Diagnostic utility and pitfalls of human T‐lymphotropic virus serology in adult T‐cell leukemia/lymphoma: evidence from a 25‐year pathology‐based nonendemic cohort

Abstract

Adult T‐cell leukemia/lymphoma (ATLL) is a distinct type of peripheral T‐cell lymphoma (PTCL) driven by human T‐lymphotropic virus type I (HTLV‐1)–infected T cells, but diagnosis can be confounded by histological overlap with non‐ATLL PTCL. While molecular testing is the diagnostic gold standard in endemic areas, HTLV‐1/2 serology is often used as a surrogate in nonendemic settings, yet its accuracy and limitations remain unclear. We retrospectively analyzed 881 PTCL cases over 25 years at two tertiary referral hospitals in Taiwan, where HTLV‐1 is nonendemic. Serology was available in 48.2% of cases. Molecular confirmation was performed using HBZ in situ hybridization and tax quantitative polymerase chain reaction. Among the 44 seropositive PTCL patients with tissue available, 37 were diagnosed and molecularly confirmed as ATLL. Of the seven initially diagnosed as non‐ATLL PTCL, three were reclassified as ATLL, while four showed no molecular evidence, including three likely false seropositives with borderline signal‐to‐cutoff (S/CO) values in serology assay and T follicular helper cell phenotype. In seropositive PTCL, clinicopathologic interpretation without molecular testing yielded 100% positive predictive value, 93% sensitivity, and 93% accuracy, compared with 91% positive predictive value by serology alone. In addition, retrospective molecular screening in 59 PTCL without prior serologic data revealed one case of ATLL that had initially been overlooked. In conclusion, our findings support routine HTLV‐1/2 serologic testing in all newly diagnosed PTCLs. However, interpretation should be guided by clinicopathological correlation and serology index values. Borderline index results warrant molecular confirmation.

Introduction

Adult T‐cell leukemia/lymphoma (ATLL) is a type of peripheral T‐cell lymphoma (PTCL) resulting from the monoclonal expansion of T cells infected by human T‐lymphotropic virus type I (HTLV‐1) [1, 2]. The incidence of ATLL exhibits considerable geographical variation, with higher rates in HTLV‐1–endemic areas such as southwestern Japan, the Caribbean, South America, inter‐tropical Africa, the Middle East, and Papua New Guinea. Once contracted, HTLV‐1 infection persists except for rare occasions [3]. The prevalence of HTLV‐1 carriers can be notably high, reaching 8% in endemic southwestern Japan [4]. In Taiwan, where HTLV‐1 is not endemic, the carrier rate is 0.034–0.058% in healthy blood donors [5, 6]. However, there are regional variations, such as higher carrier rates in Eastern Taiwan [6], and even up to 1.63% in some communities [7]. However, the lifetime cumulative incidence of ATLL in carriers remains at only 3–5% [8]. The annual incidence of ATLL in Japan stands at 8–11 cases per 100,000 population, while in nonendemic regions such as the United States and Taiwan, it is as low as 0.06 per 100,000 population [9, 10].

ATLL exhibits diverse clinical presentations, complicating its diagnosis. The Shimoyama classification categorizes ATLL into smoldering, chronic, lymphomatous, and acute subtypes [11]. Peripheral blood lymphocytosis with or without hypercalcemia raises clinical suspicion of ATLL; however, diagnosis of lymphomatous subtype ATLL is challenging due to its resemblance to other lymphomas, including ALK‐negative anaplastic large cell lymphoma (ALKnALCL), nodal T follicular helper‐cell lymphoma (nTFHL), other PTCL (including those with cytotoxic T phenotypes), and even classic Hodgkin lymphoma [12, 13, 14, 15, 16]. Correct identification of ATLL is important as the treatment strategies are unique [17].

In endemic regions, the diagnosis of ATLL typically includes demonstrating monoclonal integration of HTLV‐1 provirus by Southern blot, detecting viral transcripts in the majority of tumor cells by HTLV‐1 bZIP factor (HBZ) in situ hybridization (ISH), or identifying a high HTLV‐1 proviral load (PVL) by tax quantitative polymerase chain reaction (qPCR) in tumor tissue – based on the assumption that malignant cells harbor the viral genome [18, 19, 20]. In contrast, in nonendemic settings such as Taiwan, molecular testing for HTLV‐1 is not available in clinical diagnosis. Therefore, the diagnosis of ATLL relies primarily on identifying seropositivity to HTLV‐1/2 in PTCL patients [21, 22]. This approach assumes that (1) HTLV‐1/2 seropositivity in patients with non‐ATLL PTCL is close to the background carrier rate in the general population, and (2) HTLV‐2 infection is substantially less prevalent than HTLV‐1 and can be neglected [5]. Without molecular confirmation, some HTLV‐1/2 seropositive PTCL patients may be misdiagnosed as ATLL when they are merely HTLV‐1 carriers. In contrast, if molecular tests are available and show definitive evidence of proviral integration, ATLL should be the final diagnosis even when the clinicopathological features are atypical. Clinically, when molecular tests are unavailable, a diagnosis of ATLL relies on multidisciplinary discussion of the clinical presentation and pathological features, which can lead to occasional false‐positive or false‐negative diagnoses.

In essence, the diagnosis of ATLL in Taiwan relies on detecting HTLV‐1/2 seropositivity at the time of initial PTCL diagnosis. However, the clinical efficacy, accuracy, and limitations of this approach remain unassessed. In light of this, we aimed to incorporate HBZ ISH and tax qPCR, the current gold standard to detect HTLV‐1 integration in pathological specimens, to comprehensively assess the potential limitations in ATLL diagnosis without HTLV‐1 molecular testing in a nonendemic setting.

Materials and methods

Cohort and specimens from patients with PTCL

We conducted a retrospective observational cohort study. Formalin‐fixed paraffin‐embedded (FFPE) tissue blocks and sections bearing a diagnosis of PTCL of any type were retrieved from the pathology archives, including consultation files, of two institutions: National Taiwan University Hospital (NTUH) covering the years from 2000 to 2024 and National Taiwan University Cancer Center for the period spanning 2020 to 2024. We collected the original diagnoses from the electronic medical record. Because detailed diagnostic rationales were not consistently documented, challenging cases would have been discussed in a multidisciplinary team setting. All available glass slides were subjected to a thorough review by an experienced hematopathologist (CTY). Challenging cases were reviewed using multiheaded microscopy to reach a consensus (C‐TY, K‐CC, R‐CW, and S‐SC). Cases were excluded if slides were unavailable for review or the diagnosis of PTCL was uncertain.

For the PTCL cases, we retrieved HTLV‐1/2 serology status through the NTUH integrative Medical Database (NTUH‐iMD). The serology results were categorized as ‘ever positive’, ‘negative’, or ‘unknown’. Patients who had undergone multiple HTLV serology tests were classified as ‘ever positive’ if any test was positive. For HTLV‐1/2 serology, we used an enzyme immunoassay method (Murex HTLV‐1/2; Murex, Dartford, UK) until September 2012. Subsequently, we used a chemiluminescent microparticle immunoassay method (Architect; Abbott, Abbott Park, IL, USA). Clinical information was also obtained through NTUH‐iMD and electronic medical records.

This study was conducted in accordance with the Declaration of Helsinki. Ethics approval was granted by the institutional research ethics committees (REC number: 202309004RIND and 202407071RINA). The committees approved a waiver of consent for this delinked, retrospective observational study.

HBZ ISH for detecting HTLV‐1 virus transcripts in tumor cells

For PTCL cases suspected to be ATLL, we employed the HBZ ISH technique to ascertain the presence of HTLV‐1 virus transcripts within tumor cells. We utilized a commercially available HBZ probe (catalog no. 432901; ACDBio Inc., Newark, CA, USA). The ISH assay was conducted according to the manufacturer's protocols. The designation of ATLL relied on the degree of positivity for HBZ [20]. When HBZ ISH was negative, we performed ISH for a housekeeping gene PPIB (catalog no. 313901; ACDBio Inc., Newark, CA, USA) to ensure the RNA quality. Cases lacking both detectable HBZ and PPIB signals were considered invalid. We used cell blocks from established cell lines as experimental controls, including YT (a gift from Prof. Chung‐Wu Lin, Department of Molecular Medicine, Koo Foundation Sun Yat‐Sen Cancer Center; originally obtained from DSMZ, Braunschweig, Germany) [23, 24], an Epstein–Barr virus (EBV)‐associated NK lymphoma cell line, as the negative control, and HuT‐102 (purchased from the American Type Culture Collection [ATCC], Manassas, VA, USA), an HTLV‐1‐positive T‐cell lymphoma cell line, as the positive control (supplementary material, Figure S1).

tax qPCR for HTLV‐1 proviral load (PVL)

HTLV‐1 PVL was determined using an in‐house duplex qPCR targeting the tax viral gene and RPPH1 human housekeeping gene, performed on DNA extracted from FFPE tumor tissue using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany). An in‐house plasmid containing partial sequences of both tax and RPPH1 was constructed to generate the standard curve. This design simulates stable viral integration in the cell line control [25], and it minimizes dilutional errors and maximizes sample throughput per run (supplementary material, Figure S2) [26].

PCR was performed using TaqMan Fast Advanced Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) and template DNA on QuantStudio 3 real‐time PCR instrument (Thermo Fisher Scientific). The standard curve was established by serial dilutions of the in‐house tax‐RPPH1 plasmid (supplementary material, Figure S2).

PVL was calculated as the number of tax copies per 100 nucleated cells, assuming two alleles of RPPH1 per diploid genome. DNA from YT cells served as a negative control. PVL of DNA from TL‐Om1, an ATLL cell line developed by Sugamura et al [27], is approximately 100%, as expected [25]. For HuT‐102, the PVL is approximately 550%, in agreement with prior studies reporting values between 400% and 600% (supplementary material, Figure S3) [28, 29].

Immunohistochemical and in situ hybridization study

To characterize immunophenotypes, immunohistochemical (IHC) staining was performed as needed using an automated immunostainer (BENCHMARK XT or ULTRA, Ventana, Tucson, AZ, USA) to support diagnostic confirmation. The antibodies and staining conditions are detailed in supplementary material, Table S1.

EBV detection was conducted by in situ hybridization using an EBV‐encoded small RNA (EBER) DNP probe, the ISH iView kit (Ventana), and an automated immunostainer (BENCHMARK XT or ULTRA, Ventana).

Statistical analysis

Missing data were addressed using the available‐case method. Parameters were compared using the Wilcoxon rank‐sum test (α = 0.05). To account for multiple comparisons, p values were adjusted using the Benjamini‐Hochberg method (α = 0.05) [30]. We used logistic regression for serology check rate (α = 0.05). All statistical analyses were performed using the R language and RStudio [31, 32].

For details, refer to the Supplementary materials and methods.

Results

Determination of background noise of molecular tests

To determine the cutoff value of HTLV‐1 infection, we randomly selected 57 PTCL tumor specimens from 368 patients seronegative for HTLV‐1/2 (supplementary material, Table S2). In this group, the PVL ranged from 0% to 0.03%, with a standard deviation of 0.02%. Based on this distribution, we defined a PVL >0.1% (up to 5 standard deviations) as indicative of HTLV‐1 infection to strictly prevent a false‐positive result.

To determine the threshold of ATLL diagnosis, we aimed to model the PVL distribution in HTLV‐1 carriers who might happen to have non‐ATLL PTCL. To this end, we measured the PVL in non‐PTCL tumor samples from HTLV‐1/2–seropositive patients, as described in the literature (supplementary material, Figure S4) [20]. We analyzed 25 tumor specimens from 24 non‐PTCL patients (13 with solid cancers, 4 with benign mesenchymal tumors, and 8 with hematologic neoplasia collected from NTUH‐iMD between 2000 and 2024) but seropositive for HTLV‐1/2 (supplementary material, Table S3). The mean PVL in this group, plus three standard deviations, was 8.02%. Accordingly, and in line with the previous study [20], we set a working threshold of 10% PVL for the diagnosis of ATLL (supplementary material, Figure S3).

Therefore, PVL ranging from 0.1% to 10% suggests an HTLV‐1 carrier status, while those PTCL patients with PVL of more than 10% indicate ATLL patients. However, it is important to note that a PVL of 0% in tumor tissue does not exclude HTLV‐1 infection, given the stringent criterion used.

Notably, one esophageal carcinoma (SEPOpt0013, shaded row in supplementary material, Table S3) from a 46‐year‐old man seropositive for HTLV‐1/2 exhibited a PVL of 11.68%, exceeding the 10% threshold. He subsequently developed acute subtype ATLL 3 years later. Retrospective review of the initial esophageal carcinoma pathology slides revealed no evidence of ATLL at that time. However, the presence of occult ATLL at the time of esophageal carcinoma diagnosis could not be excluded.

HTLV‐1/2 seropositive PTCL is not necessarily ATLL

A total of 881 PTCL patients were included in this study. Among these, 57 (6.5%) were seropositive for HTLV‐1/2, 368 (41.8%) were seronegative, and 456 (51.8%) were not tested for HTLV‐1/2 serology.

Among the 57 seropositive cases (Figure 1 and supplementary material, Table S4), 11 were excluded due to the absence of FFPE material for molecular testing, and an additional 2 were excluded due to invalid results in both HBZ ISH and PVL by tax qPCR. Of the remaining 44 cases, 37 were originally diagnosed as ATLL and now confirmed by molecular testing (Figure 1, yellow rectangle frame; a representative case in Figure 2). Three cases, including two ALKnALCL and one PTCL, not otherwise specified (PTCL NOS), were reclassified as ATLL based on molecular evidence (Figure 1, green rectangle frame; green shaded in supplementary material, Table S4 and Figure S5). Four cases – two originally diagnosed as nodal T follicular helper cell lymphoma, angioimmunoblastic type (nTFHL‐AI, PTCLpt0685, and PTCLpt0822), one PTCL NOS (PTCLpt0575), and one NK‐large granular lymphocytic leukemia (NK‐LGL, PTCLpt0938) – showed no molecular evidence of ATLL despite HTLV‐1/2 seropositivity (Figure 1, blue rectangle frame; blue shaded in supplementary material, Table S4). Among these four patients, two cases (PTCLpt0685, a 58‐year‐old man; Figure 3; and PTCLpt0822, a 37‐year‐old man) showed classic histological features of nTFHL‐AI, including prominent vascular proliferation, hyperplastic follicular dendritic network, and medium‐sized lymphoid cells with pale or clear cytoplasm. Both tumors expressed sufficient follicular helper T markers. HBZ ISH score was 0 with well‐preserved PPIB expression, and PVL was 0% (RPPH1/tumor percentage: 8,834/80% for PTCLpt0685; 36,484/30% for PTCLpt0822). These findings supported a diagnosis of nTFHL‐AI, not ATLL, although the HTLV 1/2 serology was positive in these two patients. PTCLpt0575 (supplementary material, Figure S6) was a gastric biopsy specimen from a 78‐year‐old woman. Histology showed atypical medium to large‐sized lymphoid cells positive for CD3, GATA3, and BCL‐6, but negative for CD20, CD30, EBER, CD4, CD8, CD56, or TIA1. PET imaging revealed a gastric tumor with extensive abdominal lymphadenopathy. A diagnosis of PTCL, NOS with possible helper T‐cell differentiation was favored. HBZ ISH was scored 0 with preserved PPIB expression. PVL was 0% (RPPH1: 10,929 copies; estimated tumor percentage: 30%). PTCLpt0938 (Figure 4) was a 38‐year‐old man with increased large granular lymphocytes in peripheral blood (WBC: 10.1 × 10³/μl; lymphocytes: 22%; atypical lymphocytes: 52%). Flow cytometry revealed atypical NK cells with the immunophenotype CD2(+), CD7(+), sCD3(−), CD4(−), CD8(−), TCRβ(−), and TCRγ(−). Bone marrow PVL was 4.6% (tax copies: 50; RPPH1: 2,183; estimated tumor percentage: 30%). PVL from peripheral blood was 0.3% (tax copies: 45; RPPH1: 27,655; estimated tumor percentage: 50%). Taken together, these findings supported a diagnosis of NK‐large granular lymphocytic leukemia in an HTLV‐1 carrier.

Figure 1.

Tile plot of peripheral T‐cell lymphomas seropositive for HTLV‐1/2. Each column represents a tumor sample. The first row is the original diagnosis. The second and third rows are estimated tumor percentage and HTLV‐1 PVL by tax quantitative polymerase chain reaction. The fourth row is HBZ ISH. The last row is the revised diagnosis after molecular testing for HTLV‐1. If there was no FFPE available for molecular testing, the tile is labeled in black. If the FFPE was insufficient or the experiment was invalid, the tile is labeled in light gray. Most cases were confirmed to be ATLL after molecular testing (yellow rectangle frame). Notably, three cases were reclassified into ATLL based on molecular evidence (green rectangle frame), while four seropositive tumors showed no molecular evidence of ATLL (blue rectangle frame). FFPE, formalin‐fixed paraffin‐embedded tissue block; ISH, in situ hybridization; PVL, proviral load.

Figure 2.

A representative case of ATLL with follicular helper T‐cell phenotype and focal increased EBV‐infected cells. This 41‐year‐old man (PTCLpt0847) presented with generalized lymphadenopathy without atypical lymphocytes in peripheral blood. (A) Core needle biopsy of a lymph node reveals diffuse infiltration of large‐sized atypical lymphoid cells (H&E, original magnification: ×400). (B–F) Immunohistochemistry shows a follicular helper T‐cell phenotype: (B) CD20(−), (C) CD3(+), (D) CD4(+), CD8(−), TIA1(−), CD10(−), BCL6(+), (E) PD1(+), and (F) ICOS(+). HTLV‐1/2 serology was positive with an S/CO of 158.29. A diagnosis of ATLL was rendered (original magnification: all ×400). (G, H) Notably, focal tight aggregates of EBV‐positive lymphoid cells were identified by EBER in situ hybridization, which would raise concern for EBV‐associated lymphoma (original magnification: ×100 and ×400). (I) Retrospective testing revealed tumor cell positivity for HBZ by in situ hybridization (score: 2+) (original magnification: ×400). Proviral load was 108% (tax: 16,895 copies; RPPH1: 31,297 copies; estimated tumor content: 60%), confirming the diagnosis molecularly.

Figure 3.

A case of nodal T follicular helper cell lymphoma, angioimmunoblastic type, seropositive for HTLV‐1. A 58‐year‐old man presented with generalized lymphadenopathy without peripheral blood leukemic cells or hypercalcemia. Lymph node biopsy revealed nodal T follicular helper cell lymphoma, angioimmunoblastic type. Bone marrow involvement confirmed stage IV disease. He initially received CHOEP chemotherapy (cyclophosphamide, doxorubicin, vincristine, etoposide, and prednisone) and achieved complete remission. However, the disease relapsed and was refractory to multiple salvage regimens. He died 2 years after diagnosis. (A) Low‐power view showing effaced nodal architecture (H&E, original magnification ×20). (B) Residual regressed follicles and prominent interfollicular vascular proliferation are noted (H&E, original magnification ×100). (C) High‐power view demonstrates medium‐sized tumor cells with abundant pale cytoplasm (H&E, original magnification ×400). (D–G) Immunohistochemistry shows tumor cells positive for (D) CD3, (E) CD4, (F) PD‐1, and (G) ICOS, while negative for CD8 and TIA1, consistent with a follicular helper T phenotype. (H) Tumor cells are negative for HBZ (in situ hybridization, original magnification ×400). (I) RNA quality is confirmed by strong PPIB expression (score 3+, in situ hybridization, original magnification ×400).

Figure 4.

A case of NK‐large granular lymphocytic leukemia in HTLV‐1 carrier. A 38‐year‐old man presented with chronic lymphocytosis and large granular lymphocytes on peripheral smear. Hemogram revealed WBC 10.1 × 10³/μl (lymphocytes 22%, atypical lymphocytes 52%), hemoglobin 15.6 g/dl, and platelets 239 × 10³/μl. Flow cytometry demonstrated an NK‐cell phenotype, further supported by polyclonal pattern in T‐cell clonality testing (BIOMED‐2 protocol). A diagnosis of NK‐large granular lymphocytic leukemia (NK‐LGL) was established. The patient remained untreated and disease‐free during 5 years of outpatient follow‐up. PVL in bone marrow and peripheral blood ranged from 0.1% to 10% (estimated tumor percentage: about 50%), supporting HTLV‐1 carrier status without molecular evidence of ATLL. (A) Peripheral blood smear showing large granular lymphocytes (Liu stain, original magnification ×1,000). (B) Flow cytometry of peripheral blood demonstrating atypical lymphoid cells positive for CD45, CD2, and CD7, and negative for CD13, CD33, CD19, surface CD3, CD5, CD56, CD4, CD25, TCRαβ, and TCRγδ. (C) Bone marrow biopsy showing small lymphoid infiltrates (H&E, original magnification: ×400). (D–F) Immunohistochemistry shows the small lymphoid cells are positive for (D) TIA‐1 and (E) CD2, and (F) negative for CD3 (immunostain, original magnification: ×400).

In summary, for PTCL patients with seropositive HTLV‐1/2, conventional clinicopathologic interpretation (original diagnosis) without molecular testing yielded 100% positive predictive value (PPV) (37/37), 93% sensitivity (37/40), and 93% accuracy (41/44) (supplementary material, Table S5).

The mechanism of false seropositivity

The purely serology‐based diagnosis for ATLL in our setting achieved 91% PPV (40/44), which is substantially lower than the theoretical probability of >99% expected from a background HTLV‐1 seroprevalence of 0.034–0.058% due to the four false‐positive cases. Intriguingly, one case (PTCLpt0685) of non‐ATLL PTCL seropositive for HTLV‐1/2 underwent multiple HTLV‐1/2 serologic testing and showed inconsistent results (supplementary material, Table S4). These findings raised concerns regarding potential false seropositivity due to borderline signal‐to‐cutoff (S/CO) values.

To explore this hypothesis, we collected S/CO values from all HTLV‐1/2 Architect assay results in our PTCL cohort (revised diagnosis: ATLL n = 24; non‐ATLL PTCL n = 321). Optical density values from the Murex assay were unavailable because they were not recorded in NTUH‐iMD. Figure 5 shows the S/CO values of each case, grouped by lymphoma category. Pairwise comparisons (supplementary material, Table S6) among the 10 non‐ATLL categories shown in Figure 5 revealed that certain PTCL categories displayed skewed S/CO distributions. Specifically, PTCLs with helper T‐cell phenotypes, such as nTFHL and mycosis fungoides, tended to show higher S/CO values. In contrast, PTCL categories with cytotoxic T‐cell or NK‐cell phenotypes, such as extranodal NK/T‐cell lymphoma, had values predominantly below 0.3.

Figure 5.

Signal‐to‐cutoff ratios in HTLV‐1/2 serologic assays across peripheral T‐cell lymphoma categories and other tumors. This violin plot shows the distribution of S/CO values among various PTCL and HTLV‐1/2 seropositive patients with non‐PTCL tumors. The dashed line indicates the serology assay cutoff (S/CO = 1), and the dotted line marks S/CO = 0.3. Notably, several non‐ATLL PTCL categories with helper T‐cell phenotypes exhibited right‐skewed S/CO distributions, with some values just exceeding the serology threshold. In contrast, PTCLs of cytotoxic T or NK phenotype showed consistently low S/CO values, mostly <0.3. One exception was a case of NK‐large granular lymphocytic leukemia, subsequently confirmed as an HTLV‐1 carrier. Diagnosis abbreviations: ALKpALCL: ALK‐positive anaplastic large cell lymphoma; ALKnALCL: ALK‐negative anaplastic large cell lymphoma; CTCL: MF or SS: cutaneous T‐cell lymphoma, including mycosis fungoides and Sézary syndrome; CTCL: others or LyP: cutaneous T‐cell lymphomas other than MF/SS, including primary cutaneous ALCL and lymphomatoid papulosis; ENKTL: extranodal NK/T‐cell lymphoma; nTFHL: nodal T follicular helper‐cell lymphoma; Other PTCL: NOS or unclassified: PTCL, NOS and unclassified cases; PTCL: peripheral T‐cell lymphoma; pgiPTCL: primary gastrointestinal PTCL; S/CO: signal‐to‐cutoff.

Among the four non‐ATLL PTCL patients with HTLV‐1/2‐seropositivity in our cohort, three belonged to the diagnosis group with skewed S/CO values. The remaining one exhibited an NK phenotype and HTLV‐1 carrier status. These findings further indicated that the borderline seropositivity might be artifactual, representing false HTLV‐1/2 seropositivity due to non‐specific cross‐reactivity.

To delineate the possibility of false seropositivity, we determined the upper bound of S/CO values in non‐ATLL PTCL (mean + 3 standard deviations, back‐transformed from log scale) as 0.974, while the lower bound in ATLL (mean – 3 standard deviations) was 25.1. This marked separation underscores a diagnostic gray zone between S/CO values of 1 and 25. Caution is therefore warranted when interpreting HTLV‐1/2 seropositivity within this range, particularly in PTCLs lacking definitive molecular evidence of ATLL. Notably, all the 4 false seropositive patients displayed S/CO values between 1.28 and 6.08, supporting a non‐ATLL diagnosis with our data by combining the negative molecular testing results.

HTLV‐1/2 serologic testing has become more common over time but remains incomplete

In our cohort, over half of the PTCL patients lacked HTLV‐1/2 serologic data. Figure 6A illustrates the temporal trends in HTLV‐1/2 serology testing and ATLL detection rates among all PTCL cases. Both testing and ATLL diagnosis rates accelerated between 2005 and 2010. However, although serology testing steadily increased after 2010 and reached approximately 80%, the rate of ATLL diagnosis did not increase accordingly. Overall, the annual increase in HTLV serology testing was 16% (odds ratio [OR] per year: 1.16; 95% CI: 1.14–1.19; p < 0.0001 by overall logistic regression).

Figure 6.

Temporal trends in HTLV‐1/2 serologic testing and ATLL diagnosis rates. (A) Line plot showing the annual rate of HTLV‐1/2 serologic testing (blue) and the proportion of PTCL cases diagnosed as ATLL, based on original pathological reports (orange) and molecular confirmation (red). Dashed lines represent LOESS‐smoothed curves to highlight overall trends. The serologic testing rate increased sharply after 2005, accompanied by a rise in ATLL diagnoses. However, since 2010, ATLL diagnosis rates plateaued despite continued increases in testing. (B) PTCL category‐stratified trends of HTLV‐1/2 serologic testing. Categories were grouped based on disease biology and their potential risk of being ATLL mimics. Testing rates increased across most categories, including CTCL (MF and SS). Notably, the testing rate in ALKnALCL, a PTCL type with a higher risk of misclassified ATLL, has declined in recent years, raising concern for missed diagnoses. ALKpALCL: ALK‐positive anaplastic large cell lymphoma; ALKnALCL: ALK‐negative anaplastic large cell lymphoma; ATLL: adult T‐cell leukemia/lymphoma; CTCL: MF or SS: cutaneous T‐cell lymphoma, including mycosis fungoides and Sézary syndrome; CTCL: others or LyP: cutaneous T‐cell lymphomas other than MF/SS, including primary cutaneous ALCL and lymphomatoid papulosis; ENKTL: extranodal NK/T‐cell lymphoma; nTFHL: nodal T follicular helper‐cell lymphoma; Other PTCL: NOS or unclassified: PTCL, NOS and unclassified cases; PTCL: peripheral T‐cell lymphoma; pgiPTCL: primary gastrointestinal PTCL.

We next examined whether serology testing varied by PTCL categories. As shown in Figure 6B, testing rates have generally increased over time across most categories, particularly in cutaneous T‐cell lymphoma of mycosis fungoides or Sézary syndrome (OR per year: 1.22; 95% CI: 1.10–1.37; p = 0.0003 in stratified logistic regression), yet 25% of cases still lacked testing. In addition, the testing rate for ALKnALCL dropped, even though this type carries a possibility of being ATLL, raising a concern for missed diagnoses.

To evaluate whether missing serologic data leads to missed ATLL in real‐world settings, we retrospectively selected 59 out of 513 PTCL cases without HTLV‐1/2 serology for molecular tests. None of these cases was initially diagnosed as ATLL. We intentionally oversampled PTCL types with clinicopathologic features mimicking ATLL such as mycosis fungoides, nTFHL, and PTCL NOS (supplementary material, Table S7). Among them, two cases (3.3%, 2/59) diagnosed in 2024 showed HTLV‐1 PVL >0.1%. One tumor, from a 70‐year‐old woman with unclassified PTCL, exhibited a helper T‐cell phenotype, strong HBZ ISH positivity (score 3+), and a PVL of 156%, consistent with ATLL (supplementary material, Figure S7). The other one, from a 69‐year‐old man with mycosis fungoides, showed a CD3(+)CD8(+)CD4(−) phenotype with CD7 loss and a PVL of 1.6%, though estimated tumor content was also only about 1%. HBZ ISH was inconclusive due to inadequate RNA quality. While molecular evidence was insufficient to confirm ATLL, the findings still raised clinical suspicion in this patient (supplementary material, Figure S8).

Together, these results support the routine use of HTLV‐1/2 serologic testing in newly diagnosed PTCL cases to reduce the risk of missing ATLL diagnoses.

Discussion

In this large (n = 881) PTCL cohort study spanning 25 years in two tertiary referral hospitals in an HTLV‐1 nonendemic setting, we identified three diagnostic pitfalls for ATLL. The most striking finding was that approximately 10% (4/44) of HTLV‐1/2–seropositive PTCLs were not ATLL, largely due to potential false seropositivity associated with helper T‐cell tumor phenotypes. In addition, ATLL closely mimicked other PTCLs, leading to frequent misclassification. Finally, even with serologic testing rates approaching 80%, retrospective molecular screening in cases without prior serologic data revealed ATLL that had initially been overlooked.

In HTLV‐1 nonendemic regions, molecular testing is not routinely employed for ATLL diagnosis. Instead, diagnosis typically relies on pathologic identification of PTCL followed by clinician‐initiated HTLV‐1/2 serologic testing. At our institution, increasing awareness of ATLL as a diagnostic mimic has raised the serology testing rate to 80% in recent years. Nevertheless, cases are still underrecognized. Some may question the need for serologic testing in all PTCLs, especially when a known driver mutation or virus is identified, such as ALK in ALK‐positive anaplastic large cell lymphoma or EBV in extranodal NK/T‐cell lymphoma. However, ATLL can trigger marked proliferation of EBV‐infected bystander cells, as illustrated in Figure 2 and previous studies [33, 34, 35], potentially leading to misclassification as EBV‐associated lymphoma. Our findings support routine HTLV‐1/2 serology testing during PTCL workup, regardless of initial clinical suspicion, to minimize diagnostic bias.

Notably, approximately 10% (4/44) of HTLV‐1/2–seropositive PTCLs in our cohort did not meet molecular criteria for ATLL. While the presence of non‐ATLL PTCL in HTLV‐1 carriers, such as reported cases and case PTCLpt0938 in our study [36], is an obvious possibility, our data suggest that false seropositivity is the predominant mechanism. This phenomenon appears to be associated with tumors of helper T phenotype and possible antibody cross‐reactivity. All molecularly confirmed ATLL cases in our study had S/CO values exceeding 25, consistent with prior reports indicating that S/CO <20 is almost invariably false positive for HTLV‐1 infection [37, 38, 39]. Accordingly, we recommend interpreting HTLV‐1/2 serology results in the context of S/CO values to minimize diagnostic errors.

Although the Architect HTLV‐1/2 assay claims a specificity >99% [40, 41], this does not appear true in the context of helper T‐cell–derived PTCLs. In particular, nTFHL‐AI may exhibit increased rates of false‐positive serologic reactivity. Several case reports have documented co‐occurrence of false‐positive HIV serology with nTFHL‐AI [42, 43, 44]. In addition, one patient was seropositive for HTLV‐1/2 but lacked monoclonal proviral integration by Southern blot [45]. These findings, together with the phenotypic overlap between ATLL and nTFHL [12, 13, 14], underscore the need for cautious interpretation of serologic data. Notably, while nTFHL often exhibits a hyperplastic follicular dendritic network, such a feature is uncommon in ATLL [14, 45]. In ambiguous cases, molecular confirmation remains essential.

Current standard molecular tests for ATLL in FFPE specimens include PVL measurement by tax qPCR and HBZ ISH, both of which have known limitations. As shown in our study and in a previous report [20], HBZ ISH is less sensitive because its target, viral RNA, is often degraded in archival FFPE tissue. However, HBZ ISH better correlates with tumor distribution when tumor burden is low [15, 20]. On the contrary, a major limitation of PVL is its reliance on sufficient tumor content (typically ≥30%), which is often unmet in cases like nTFHL‐AI. To address this, we explored the utility of PVL testing in samples with lower tumor percentages. Notably, a PVL >10% still provided meaningful support for ATLL diagnosis even in low tumor‐content settings (e.g., PTCLpt0452 and PTCLpt1053; see Figure 1). Nevertheless, cases with both low PVL (<10%) and low tumor percentage (<30%) remain diagnostically inconclusive (supplementary material, Figure S8).

This study has several limitations. First, it is a retrospective observational study based on archival FFPE specimens. ATLL cases diagnosed in peripheral blood could not be re‐evaluated. Second, the number of HTLV‐1/2–seropositive PTCLs not meeting molecular criteria for ATLL is relatively small, limiting our ability to explore mechanisms such as HTLV‐2 carriers. Third, the inference of potential false seropositivity is based on borderline S/CO values and skewed S/CO distributions in certain PTCL categories, without confirmatory western blot analysis. Lastly, all serologic assessments were conducted using two commercial platforms, with only one assay providing available index values for further analysis.

In summary, we recommend routine HTLV‐1/2 serologic testing in all newly diagnosed PTCL cases to avoid missing ATLL. However, a substantial proportion (~10%) of HTLV‐1/2 seropositive PTCLs are not ATLL. This discrepancy likely results from potential false seroreactivity in helper T‐cell–derived PTCLs with borderline serology values and from genuine non‐ATLL PTCLs in HTLV‐1 carriers. These findings underscore the importance of careful clinicopathologic correlation and molecular confirmation, rather than sole reliance on serologic status, in the diagnosis of ATLL.

Author contributions statement

C‐TY and W‐CC conceived and designed the study. C‐TY, S‐SC, R‐CW and K‐CC reviewed the diagnoses. C‐TY, L‐YS and KK established the tax‐qPCR protocol. C‐TY, L‐YS, KK, H‐BY, S‐HL, K‐CC, C‐JK and H‐HH performed experiments and analysis. C‐TY and W‐CC wrote the paper. W‐CC and S‐SC critically revised the paper.

Declaration of generative AI and AI‐assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT in order to enhance the English quality of the manuscript. After using this tool, the authors carefully reviewed and edited the content as needed and take full responsibility for the content of the publication.

Supporting information

Supplementary materials and methods

Figure S1. HBZ in situ hybridization in cell line control

Figure S2. In‐house HTLV‐1 tax quantitative polymerase chain reaction

Figure S3. HTLV‐1 proviral load (PVL) reference for cutoff determination estimated by tax quantitative polymerase chain reaction

Figure S4. Schematic of HTLV‐1 proviral load across different contexts

Figure S5. A representative ATLL case misdiagnosed as ALK‐negative anaplastic large cell lymphoma

Figure S6. A case of peripheral T‐cell lymphoma, not otherwise specified, seropositive for HTLV‐1

Figure S7. A case of peripheral T‐cell lymphoma without initial HTLV‐1/2 serology, later confirmed as adult T‐cell leukemia/lymphoma by molecular testing

Figure S8. A case of peripheral T‐cell lymphoma (PTCL) initially diagnosed as mycosis fungoides without HTLV‐1/2 serologic testing, later found to have detectable HTLV‐1 proviral load

Table S1. Antibody list for immunohistochemical study

Table S2. HTLV1 PVL testing in 57 PTCL patients seronegative for HTLV‐1/2

Table S3. HTLV1 PVL in 25 non‐PTCL patients ever seropositive for HTLV‐1/2

Table S4. Fifty‐seven PTCL patients ever seropositive for HTLV‐1/2

Table S5. Confusion matrix of original diagnosis and revised diagnosis after HTLV‐1 molecular testing in HTLV‐1/2–seropositive PTCL patients

Table S6. Pairwise comparison of S/CO values across different lymphoma categories

Table S7. PTCL cases without HTLV‐1/2 serological data selected for HTLV‐1 molecular testing

Data availability statement

For original data, contact the corresponding author.