituximab in univariable analysis. Multivariable analysis adjusted by Fine and Gray's model showed that initial serum HBV DNA titer (HR, 9.94; 95% CI, 1.91 to 51.6; P = 0.0063) and the use of rituximab (HR, 3.19; 95% CI, 1.14 to 8.96; P = 0.027) were the independent predictors of virologic breakthrough (Fig. 2A and Table 4). The additional data showed that the only significant predictor of virologic breakthrough with confirmed genotypic resistance was an initial HBV DNA titer of ≥2,000 IU/ml by multivariable analysis (see Table S2 in the supplemental material).

Fig 2.

Cumulative probabilities for hepatic events. Shown are virologic breakthrough (VB) (A) and withdrawal hepatitis (WH) (B) according to serum HBV DNA titers (Gray's test).

Table 4.

Risk factors for virologic breakthrough by univariable and multivariable analyses^a

Risk factor for VB	Univariable analysis		Multivariable analysis
	Unadjusted HR (95% CI)	P value	Adjusted HR (95% CI)	P value
Gender
Female	1.00
Male	3.04 (0.93–9.96)	0.066
Age, yr
<50	1.00
≥50	0.93 (0.33–2.64)	0.890
HBeAg
Negative	1.00
Positive	3.06 (1.10–8.47)	0.031
HBV DNA titer, IU/ml
<2,000	1.00		1.00
≥2,000	9.62 (2.01–45.9)	0.005	9.94 (1.91–51.6)	0.0063
Liver cirrhosis
No	1.00
Yes	1.14 (0.37–3.53)	0.830
Type of malignancy
Oncologic	1.00
Hematologic	2.39 (0.85–6.70)	0.098
Use of steroid
No	1.00
Yes	1.00 (0.30–3.37)	1.000
Use of anthracycline
No	1.00
Yes	1.91 (0.70–5.23)	0.210
Use of rituximab
No	1.00		1.00
Yes	3.28 (1.01–10.7)	0.048	3.19 (1.14–8.96)	0.027
Primary liver cancerb
No	1.00
Yes	0.74 (0.10–5.25)	0.760
Distant metastasis
No	1.00
Yes	0.82 (0.26–2.59)	0.730

^aCI, confidence interval; HR, hazard ratio; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; VB; virologic breakthrough.

^bPrimary liver cancer includes either hepatocellular carcinoma or cholangiocarcinoma but not metastatic liver cancer.

Withdrawal hepatitis after discontinuation of lamivudine.

Fifteen patients developed HBV withdrawal hepatitis among the 61 patients who had been followed for more than 3 months after discontinuation of prophylactic LAM therapy. The overall actuarial rates for withdrawal hepatitis after discontinuation of LAM at 1, 3, and 6 months were 0%, 16.5%, and 25.4%, respectively (Fig. 1B). The median time to withdrawal hepatitis after discontinuation of LAM therapy was 2.4 months (range, 1.6 to 5.9 months) (see Table S1 in the supplemental material).

Table 5 summarizes the main baseline characteristics and clinical outcomes of the 15 patients with withdrawal hepatitis. Of the 15 patients who experienced withdrawal flare, 5 resumed LAM, 6 were started on ETV as rescue therapy, 1 patient added ADV to LAM, and 3 patients were carefully followed without receiving any antiviral agent. One died of progressive hepatic failure despite ADV add-on therapy. LAM resistance-associated mutations (L180M and M204I) were detected in this patient's serum. The remaining 14 patients tested negative for genotypic resistance.

Table 5.

Main baseline characteristics and clinical outcomes of the 15 patients who developed withdrawal hepatitis^a

Patient no.	Age (yr)	Gender	HBeAg/anti-HBe	HBV DNA at baseline (IU/ml)	Pre-exacerbation HBV DNA nadir (IU/ml)	Liver cirrhosis	Underlying malignancy	HBV DNA at WH (IU/ml)	ALT level at WH (IU/liter)	Duration of LAM after completion of chemotherapy (days)	Time to WH after end of LAM (days)	Rescue therapy	Clinical course
1	58	F	−/+	9.06 × 107	Undetectable	No	Breast cancer	>1.1 × 108	384	103	93	LAM	Improvement
2	57	F	−/+	1.08 × 105	Undetectable	Yes	Breast cancer	2.05 × 105	219	92	74	ETV	Improvement
3	49	F	−/+	Undetectable	Undetectable	No	Breast cancer	2.49 × 107	245	113	174	LAM	Improvement
4	50	M	−/+	3.44 × 104	1,650	No	Colorectal cancer	4.41 × 107	795	166	104	Observation	Improvement
5	39	F	−/+	9.42 × 102	Undetectable	No	Breast cancer	1.58 × 107	107	40	61	LAM	Improvement
6	49	M	−/+	Undetectable	Undetectable	No	NHL	1.41 × 106	1,771	0	177	LAM	Improvement
7	62	M	−/−	2.98 × 105	38	Yes	Colorectal cancer	4.25 × 106	95	54	48	ETV	Improvement
8	60	M	−/+	8.35 × 102	NAb	No	NHL	7.11 × 107	2,762	30	86	ETV	Improvement
9	56	F	−/+	1.20 × 10	Undetectable	No	NHL	3.21 × 106	1,750	72	172	ETV	Improvement
10	59	M	−/−	2.11 × 106	NA	No	NHL	>1.1 × 108	507	96	72	ETV	Improvement
11	59	M	−/+	1.21 × 105	NA	No	Non-small cell lung cancer	>1.1 × 108	323	0	73	LAM	Death due to disease progression
12	60	M	NA	1.23 × 105	240	Yes	Leukemia	2.41 × 105	111	67	152	ETV	Improvement
13	52	M	+/−	1.36 × 106	44	Yes	Stomach cancer	2.67 × 105	118	111	69	Observation	Loss to follow up
14	47	F	−/+	1.74 × 107	51	Yes	Breast cancer	3.00 × 105	48	154	62	Observation	Improvement
15	53	F	NA	NA	NA	Yes	NHL	2.17 × 107	56	0	59	LAM + ADV	Death due to hepatic failure

^aF, female; M, male; ADV, adefovir; anti-HBe, hepatitis B e antibody; ETV, entecavir; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; LAM, lamivudine; NHL, non-Hodgkin's lymphoma; WH, withdrawal hepatitis.

^bNA, laboratory data were not available at either baseline or nadir.

In univariable analysis, the significant predictors of withdrawal hepatitis were an initial serum HBV DNA titer of ≥2,000 IU/ml, liver cirrhosis, hematologic malignancy, the use of rituximab, and distant metastasis. The Cox regression model showed that initial serum HBV DNA titer (HR, 5.90; 95% CI, 1.64 to 21.3; P = 0.007), liver cirrhosis (HR, 10.4; 95% CI, 2.38 to 44.9; P = 0.002), and distant metastasis (HR, 5.14; 95% CI, 1.54 to 17.2; P = 0.008) were the independent risk factors for withdrawal hepatitis (Fig. 2B and Table 6). The association between the total duration of LAM therapy and the occurrence of withdrawal hepatitis was not statistically significant. Prolonged exposure to additional LAM therapy after completion of chemotherapy tended to preclude withdrawal hepatitis from occurring, although this did not reach significant difference (P = 0.091) (Table 7).

Table 6.

Risk factors for withdrawal hepatitis by univariable and multivariable analyses^a

Risk factor for WH	Univariable analysis		Multivariable analysis
	Unadjusted HR (95% CI)	P value	Adjusted HR (95% CI)	P value
Gender
Female	1.00
Male	1.46 (0.53–4.02)	0.468
Age, yr
<50	1.00
≥50	1.53 (0.49–4.80)	0.470
HBeAg
Negative	1.00
Positive	1.12 (1.15–8.58)	0.916
HBV DNA titer, IU/ml
<2,000	1.00		1.00
≥2,000	5.55 (1.85–16.7)	0.002	5.90 (1.64–21.3)	0.007
Liver cirrhosis
No	1.00		1.00
Yes	9.68 (3.37–27.8)	<0.0001	10.4 (2.38–44.9)	0.002
Type of malignancy
Oncologic	1.00
Hematologic	2.73 (0.97–7.68)	0.057
Use of steroid
No	1.00
Yes	0.59 (0.21–1.65)	0.313
Use of anthracycline
No	1.00
Yes	1.34 (0.49–3.70)	0.571
Use of rituximab
No	1.00
Yes	3.02 (0.96–9.48)	0.059
Distant metastasis
No	1.00		1.00
Yes	2.81 (0.98–8.03)	0.054	5.14 (1.54–17.2)	0.008

^aCI, confidence interval; HR, hazard ratio; HBeAg, hepatitis B e antigen; HBV, hepatitis B virus; WH, withdrawal hepatitis. Note that since none of the patients with primary liver cancer showed withdrawal hepatitis, the hazard ratio of primary liver cancer for withdrawal hepatitis could not be estimated.

Table 7.

Relationship between the duration of prophylactic lamivudine therapy and withdrawal hepatitis^a

LAM therapy duration	Result	Mean ± SD duration (days)	P value
Total	WH−	304.11 ± 203.91	0.195
	WH+	226.07 ± 97.75
Additional therapy	WH−	142.61 ± 184.49	0.091
	WH+	84.46 ± 47.44

^aLAM, lamivudine; WH, withdrawal hepatitis.

The results from the conditional inference tree structure indicate that 30% of patients receiving additional LAM therapy for less than 166 days showed withdrawal hepatitis, while none of the patients taking it for more than 166 days showed withdrawal hepatitis (P = 0.249) (Fig. 3 and Table 8).

Fig 3.

The conditional inference tree structure according to the duration of additional lamivudine (LAM) therapy. Node 2 indicates the population who received additional LAM for less than 166 days after completing chemotherapy, and node 3 indicates the population receiving it for more than 166 days. The gray box indicates the subjects without withdrawal hepatitis, and the black box indicates those with withdrawal hepatitis.

Table 8.

Determination of the optimal duration of additional lamivudine treatment using the conditional inference tree structure^a

Additional LAM duration (days)	Total no. of patients	No. of patients with WH	Frequency of WH
≤166	50	15	0.3
>166	11	0	0

^aLAM, lamivudine; WH, withdrawal hepatitis.

DISCUSSION

In this retrospective evaluation of 110 HBsAg-positive patients with oncologic and hematologic malignancies undergoing chemotherapy, we analyzed the predictors of virologic breakthrough and withdrawal hepatitis. We demonstrated that a high viral load (≥2,000 IU/ml) was a significant predictor of both virologic breakthrough and withdrawal hepatitis. These findings provide convincing evidence beyond the expert opinion in the recent updated guidelines (19), highlighting the role of a viral load in the setting of prevention of HBV reactivation following cytotoxic chemotherapy.

When prophylactic LAM therapy is utilized during chemotherapy, HBV replication might be kept under control, preventing hepatitis flare from occurring. Therefore, based on the recent studies (20), the prophylactic use of LAM for HBsAg-positive patients undergoing chemotherapy is routinely recommended; however, the optimal duration of LAM therapy after completion of chemotherapy is still under debate. Several investigators have suggested that LAM treatment of inactive HBV carriers should continue for 6 months following termination of chemotherapy (18, 22, 24), while others insist on up to a year of postchemotherapy LAM treatment as there have been cases with HBV reactivation 12 months after completion of chemotherapy (12, 26).

Withdrawal hepatitis usually occurs when LAM is withdrawn shortly after completion of chemotherapy and the dormant HBV is allowed to replicate. In this study, about 25% of the patients who had completed chemotherapy developed withdrawal hepatitis within a median of 2.4 months after discontinuation of LAM. However, prolonged exposure to LAM also has been known to induce LAM resistance-associated mutations in the HBV genome, increasing the potential for virologic or biochemical breakthrough. The results from our study suggest that the optimal duration of additional LAM therapy should be longer than at least 166 days to prevent withdrawal hepatitis. Given the high emergence rate of LAM resistance mutations, the optimal duration of additional LAM treatment would be around 5 to 6 months.

Our results showed that the independent predictors of virologic breakthrough were high initial HBV DNA titer (≥2,000 IU/ml) and rituximab treatment. A high pretreatment viral load of HBV is associated with an increase in the rate of LAM resistance, leading to virologic and biochemical breakthrough (30). The emergence rate of YMDD motif mutants has been reported at 24% by 1 year and up to 70% by 4 years in CHB patients treated with LAM (14). In the present study, virologic breakthrough occurred in 15 of 110 patients (13.6%) and LAM resistance-associated mutations were found in only 7 patients (6.4%) by 2 years. Thus, the incidence of the YMDD motif mutation among cancer patients receiving chemotherapy is likely to be lower than among those with CHB, because the viral load of HBV in CHB patients is much higher than those in the usual cancer patients carrying HBsAg. The early rise in HBV DNA titers due to the immunosuppressive effect of chemotherapy might partly account for the 8 patients who developed virologic breakthrough without evidence of LAM resistance.

HBV reactivation is more common even in lymphoma patients with prior resolved hepatitis B when rituximab, a humanized anti-CD20 monoclonal antibody, is used as a chemotherapeutic regimen (5, 28). Treatment with rituximab induces apoptosis of malignant and normal B lymphocytes and profound B-cell depletion for 6 to 9 months (21). Accordingly, HBsAg-positive patients with a high viral load undergoing cytotoxic chemotherapy or those receiving rituximab for non-Hodgkin lymphoma may be at risk of an emergence of LAM-resistant strains of HBV and hence should be protected by antiviral agents with a high genetic barrier to resistance. In the future, a large-scale, prospective study is warranted to compare the efficacy and cost-effectiveness of antiviral prophylaxis between LAM and other more potent nucleoside or nucleotide analogues for HBsAg-positive and/or anti-hepatitis B core IgG-positive patients with non-Hodgkin lymphoma receiving rituximab.

The cumulative incidence of virologic breakthrough has increased in large part after 12 months as the number of at-risk patients has much decreased after 12 months, while virologic breakthrough has still occurred during the same period. The vast majority of patients (65/80 [81.3%]) were censored after 12 months because LAM therapy was withdrawn (n = 50) or withdrawal hepatitis occurred (n = 15) within 12 months, so that they were no longer at risk by definition, for virologic breakthrough (see Table S3 in the supplemental material). However, none of the patients developed virologic breakthrough after 2 years. Withdrawal hepatitis occurred in 11 of the 15 patients (73.3%) between 6 and 12 months, which accounted for the high actuarial rate of withdrawal hepatitis after 6 months. However, none of the patients developed withdrawal hepatitis after 1 year (see Table S3).

In the present study, the significant predictors of withdrawal hepatitis by multivariable analysis were a high baseline HBV DNA titer (≥2,000 IU/ml), the presence of liver cirrhosis, and distant metastasis. According to previous reports, the prechemotherapy viral load of HBV has also been significantly associated with withdrawal flare due to HBV reactivation, which is in line with our results (10, 15). In the present study, the patients with HBV-related cirrhosis and normal aminotransferase levels, who are now indicated for lifelong antiviral treatment (19), might have been included in the study population by chance. LAM was usually withdrawn 3 to 6 months after completion of chemotherapy in these patients, since South Korean public health insurance did not cover the cost of antiviral therapy in cirrhotic patients with normal aminotransferase levels during our study period. Nonetheless, based on the results of our study, liver cirrhosis was definitely a significant risk factor for withdrawal hepatitis, offering convincing evidence that cirrhotic patients should be treated with lifelong antiviral therapy even in the setting of prevention of HBV reactivation. The patients with distant metastasis received shorter-duration therapy of additional LAM due to the poor oral tolerability than those without distant metastasis, which could be a possible reason for a higher occurrence rate of withdrawal hepatitis (data not shown).

Unfortunately, the duration of prophylactic LAM after completion of chemotherapy was highly varied among patients in our study. It is well acknowledged that the risk of virologic breakthrough and withdrawal hepatitis may vary with the duration of LAM therapy (30), and as a result, this potential confounding factor might be a limitation of the present study. Moreover, the various treatment strategies employed against prophylactic failure provided no common approach regarding the optimal management algorithm for hepatic events, such as virologic breakthrough and withdrawal hepatitis. The number of incurred events in our study was too small to determine the optimal duration of LAM treatment and compare the efficacy of rescue therapy. Thus, a similar study on a larger scale might be necessary to validate our findings in the future. Nonetheless, to our knowledge, this is the first study to analyze risk factors for prophylactic failure in HBsAg-positive patients with oncologic and hematologic malignancies receiving antiviral prophylaxis by adjusting competing risk events, such as malignancy-related death.

In conclusion, when HBsAg-positive patients with oncologic and hematologic malignancies undergo chemotherapy and LAM prophylaxis, careful selection of high-risk patients for prophylactic failure is mandatory to maximize the therapeutic efficacy of chemotherapy and reduce the risk of virologic breakthrough and hepatitis flare. Thus, it should be feasible to adopt individualized therapy for the high-risk patients: antiviral agents with a high genetic barrier to resistance for those with high serum HBV DNA levels (≥2,000 IU/ml) or hematologic malignancies, especially using rituximab, and extended duration of antiviral prophylaxis for those who have a high viral load, underlying liver cirrhosis, or distant metastasis before chemotherapy. Therefore, risk stratification and tailored therapies might eventually prevent HBV reactivation resulting from prophylactic failure and avoid unwanted interruption or delay of chemotherapy.