Prognostic role of whole blood Epstein‑Barr virus DNA load in angioimmunoblastic T-cell lymphoma

Abstract

Objective

The prognostic role of Epstein-Barr virus (EBV) DNA in aggressive angioimmunoblastic T-cell lymphoma (AITL) remains unclear. This study aimed to investigate the prognostic significance of whole blood EBV DNA at diagnosis in AITL patients, establish an optimal prognostic cutoff, and develop a novel prognostic index incorporating this viral biomarker.

Methods

This multicenter retrospective study included 323 newly diagnosed AITL patients. The optimal cutoff of EBV DNA was determined using maximally selected rank statistics. Independent prognostic factors identified by multivariate Cox regression were used to construct a novel prognostic index (AITL prognostic index [API]), which was internally validated and compared with existing models.

Results

At diagnosis, 241 (74.6%) AITL patients were EBV DNA positive, with 143 (44.3%) exceeding the optimal cutoff for EBV DNA load of 6110 copies/mL. EBV DNA > 6110 copies/mL was established as an independent adverse factor for overall survival (HR, 1.553; p =.006), along with age ≥ 60 years, ECOG ≥ 2, platelets < 150 × 10⁹/L, and extranodal sites > 1. Based on these thresholds, API was used to stratify patients into four distinct risk groups with significantly different median OS (p <.001) and demonstrated superior discriminatory power (Harrell’s C-statistic: 0.741; Akaike Information Criterion value: 1829.1) compared to IPI, PIT, PIAI, and AITL score.

Conclusions

Whole blood EBV DNA load is a powerful, independent prognostic index in AITL. The novel prognostic index API integrates a quantitative EBV DNA cutoff, providing superior risk stratification over established prognostic models, and guideing clinical decision-making for AITL patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13027-026-00739-6.

Introduction

Angioimmunoblastic T-cell lymphoma (AITL) is the second most common subtype of peripheral T-cell lymphoma (PTCL), accounting for approximately 1%-2% of all non-Hodgkin lymphomas and 15%-20% of all PTCL cases [1, 2]. AITL originates from T follicular helper (TFH) cells and is classified as a subtype of nodal TFH cell lymphomas (nTFHLs) in the 5th edition of the World Health Organization (WHO) classification [3]. AITL follows an aggressive clinical course with a poor prognosis, characterized by progressive lymphadenopathy, hepatosplenomegaly, constitutional symptoms, anemia, and hypergammaglobulinemia [4]. Previous studies have reported the 5-year overall survival (OS) at 32%-41% [5] and event/progression/failure-free survival at 18%-38% [4, 6, 7]. Currently, no standard treatment strategy exists for AITL [8]. First-line treatment is based on CHOP (cyclophosphamide, doxorubicin, vincristine, prednisone) chemotherapy, and while the addition of etoposide (CHOEP) may further improve progression-free survival (PFS), it still has limitations in efficacy [9–11]. Therefore, there is an urgent need to improve treatment strategies and refine risk stratification for AITL patients.

Epstein-Barr virus (EBV) is a ubiquitous human γ-herpesvirus that primarily infects host B cells and establishes lifelong latency. It is involved in the development of lymphoid malignancies, including Burkitt lymphoma (BL), natural killer (NK)/T-cell lymphoma (NKTL) and Hodgkin lymphoma (HL) [12–15]. In AITL, EBV infection has also been implicated in disease pathogenesis and progression [8, 16, 17]. The diagnostic gold standard for EBV active infection is the detection of EBV-encoded small RNAs (EBERs) by in situ hybridization (ISH) [18], whereas its clinical application is limited by sample requirements and technical expertise. An alternative method for viral detection and quantification is the use of real-time quantitative polymerase chain reaction (RQ-PCR) to detect circulating EBV DNA. As EBV DNA is readily obtainable from peripheral blood, its dynamic changes can be monitored long-term at initial diagnosis and post-treatment to assess the disease burden.

Pre-treatment EBV DNA positivity has been demonstrated to correlate with poor prognosis in various EBV-related lymphomas, including HL [19], extranodal NKTL (ENKTL) [14], and diffuse large B-cell lymphoma (DLBCL) [20]. Meanwhile, the prognostic value of EBER status remains controversial [21, 22]. As for AITL, few studies have focused on the prognostic role of EBV DNA [23], as well as the dynamic changes in EBV DNA load post-treatment and at relapse. To better understand the prognostic role of EBV DNA in AITL patients, this study aimed to explore the correlation between clinical features, treatment response and survival outcomes depending on the presence and load of whole blood EBV DNA at diagnosis, as well as its subsequent dynamic changes. As a result, a prognostic index was established that incorporates EBV DNA load at diagnosis as a risk factor.

Methods

Patients

This multicenter retrospective study enrolled newly diagnosed adult AITL patients from four tertiary hospitals in China between August 2015 and December 2023. Diagnosis was confirmed by histopathology according to the 2016 WHO classification criteria [24]. Patients lacking baseline EBV data were excluded. Approval was obtained from the institutional review board at the coordinating center (The First Affiliated Hospital of Medical School, Zhejiang University, Hangzhou, China) and also at each participating center as per the institutional standard. Informed consent was obtained from all participants.

Data were collected on baseline clinical and disease characteristics, first-line treatment, treatment outcomes, survival status and cause of death. Treatment regimen selection and consideration for autologous stem cell transplantation (ASCT) were determined by the treating physician. Response evaluation was categorized as complete remission (CR), partial response (PR), stable disease (SD), or progressive disease (PD), in accordance with the International Working Group Criteria. Overall response rate (ORR) was defined as the proportion of PR or better. EBV-DNA was routinely measured at the end of each treatment cycle and during subsequent follow-up visits.

Identification of EBV DNA and EBER

EBV DNA in whole blood samples was detected using RQ-PCR in the molecular microbiology laboratory of our institution after DNA extraction with the EBV PCR Fluorescence Quantitative Diagnostic Kit (ZJ Bio-Tech, Shanghai, China). The results were expressed as copies/mL of total EBV-DNA, with copy numbers analyzed using a standard curve. The detection limit for this assay was 500 copies/mL, and values exceeding this threshold were considered EBV DNA positive. If EBV DNA was less than 500 copies/mL, it was recorded as 0 and considered EBV DNA negative.

The presence of EBV infection in paraffin-embedded sections was assessed using EBER-ISH. Consistent with the relevant literature and the pathological findings from this study [7, 25], EBER-ISH positivity was determined by counting the number of EBER-positive cells per high-power field (HPF). Cases were classified into three groups: EBER positivity (defined as ≥ 30 cells/HPF in the highest-density region), EBER weak positivity (> 0 but < 30 cells/HPF), and EBER negativity (0 cells/HPF).

Prognostic index

The International Prognostic Index (IPI) defines four risk groups based on age, Eastern Cooperative Oncology Group (ECOG) performance status (PS), Ann Arbor stage, lactate dehydrogenase (LDH) level, and the number of extranodal sites: low (0–1), low-intermediate (2), high- intermediate (3), and high (4–5) [26]. The Prognostic Index for T-cell lymphoma (PIT) establishes four risk groups based on age, ECOG PS, LDH level, and bone marrow involvement: Group 1 (0), Group 2 (1), Group 3 (2), and Group 4 (3–4) [27]. The Prognostic Index for AITL (PIAI) delineates two risk groups based on age, B symptoms, ECOG PS, platelet count and the number of extranodal sites: low (0–1) and high (2–5) [7]. The AITL score was determined by age, ECOG PS, β2-microglobulin (β2-MG) and C-reactive protein (CRP) levels, classifying patients into three risk groups: low (0-1), intermediate (2), and high (3–4) [28].

Statistical analysis

The optimal cutoff value of EBV DNA at diagnosis was established using the method of maximally selected rank statistics (maxstat package, R software, Vienna, Austria). To address the left-censoring of undetectable viral loads, EBV DNA levels below 500 copies/mL were imputed at 500 copies/mL. Given the exponential distribution of EBV DNA levels across the cohort, all viral load measurements underwent log₁₀ transformation prior to statistical analysis. To avoid overfitting, we performed internal validation through nonparametric bootstrap resampling with 1,000 iterations and calculated the 95% bootstrap confidence interval (CI) for the optimal cutoff.

Categorical variables were compared using Chi-square test or Fisher’s exact test, while continuous variables were compared using Mann-Whitney tests. OS was calculated from the date of initial diagnosis to the date of death from any cause, with censoring for patients alive at the last follow-up. Survival curves were estimated using the Kaplan-Meier method and survival differences between groups were evaluated using the log-rank test, with Bonferroni correction applied for multiple comparisons. Univariate and multivariate analyses were performed using the Cox proportional hazards model to evaluate the prognostic impact of different variables on OS, with hazard ratios (HRs) used as a summary measure. Covariates identified as having an independent prognostic impact on OS were used to create a novel prognostic index (AITL prognostic index [API]). When comparing all prognostic indices, a measure of global fit (Akaike Information Criterion [AIC]) and concordance index (Harrell’s C-statistic) were used, with lower AIC values indicating a better fit and higher Harrell’s C-statistic values indicating better discrimination [29]. The internal validation of the new prognostic model was performed using bootstrap resampling. A two-tailed p value < 0.05 was considered statistically significant. Statistical analyses were performed using SPSS (version 26.0), R software (version 4.4.0), and GraphPad Prism 9.

Results

Patient characteristics stratified by EBV DNA status

A total of 323 AITL patients were included in this study. The median age was 67 years (interquartile range [IQR], 57.5–73.0), with 189 (58.5%) males and 225 (69.7%) patients aged ≥ 60 years. A total of 314 patients (97.2%) presented with Ann Arbor stage III-IV, and 269 patients (83.3%) had higher β2-MG levels (> 2.5 mg/L). More than half of the patients (62.8%) had elevated LDH levels (> 250 U/L).

Whole blood EBV DNA was detected in 241 (74.6%) patients at diagnosis. The median EBV DNA load of AITL was 4100 copies/mL (range, 0-8.71 × 10⁶ copies/mL). The clinical characteristics were compared between two groups with different EBV DNA status in Table 1. Patients with EBV DNA positivity exhibited higher LDH (67.2% vs. 50.0%, p =.008), β2-MG (87.6% vs. 70.7%, p <.001) and CRP (72.6% vs. 57.3%, p =.015), along with reduced albumin levels (median, 35.9 vs. 39.8 g/L, p <.001) compared to EBV DNA-negative cases. Additionally, EBV DNA-positive status correlated with poorer ECOG PS (51.0% vs. 34.1%, p =.012), increased bone marrow involvement (58.1% vs. 32.9%, p <.001), and higher-risk prognostic scores, including IPI ≥ 3 (70.5% vs. 54.9%, p <.001), PIT ≥ 2 (78.8% vs. 58.5%, p =.039), PIAI ≥ 2 (70.5% vs. 57.3%, p =.001) and AITL score ≥ 2 (86.7% vs. 73.2%, p =.005).

Table 1.

Clinical characteristics of AITL patients according to the EBV DNA status and cutoff

Characteristic	EBV DNA-negative	EBV DNA-positive	p value		EBV DNA ≤ 6110 copies/mL	EBV DNA > 6110 copies/mL	p value	Total
	(n = 82)	(n = 241)			(n = 180)	(n = 143)		(n = 323)
Male sex	41 (50.0)	148 (61.4)	0.093		104 (57.8)	85 (59.4)	0.851	189 (58.5)
Age, y	64.5 (55.3–71.8)	67.0 (58.0–74.0)	0.116		64.0 (57.0–72.0)	68.0 (59.0–75.0)	0.029	67.0 (57.5–73.0)
Age ≥ 60 y	52 (63.4)	173 (71.8)	0.199		120 (66.7)	105 (73.4)	0.234	225 (69.7)
B symptoms	40 (48.8)	143 (59.3)	0.124		82 (45.6)	101 (70.6)	< 0.001	183 (56.7)
Stage Ⅲ-Ⅳ	81 (98.8)	233 (96.7)	0.457		174 (96.7)	140 (97.9)	0.736	314 (97.2)
ECOG ≥ 2	28 (34.1)	123 (51.0)	0.012		67 (37.2)	84 (58.7)	< 0.001	151 (46.7)
Hemoglobin < 10 g/dL	18 (22.0)	78 (32.4)	0.101		43 (23.9)	53 (37.1)	0.014	96 (29.7)
PLT < 150 × 109/L	29 (35.4)	99 (41.1)	0.434		59 (32.8)	69 (48.3)	0.007	128 (39.6)
LMR < 1.7	37 (45.1)	122 (50.6)	0.464		88 (48.9)	71 (49.7)	0.981	159 (49.2)
LDH > ULN	41 (50.0)	162 (67.2)	0.008		97 (53.9)	106 (74.1)	< 0.001	203 (62.8)
β2M > ULN	58 (70.7)	211 (87.6)	< 0.001		138 (76.7)	131 (91.6)	< 0.001	269 (83.3)
CRP > ULN	47 (57.3)	175 (72.6)	0.015		108 (60.0)	114 (79.7)	< 0.001	222 (68.7)
Creatinine, µmol/L	69.0 (56.5–77.8)	72.0 (59.0–87.0)	0.291		69.0 (56.0-81.3)	74.0 (60.5–88.5)	0.026	71.0 (58.5–85.0)
Albumin, g/L	39.8 (33.9–45.5)	35.9 (31.5–40.0)	< 0.001		38.3 (33.2–43.6)	35.1 (30.9–38.8)	< 0.001	36.4 (32.2–41.9)
Extranodal sites > 1	8 (9.8)	39 (16.2)	0.213		23 (12.8)	24 (16.8)	0.392	47 (14.6)
Bone marrow involvement	27 (32.9)	140 (58.1)	< 0.001		77 (42.8)	90 (62.9)	< 0.001	167 (51.7)
IPI ≥ 3	45 (54.9)	170 (70.5)	0.014		106 (58.9)	109 (76.2)	0.002	215 (66.6)
PIT ≥ 2	48 (58.5)	190 (78.8)	< 0.001		115 (63.9)	123 (86.0)	< 0.001	238 (73.7)
PIAI ≥ 2	47 (57.3)	170 (70.5)	0.039		106 (58.9)	111 (77.6)	0.001	217 (67.2)
AITL score ≥ 2	60 (73.2)	209 (86.7)	0.005		139 (77.2)	130 (90.9)	0.001	269 (83.3)
EBER-ISH			0.001				< 0.001
Negative	26 (31.7)	38 (15.8)			51 (28.3)	13 (9.1)		64 (19.8)
Weak positive	42 (51.2)	126 (52.3)			93 (51.7)	75 (52.4)		168 (52.0)
Positive	10 (12.2)	64 (26.6)			24 (13.3)	50 (35.0)		74 (22.9)
Unknown	4 (4.9)	13 (5.4)			12 (6.7)	5 (3.5)		17 (5.3)

Data are shown as median (25%, 75%) or n (%). AITL, angioimmunoblastic T-cell lymphoma; EBV, Epstein-Barr virus; ECOG, Eastern Cooperative Oncology Group; PLT, platelet; LMR, lymphocyte-to-monocyte ratio; LDH, lactate dehydrogenase; ULN, upper limit of normal; β2M, β2-microglobulin; CRP, C-reactive protein; IPI, International Prognostic Index; PIT, Prognostic Index for T-cell lymphoma; PIAI, Prognostic Index for AITL; EBER, EBV-encoded small RNA

Among the 306 patients with available EBER-ISH testing results, 168 (52.0%) were weak positive and 74 (22.9%) were positive. A statistically significant association was observed between EBV DNA positivity and EBER status (p =.001). However, other clinical characteristics did not differ according to EBV DNA positivity.

Identification of the EBV DNA cutoff for OS

Using the maximally selected rank statistics method, we identified 6110 copies/mL (log₁₀ [EBV DNA]: 3.786; p <.001) as the optimal cutoff for OS. The 95% bootstrap CIs for this threshold were 629–20,400 copies/mL (log₁₀ 2.799–4.310) (Figure S1). In our cohort, 143 patients (44.3%) had EBV DNA levels exceeding this cutoff value.

Consistent with the findings observed between EBV DNA-positive and negative groups, patients with EBV DNA > 6110 copies/mL (vs. ≤ 6110 copies/mL) exhibited more aggressive disease features, including elevated LDH, β2-MG and CRP levels, along with reduced albumin, poorer ECOG PS, more frequent bone marrow involvement, and worse prognostic scores (all p <.05) (Table 1). Furthermore, the group with higher EBV DNA load had an older age (median, 68.0 vs. 64.0 years, p =.029), with higher rates of B symptoms (70.6% vs. 45.6%, p <.001), lower hemoglobin (37.1% vs. 23.9%, p =.014) and platelet levels (48.3% vs. 32.8%, p =.017), and elevated creatinine (median, 74.0 vs. 69.0 µmol/L, p =.026). Notably, EBV DNA > 6110 copies/mL was also significantly associated with EBER-ISH positivity in tumor tissue (p <.001).

Treatment regimens and responses

Treatment data were available for 306 patients, with the majority (82.7%) receiving CHOP chemotherapy with or without etoposide (61.7% and 20.9%, respectively). A total of 55 patients (18.0%) were treated with chidamide, 17 patients underwent consolidative ASCT, and one patient received ASCT in relapse. The selection of first-line chemotherapy regimens did not differ significantly across patients stratified by EBV DNA status or EBV DNA load.

Among the 291 patients with obtainable treatment responses, 131 (45.0%) achieved CR and 83 (28.5%) achieved PR, for an ORR of 73.5%. EBV DNA-positive patients demonstrated significantly lower CR rate than EBV DNA-negative patients (39.9% vs. 59.0%, p =.003), though the difference in ORR did not reach statistical significance (70.8% vs. 80.8%, p =.091). In contrast, stratification by the EBV DNA cutoff revealed more pronounced disparities. EBV DNA > 6110 copies/mL was significantly associated with inferior CR rate (34.4% vs. 53.0%, p =.002) and ORR (61.6% vs. 82.5%, p <.001). The complete treatment details were summarized in Table S1.

Impact of EBV DNA load at diagnosis on survival

To evaluate the prognostic impact of viral loads at diagnosis in EBV DNA-positive patients, we compared the survival time among patients with baseline EBV DNA levels of 1 × 10², 1 × 10³, 1 × 10⁴, 1 × 10⁵ and 1 × 10⁶ copies/mL. The results demonstrated a significant viral load-dependent association with survival (Fig. 1A, p =.005). Specifically, patients with higher baseline EBV DNA loads exhibited progressively worse OS. The 1-year OS rates declined with increasing EBV DNA loads: 84.0% for 1 × 10² copies/mL, 76.8% for 1 × 10³ copies/mL, 52.9% for 1 × 10⁴ copies/mL, and 44.8% for 1 × 10⁵ copies/mL. In the univariable Cox regression analysis, the HRs for OS compared with the group exhibiting EBV DNA load of 1 × 10² copies/mL were 1.51 (95% CI, 0.77–2.96), 2.39 (95% CI, 1.22–4.66), and 2.98 (95% CI, 1.41–6.31) for patients with EBV DNA loads of 1 × 10³, 1 × 10⁴ and 1 × 10⁵ copies/mL, respectively (p =.002). The 1 × 10⁶ cohort (n = 7) was excluded from comparison due to its limited statistical power.

Fig. 1.

OS of AITL patients according to the (A) baseline EBV DNA loads, (B) EBV DNA status, (C) EBV DNA cutoff, and (D) EBER-ISH status

Prognostic analysis

At the median follow-up of 45.5 months, 185 deaths were recorded, with disease progression (49.2%) being the most common cause, followed by infection (5.9%) and secondary malignancies (4.3%). The median OS for the entire cohort was 29.2 months (95% CI, 20.3–41.9 months), with a 1-year OS rate of 68.1% (95% CI, 63.2%–73.4%) and a 3-year OS rate of 46.7% (95% CI, 41.4%-52.6%).

Compared to patients with EBV DNA positivity, EBV DNA-negative patients achieved significantly better OS (median, 81.2 vs. 20.3 months, p <.001) (Fig. 1B). For the groups with EBV DNA negativity and positivity, the 1-year and 3-year OS rates were 78.0% vs. 64.7% and 65.3% vs. 40.2%, respectively. This survival advantage was also evident when stratified by viral load: patients with EBV DNA ≤ 6110 copies/mL had better survival outcomes than those with higher loads (median OS, 53.6 vs. 14.6 months, p <.001), with significantly improved 1-year (80.6% vs. 52.4%) and 3-year OS rates (58.3% vs. 31.9%) (Fig. 1C). In contrast, the EBER-ISH status showed no significant impacts on OS in AITL patients (Fig. 1D, p =.09).

The univariate analysis revealed that, in addition to EBV DNA positivity and a higher viral load (> 6110 copies/mL), age ≥ 60 years, the presence of B symptoms, ECOG PS ≥ 2, hemoglobin < 10 g/dL, platelets < 150 × 10⁹/L, LDH > 250 U/L, β2-MG > 2.5 mg/L, CRP > 8 mg/L, albumin < 30 g/L, creatinine > 133 µmol/L, bone marrow involvement, and extranodal sites > 1 were also significantly related to inferior OS (Table 2). After adjusting these variables in the multivariate analysis, age ≥ 60 years (HR, 1.839; 95% CI, 1.257–2.689), ECOG PS ≥ 2 (HR, 2.057; 95% CI, 1.425–2.969), platelets < 150 × 10⁹/L (HR, 2.043; 95% CI, 1.454–2.872), extranodal sites > 1 (HR, 1.771; 95% CI, 1.179–2.660), and EBV DNA > 6110 copies/mL at diagnosis (HR, 1.553; 95% CI, 1.137–2.122) remained independent adverse prognostic factors for OS (Table 2).

Table 2.

Univariate and multivariate regression analysis for OS

Characteristic	Univariate analysis		Multivariate analysis
	HR (95% CI)	p value	HR (95% CI)	p value
Male sex	0.776 (0.576–1.046)	0.096
Age ≥ 60 y	2.115 (1.488–3.007)	< 0.001	1.839 (1.257–2.689)	0.002
B symptoms	2.218 (1.633–3.013)	< 0.001	1.157 (0.795–1.682)	0.446
Stage Ⅲ-Ⅳ	2.419 (0.773–7.575)	0.129
ECOG ≥ 2	3.476 (2.568–4.705)	< 0.001	2.057 (1.425–2.969)	< 0.001
Hemoglobin < 10 g/dL	2.383 (1.767–3.213)	< 0.001	1.304 (0.934–1.821)	0.119
PLT < 150 × 109/L	2.980 (2.220–4.000)	< 0.001	2.043 (1.454–2.872)	< 0.001
LMR < 1.7	1.253 (0.938–1.673)	0.126
LDH > ULN	1.588 (1.162–2.170)	0.004	1.133 (0.803–1.597)	0.477
β2M > ULN	2.544 (1.577–4.102)	< 0.001	1.019 (0.595–1.746)	0.944
CRP > ULN	2.557 (1.793–3.647)	< 0.001	1.246 (0.820–1.892)	0.303
Albumin < 30 g/L	2.988 (2.099–4.255)	< 0.001	1.241 (0.811–1.897)	0.320
Creatinine > 133 µmol/L	2.529 (1.402–4.562)	0.002	1.235 (0.649–2.352)	0.520
Bone marrow involvement	2.195 (1.627–2.962)	< 0.001	1.028 (0.716–1.475)	0.882
Extranodal sites > 1	2.215 (1.550–3.166)	< 0.001	1.771 (1.179–2.660)	0.006
EBV DNA- positivity	1.990 (1.368–2.894)	< 0.001
EBV DNA > 6110 copies/mL	2.153 (1.608–2.882)	< 0.001	1.553 (1.137–2.122)	0.006
EBER-ISH
Weak positive	1.305 (0.874–1.948)	0.193
Positive	1.478 (0.939–2.325)	0.091

OS, overall survival; HR, hazard ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; PLT, platelet; LMR, lymphocyte-to-monocyte ratio; LDH, lactate dehydrogenase; ULN, upper limit of normal; β2M, β2-microglobulin; CRP, C-reactive protein; EBV, Epstein-Barr virus; EBER, EBV-encoded small RNA

Longitudinal EBV DNA monitoring and clinical outcomes

Serial EBV DNA measurement was performed in 208 patients following first-line therapy, with additional monitoring during treatment and at the time of disease relapse or progression. Patients were classified into four patterns on the basis of EBV DNA dynamics: Category 1 (positive to negative, n = 82), Category 2 (persistently positive, n = 68), Category 3 (persistently negative, n = 49), and Category 4 (negative to positive, n = 9).

The treatment responses stratified by EBV DNA patterns were presented in Fig. 2A. EBV DNA-negative patients at diagnosis showed no significant association between post-treatment viral status and response (ORR: 87.8% vs. 77.8%, p =.597; CR rate: 67.3% vs. 44.4%, p =.262). In contrast, for EBV DNA-positive patients at diagnosis, conversion to EBV DNA-negativity after first-line therapy predicted superior outcomes, with significantly higher ORR (93.9% vs. 66.2%, p <.001) and CR rate (59.8% vs. 35.3%, p =.003) compared to those with persistent EBV DNA positivity.

Fig. 2.

Dynamic changes in EBV DNA could predict treatment efficacy and OS in AITL patients. (A) Serial EBV DNA monitoring and clinical outcomes. (B) OS stratified by post-treatment EBV DNA dynamics. (C) Serial EBV DNA monitoring in baseline EBV DNA-positive patients with CR. (D) Serial EBV DNA monitoring in baseline EBV DNA-positive patients with PR. (E) Serial EBV DNA monitoring in baseline EBV DNA-positive patients with SD/PD

Subsequently, the prognostic value of EBV-DNA changes after therapy was analyzed. Patients with persistent EBV DNA negativity (Category 3) had the longest median OS (81.2 months; 95% CI, 48.1-114.3 months), whereas those with persistent EBV DNA positivity (Category 2) exhibited the shortest median OS (28.5 months; 95% CI, 15.2–41.8 months) (Fig. 2B). Pairwise analyses demonstrated statistically significant differences in OS when comparing Category 1 with Category 2 and Category 2 with Category 3 after Bonferroni adjustment (p <.05 for both), with non-significant differences among other groups.

In EBV DNA-positive patients at diagnosis, the change in EBV-DNA loads by clinical course was associated with treatment responses. Among patients who initially achieved remission (CR/PR) with first-line therapy, the majority exhibited declining EBV DNA levels during and after treatment (CR: 59/73 [80.8%]; PR: 31/49 [63.3%]) (Fig. 2C, D). Among these responders, 36/56 (64.3%) patients demonstrated EBV DNA elevation at disease recurrence, with significantly higher viral loads at relapse compared to post-treatment measurements (median: 2810 vs. 0 copies/mL; log₁₀ (EBV DNA + 1): 3.449 vs. 0, p <.001). Conversely, 18/25 (72.0%) patients who had no response to first-line therapy (SD/PD) maintained EBV DNA positivity throughout the treatment, showing no significant treatment-induced reduction in viral loads (median: 11,195 vs. 4640 copies/mL; log₁₀ (EBV DNA + 1): 4.046 vs. 3.667, p =.08) (Fig. 2E).

Prognostic index development

On the basis of multivariate analysis, we established a novel prognostic index for AITL (API) that incorporated five independent risk factors for OS, each assigned 1 point: age ≥ 60 years, ECOG PS ≥ 2, platelets < 150 × 10⁹/L, extranodal sites > 1, and EBV DNA > 6110 copies/mL at diagnosis. Significant differences in OS were observed between adjacent risk categories: those scored 1 and 2 (p =.002), 2 and 3 (p <.001), as well as 3 as 4 (p =.007). However, survival outcomes did not differ between patients who scored 0 and 1 (p =.078) as well as 4 and 5 (p =.800). Consequently, patients were stratified into four distinct risk groups: low- (score ≤ 1), intermediate- (score = 2), high- (score = 3), and very high-risk (score ≥ 4). The median OS progressively decreased with increasing risk category: it was not reached (NR) in the low-risk group, 41.3 months in the intermediate-risk group, 10.7 months in the high-risk group, and 6.8 months in the very high-risk group (p <.001). The corresponding survival rates at 1-year were 93.4%, 77.2%, 47.9%, and 29.6%, while 3-year OS rates were 78.8%, 51.5%, 22.7%, and not applicable (NA) for low-, intermediate-, high- and very high-risk groups, respectively (Fig. 3).

Fig. 3.

OS stratified by AITL prognostic index

Both the API and established prognostic indices (IPI, PIT, PIAI, and AITL score) significantly predicted OS across all risk groups (p <.001 for each) (Table 3). Notably, the API exhibited superior discriminatory capacity for refined risk stratification. While the IPI showed inadequate differentiation between low- and low-intermediate-risk groups (adjusted p =.302), and the PIT failed to distinguish between Group 1 and 2 (adjusted p =.848) or Group 2 and 3 (adjusted p =.240), the API successfully overcame these limitations of both the IPI and PIT stratification systems (all adjusted p <.001), despite the exclusion of the API high-risk group in the IPI comparison and the API very-high risk group in the PIT comparison due to insufficient patient numbers (n = 3 for both) (Fig. 4). The quantitative assessment of model performance further confirmed the superiority of the API, as indicated by a lower AIC value (1829.1) and higher Harrell’s C-statistic (0.741) compared to other prognostic indices.

Table 3.

Comparison of survival and performance between the AITL prognostic index and established prognostic indices

Prognostic index	n (%)	Median OS (95% CI), mo	1-y OS (95% CI), %	3-y OS (95% CI), %	p value	Harrell C-statistic	AIC global fit
IPI					< 0.001	0.696	1878.3
Low risk (0–1)	32 (9.9)	NR (65.1-NR)	96.9 (91.0-100.0)	89.6 (79.1–100.0)
Low-intermediate risk (2)	76 (23.5)	NR (42.3-NR)	92.1 (86.2–98.4)	66.0 (55.6–78.3)
High-intermediate risk (3)	118 (36.5)	21.1 (15.8–53.6)	63.6 (55.4–72.9)	44.5 (36.1–54.7)
High risk (4–5)	97 (30.0)	9.7 (7.6–15.5)	45.4 (36.5–56.4)	19.5 (12.7–29.8)
PIT					< 0.001	0.696	1876.8
Group 1 (0)	26 (8.0)	NR (65.1-NR)	96.2 (89.0-100.0)	87.1 (74.5–100.0)
Group 2 (1)	59 (18.3)	NR (52.1-NR)	93.2 (87.0-99.9)	68.7 (57.3–82.5)
Group 3 (2)	86 (26.6)	42.3 (29.2-NR)	72.1 (63.2–82.2)	54.7 (44.9–66.6)
Group 4 (3–4)	152 (47.1)	12.7 (9.1–16.8)	51.3 (44.0-59.9)	26.1 (19.7–34.6)
PIAI					< 0.001	0.742	1843.5
Low risk (0–1)	106 (32.8)	NR (65.1-NR)	94.3 (90.0-98.8)	79.2 (71.4–87.9)
High risk (2–5)	217 (67.2)	15.2 (11.1–18.3)	55.3 (49.1–62.3)	30.6 (24.8–37.6)
AITL score					< 0.001	0.706	1872.2
Low risk (0–1)	54 (16.7)	NR (65.1-NR)	96.3 (91.4–100.0)	85.3 (75.6–96.2)
Intermediate risk (2)	74 (22.9)	67.3 (37.6-NR)	85.1 (77.4–93.6)	61.7 (51.2–74.4)
High risk (3–4)	195 (60.4)	14.3 (10.1–17.3)	53.8 (47.3–61.3)	30.1 (24.1–37.6)
AITL prognostic index (API)					< 0.001	0.741	1829.1
Low risk (0–1)	106 (32.8)	NR	93.4 (88.8–98.2)	78.8 (71.1–87.4)
Intermediate risk (2)	92 (28.5)	41.3 (27.5-NR)	77.2 (69.1–86.2)	51.5 (42.0-63.3)
High risk (3)	71 (22.0)	10.7 (8.4–17.3)	47.9 (37.6–61.0)	22.7 (14.3–36.1)
Very high risk (4–5)	54 (16.7)	6.8 (4.5–10.1)	29.6 (19.6–44.7)	NA

AITL, angioimmunoblastic T-cell lymphoma; OS, overall survival; AIC, Akaike Information Criterion; IPI, International Prognostic Index; PIT, Prognostic Index for T-cell lymphoma; PIAI, Prognostic Index for AITL; NR, Not Reached (> 50% patients alive at last follow-up); NA, Not Applicable (all patients died/lost to follow-up before 3 years)

Fig. 4.

OS stratified by established prognostic indices. (A) OS stratified by IPI. (B) OS of the IPI low- and low-intermediate-risk groups stratified by API. (C) OS stratified by PIT. (D) OS of the PIT Groups 1–3 stratified by API. (E) OS stratified by PIAI. (F) OS stratified by AITL score

Due to the low prevalence of AITL and the limited availability of independent external cohorts, internal validation of the API model was performed using bootstrap resampling (n = 1000). This approach demonstrated excellent consistency in risk stratification, with survival differences between all four risk groups maintaining statistical significance (log-rank p <.05) in every resampled dataset. The model showed robust discriminative performance, exhibiting minimal variation in the Harrell’s C-statistic across bootstrap samples (mean, 0.742; 95% CI, 0.707–0.769). The model fit remained stable, as evidenced by consistent AIC values between the original model and bootstrap replicates (mean, 1823.9).

Discussion

In this study, we confirmed the prognostic role of whole blood EBV DNA in AITL patients and identified EBV DNA > 6110 copies/mL at diagnosis as an independent adverse prognostic factor. To the best of our knowledge, this is the largest cohort study to date focused on the correlation of whole blood EBV DNA with clinical features, treatment outcomes, and prognosis in AITL patients. Furthermore, both baseline EBV DNA load and its dynamic changes after first-line therapy closely correlated with treatment response and survival. Finally, we established a novel prognostic index based on the EBV DNA cutoff and four other factors, demonstrating its capability for risk stratification and prognostic prediction for AITL patients.

The prognostic significance of EBV DNA varies across different PTCL subtypes. While both EBV DNA status and levels are established independent prognostic factors in ENKTL [14, 19, 30, 31], their role in other PTCL subtypes, particularly in AITL, remains poorly understood. A study by Chan et al., which distinguished NKTL from other PTCLs (including AITL, ALCL, and PTCL-NOS), concluded that the prognostic impact of EBV DNA was confined to NKTL [32]. In contrast, further studies have consistently supported its prognostic significance in AITL. Both Zhao et al. and Chen et al. found that elevated pre-treatment plasma EBV DNA was an independent unfavorable prognostic factor for AITL patients after adjusting for PTCL pathological subtypes [33, 34]. Similarly, Kim et al. demonstrated that EBV positivity detected by ISH or PCR in 68 AITL patients was associated with poorer prognosis [35]. In a cohort of 60 AITL patients receiving DA-EPOCH (Dose-adjusted Etoposide, Prednisone, Oncovin, Cyclophosphamide, and Hydroxydaunorubicin) regimens, whole blood EBV DNA was shown to be an independent risk factor for inferior PFS and OS [23]. Despite these findings, the corresponding studies were often limited by variable EBV DNA positivity rates and small sample sizes in non-ENKTL PTCL subtypes, precluding firm conclusions.

Our study, conducted on a relatively large and specific cohort of AITL patients, reinforces the prognostic value of EBV DNA by confirming that its presence in whole blood is associated with worse OS. Furthermore, a key contribution of our work is the establishment of a clinically relevant prognostic cutoff value. Previous studies in ENKTL or PTCL have reported widely variable prognostic cutoffs (ranging from 500 to 5000 copies/mL), which may be related to disease stage, treatment regimen and the heterogeneity of PTCL pathological subtypes [19–33, 36, 37]. In contrast, our analysis is the first to identify 6110 copies/mL as the optimal prognostic cutoff for AITL patients. This quantitative threshold demonstrated superior prognostic power over a binary positive/negative result, with a higher HR and narrower 95% CI in univariate Cox regression analysis, a finding explained by the fact that the optimal EBV DNA cutoff for OS was substantially higher than the detection limit. This validates the hypothesis that the increase in EBV DNA was attributable to the release of EBV DNA from the tumor [38], thus suggesting that high EBV DNA levels, beyond mere positivity, can quantify the tumor burden, accurately assess disease severity and stratify risk, thereby better aiding disease monitoring and prognosis.

The distinct pathogenic mechanisms of EBV may contribute to the dissociation between viral load and the prognosis of AITL and other PTCLs. In ENKTL, EBV directly infects and transforms T or NK cells into tumor cells, with studies showing the virus infecting and being persistently detected in tumor cells in 71% of cases [37]. However, in AITL, EBV is most commonly found in large B-lymphoblasts, although it has occasionally been observed in neoplastic T cells, with the virus detected in 85–95% of biopsies [17]. EBV infection and clonal expansion of B cells in AITL are often interrelated [16]. Furthermore, specific EBV genes associated with immune evasion and the promotion of tumor survival are strongly expressed in AITL patients [39]. These findings suggest that EBV reactivation may contribute to the development of the tumor microenvironment by impairing the immune system and promoting the expansion of TFH cells and B cells. EBV DNA is released from tumor cells or host cells upon primary infection or reactivation, allowing plasma, whole blood and peripheral blood mononuclear cells (PBMCs) to be chosen as samples for PCR measurement of EBV DNA levels. However, the optimal sample type remains uncertain [14, 31–34, 36, 40, 41]. The EBV DNA positivity rate in AITL patients was 74.6% in our study, consistent with previously reported rates ranging from 45.5% to 87.1% [32–34, 40]. This variability has often been linked with smaller cohort sizes, sample type differences, and differences in the experimental assay [42]. When whole blood samples are selected, circulating tumor cells or other EBV-infected cells can also reflect the patient’s viral load. Besides, EBV infection can also be assessed through EBER-ISH. Theoretically, EBER-ISH might represent a better marker for active viral transcription in tumor tissue, while its consistency with tumor burden reflected by circulating EBV DNA is not necessarily high [31, 34, 43]. Therefore, establishing an international standard for EBV-DNA detection calibration across different EBV-associated tumors will be a crucial step in quality assurance.

In this study, we analyzed the relationship between EBV DNA, prognosis, and EBER-ISH in AITL, finding that whole blood EBV DNA was prognostically significant, while EBER-ISH showed low correlation with EBV DNA and no impact on OS. Previous studies regarding EBV status and its prognostic significance in AITL cases have been controversial. For instance, Eladl et al. suggested that an EBER-negative status was an unfavorable prognostic factor for PFS in younger AITL patients [44]. In a study evaluating the efficacy of rituximab combined with CHOP, EBV DNA levels were closely correlated with quantitative EBER-ISH in lymph node biopsies, and higher viral load at diagnosis was associated with shorter PFS [45]. Conversely, others have reported that the presence of EBV+ cells does not alter prognosis [4, 6]. As a result, the clinical significance of EBV DNA detection in whole blood, PBMCs and plasma samples, along with the correlation between EBV DNA and EBER-ISH, need to be validated in larger randomized cohorts.

Regarding the link between EBV DNA and clinical characteristics, we found that whole blood EBV DNA positivity was associated with elevated LDH and CRP levels, reduced albumin levels, poorer ECOG PS, and increased bone marrow involvement. Furthermore, EBV DNA > 6110 copies/mL demonstrated a deeper connection with tumor burden, exhibiting more aggressive characteristics including elevated β2-MG levels, older age, higher rates of B symptoms, lower hemoglobin and platelet levels, and elevated creatinine. Among existing prognostic models for AITL, IPI, as a tool for assessing the prognosis of lymphoma patients, lacks specificity for T-cell biology [26]; PIT is more specific to PTCL but has uncertain prognostic ability for AITL [27]; PIAI, though designed for AITL, only stratifies patients into low- and high-risk groups with insufficient discriminatory power [7]; and the AITL score is a relatively newer model with limited validation data and lacking direct viral biomarkers [28]. Based on multivariate analysis, independent prognostic factors associated with OS were identified in our study. We not only confirmed the established prognostic value of age, ECOG PS, platelet count, and the number of extranodal sites, which are already incorporated into the IPI, PIT, PIAI, and AITL score, but also identified EBV DNA as a novel prognostic factor. We further proposed and refined a new AITL prognostic score (API). On the basis, patients were stratified into four groups: low-, intermediate-, high-, and very high-risk, overcoming the limitations of IPI and PIT in distinguishing the prognosis of lower-risk patients. This renders API more biologically-based, providing it with stronger discriminatory power than IPI, PIT, PIAI, and AITL score. Notably, patients with very high-risk exhibited a particularly poor prognosis, with a 1-year OS rate of only 29.6%, which may be attributed to reduced statistical power due to the small number of patients in this risk group.

Whether dynamic monitoring of EBV DNA can be used to predict treatment outcomes, disease recurrence and survival in non-ENKTL PTCL subtypes remains unclear. In our study, for patients with pre-treatment EBV DNA positivity, baseline viral loads could negatively impact the OS of AITL patients. EBV DNA negativity after first-line treatment was associated with better ORR and CR rates. Patients who responded to treatment showed a gradual decline in EBV DNA, while a rebound of viral load often indicated disease relapse. In non-responders, viral load exhibited no significant alteration during the course of treatment. These findings suggest that the continuous post-treatment measurement of EBV DNA is helpful for predicting response and survival outcomes in AITL patients, consistent with conclusions from previous studies on EBV-associated tumors [14, 31, 33, 46]. However, for patients with pre-treatment EBV DNA negativity, an increase in viral load after first-line treatment would theoretically suggest disease progression. In our study, treatment efficacy was actually acceptable, with no significant differences in median OS compared with other EBV DNA groups. This might be due to the limited number of events, restricting the correlation between dynamic changes in EBV DNA and survival.

Despite the promising results, this study has several limitations. First, it was retrospective in design. Second, the sample size was not insufficient, and the limited number of patients in some subgroups reduced statistical power. Third, the lack of immunophenotyping and genetic mutation analysis in AITL limited our ability to fully explain the pathogenic mechanisms of EBV at the biological level, which could provide new insights into potential therapeutic targets and mechanisms of treatment resistance. Fourth, the treatment regimens analyzed for AITL were limited to CHOP/CHOEP, chidamide and ASCT, without the further discussion of novel agents such as the anti-CD20 monoclonal antibody rituximab, the anti-CD30 antibody–drug conjugate brentuximab vedotin (BV), or epigenetic modifiers like romidepsin. The future of AITL treatment should shift from chemotherapy regimens to more individualized strategies, guided by both the patients’ biological risk and the tumor molecular characteristics. Accordingly, our research will focus on the benefits of immediately enrolling high-risk patients in novel drug clinical trials rather than standard chemotherapy. Fifth, due to incomplete EBV viral load data and inconsistent monitoring time points during follow-up, it was challenging to further analyze the dynamic changes of EBV DNA following first-line treatment and at relapse. Sixth, although we performed internal validation of the prognostic model using the bootstrap method, the risk of overfitting still persisted, and its generalizability could not be assessed. Therefore, multi-center, large‐sample prospective clinical studies are required to investigate the prognostic value of EBV DNA in AITL patients and to permit its widespread clinical application as a prognostic indicator through external validation. Concurrently, a deeper understanding of AITL pathogenesis is crucial for developing more specific targeted therapies for this disease characterized by poor prognosis.

Conclusions

In conclusion, our study provides strong evidence that whole blood EBV DNA is an independent prognostic factor in AITL, with its dynamic change capable of predicting treatment efficacy and survival outcomes. We also identified EBV DNA > 6110 copies/mL at diagnosis as the optimal prognostic cutoff, and found this threshold to be associated with more clinical characteristics related to tumor burden. Furthermore, we established a novel prognostic index incorporating this EBV DNA cutoff, which demonstrated superior performance over established prognostic indices in prognostic prediction for AITL patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

The work reported in the paper has been performed by the authors, unless clearly specified in the text. Conceptualization: YZ and MD. Data curation and Methodology: YZ, MD, FY, YZ, LZ, FZ, WW, YL, FG, LH. Formal analysis and Investigation: YZ. Resources and Supervision: XY, ZC, JZ, BY, QW, JH. Writing—original draft: YZ. Writing—review and editing: all authors.

Funding

This research was funded by National Natural Science Foundation of China (No. 82270206), Zhejiang Provincial Natural Science Foundation (No. LZ22H160009) and Key Project of the Zhejiang Higher Education Association’s Special Project on “Smart Education and Digital Transformation of Teaching” (KT2023416).

Data availability

All data and materials used and analyzed are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by the institutional review board at the coordinating center (The First Affiliated Hospital of Medical School, Zhejiang University, Hangzhou, China) and at each participating center as per the institutional standard. The patients provided their written informed consent to participate in this study.

Consent for publication

Written informed consent was obtained from the individuals for the publication of any potentially identifiable images or data included in this article.

Competing interests

The authors declare no competing interests.