Pre-treatment [¹⁸F]FDG PET/CT for assessing bone marrow involvement and prognosis in patients with newly diagnosed peripheral T-cell lymphoma

ABSTRACT

Purpose

To explore the value of [¹⁸F]fluorodeoxyglucose (FDG)-positron emission tomography (PET)/computed tomography (CT) in assessing bone marrow involvement (BMI) and prognosis in newly diagnosed peripheral T-cell lymphomas (PTCLs) before treatment.

Methods

This retrospective study included 201 eligible PTCLs who received pre-bone marrow biopsy (BMB) and PET/CT. The status of bone marrow (BM) by PET was assessed using a visual examination and a quantitative index (the maximal standardized uptake value [SUV_max] of BM divided by the SUV_max of the liver [M/L]).

Results

Totally 148 patients had no evidence of BMI by PET or BMB; BMI was detected by both methods in 16 patients. The sensitivity and specificity of PET/CT for patients with confirmed BMI by BMB were 43.2% and 90.2%, respectively (κ =  0.353). In addition, 25 patients assessed by PET/CT staging (having stage I to II disease) had no evidence of BMI detected by both PET/CT and BMB. Image-guided biopsy was also recommended when PET/CT showed a focal FDG uptake outside the iliac crest. Survival analysis revealed that BMB was significant for overall survival (OS) (P = 0.020) while M/L for both progression free survival (P = 0.002) and OS (P < 0.001). In multivariate analysis, M/L (HR 1.825, 95% CI 1.071-3.110, P = 0.027) was an independent prognostic factor for OS. There were no statistical differences at the genetic level about BMI confirmed by PET or BMB.

Conclusion

PET/CT has a complementary role in assessing BMI and an ability to predict prognosis in PTCL patients.

KEYWORDS:

• Positron emission tomography/computerized tomography; bone marrow biopsy; bone marrow involvement; peripheral t-cell lymphomas

1. Introduction

The peripheral T-cell lymphomas (PTCLs), arising from mature post-thymic T cells, are a rare and heterogeneous group of neoplasms characterized by aggressive behaviors. The frequency of bone marrow involvement (BMI) for PTCL is 14.8%−25% [1, 2]. Assessment of BMI before treatment has an important role in accurate staging, survival prognosis [3, 4], and the guidance of therapy [5] in PTCL. However, due to sampling errors, bone marrow biopsy (BMB), the reference standard for the evaluation of bone marrow, has been questioned. In other words, random and inadequate sampling in bone marrow cannot represent the entire situation, and histologic and immunohistochemical findings are sometimes questionable and not unequivocal.

[¹⁸F]fluorodeoxyglucose (FDG)-positron emission tomography (PET)/computed tomography (CT) have shown to be widely useful in disease staging, response assessment and prognostication in lymphomas. For instance, quantitative PET [6, 7] image features such as maximal standardized uptake value (SUV_max), reveal prognostic values for lymphomas. Interim PET [7] is capable of evaluating chemosensitivity and guiding next-step therapy modifications. Additionally, PET-guided biopsy [8] can obtain representative tissues more accurately than a CT guided biopsy. As a consequence, there is a great progress in the area of PET/CT.

With the emergence of PET/CT, the role of BMB has been further questioned. This imaging tool has been widely used to screen patients with stage I/II Hodgkin lymphomas (HLs) [9–12] and substitute BMB in these stage I/II HLs. The results on diffuse large B cell lymphomas (DLBCLs) [10–13] are similar with these findings related to HLs, showing that a PET/CT-assessed BMI can identify advanced-stage disease to some extent. In all, focal or multifocal pattern on PET/CT is highly specific for BMI in both HLs and DLBCLs.

Considering that PTCLs are aggressive lymphomas with high avidity to FDG, several studies have reported using PET/CT for detecting BMI in patients with PTCLs [2, 14–16]. Sensitivity ranges from 18% to 89.3% while specificity ranges from 75.7% to 100%. Those studies included limited number of patients and no subgroup classification was performed. As a result, the utility of PET/CT in the initial evaluation of the BM status has not been fully elucidated in PTCL.

In this study, we explored the value of [¹⁸F]FDG-PET/CT in assessing BMI in patients with PTCL before treatment using a larger sample size. Additionally, the study evaluated the prognostic value of BMI and PET/CT in PTCL and the molecular pattern was also discussed.

2. Materials and methods

2.1. Patients

A total of 201 consecutive newly diagnosed PTCL patients who completed pre-treatment BMB and PET/CT for the first stage between January 2017 and December 2022 patients from one institution were considered eligible for this investigation (Figure 1). These patients were classified correctly according to the World Health Organization (WHO) [17], including peripheral T-cell lymphoma, not otherwise specified (PTCL-NOS), angioimmunoblastic T-cell lymphoma (AITL), nodal PTCL with T-follicular helper (TFH) phenotype (PTCL-TFH), anaplastic large cell lymphoma (ALCL), hepatosplenic T-cell lymphoma (HSTCL), monomorphic epitheliotropic intestinal T-cell lymphoma (MEITCL). Patients with a component of other active lymphomas were not enrolled. Relative laboratory data on the SUV_max of bone marrow and the liver were available.

Figure 1. A flowchart of patient selection, diagnostic performance, and survival analysis. Abbreviations: PTCL, peripheral T-cell lymphoma; PET/CT, positron emission tomography/computed tomography; SUV, standardized uptake value; PTCL-NOS, peripheral T-cell lymphoma, not otherwise specified; AITL, angioimmunoblastic T-cell lymphoma; PTCL-TFH, nodal T-follicular helper (TFH) cell lymphoma; ALCL, anaplastic large cell lymphoma; HSTCL, hepatosplenic T-cell lymphoma; MEITCL, monomorphic epitheliotropic intestinal T-cell lymphoma.

Additional criteria for inclusion for survival analysis were: adults (age >18 years) who underwent more than one year of follow-up or died within one year. Patients who gave up treatment because of financial conditions were excluded. Exclusion criteria for survival analysis were patients having other uncured malignancy cancers or severe infection before enrollment. As for subgroup analysis, several PTCL subgroups commonly encountered and confirmed according to WHO classification, such as AITL, ALCL, PTCL-TFH and so on, were discussed. The overall survival (OS) referred to the time from the first systemic chemotherapy to death for any reason; progression-free survival (PFS) was defined as the time from the initial systemic treatment to disease progression or death. The deadline for follow-up is December 30, 2022. Informed consents were obtained from all the patients or their immediate relatives to analyze their samples, and the study was approved by the ethics committee of our review board.

2.2. PET/CT

All the patients were scanned on a dedicated PET/CT scanner (Biograph version; Siemens, Germany). The patients had been fasting for at least 4-6 h and blood glucose levels were required to be less than 10 mmol /L before [¹⁸F]FDG injection (3.75-5.55 MBq/kg). Scanning was started from the basal skull to mid-thigh after an uptake time of 40-60 min. Low-dose CT without intravenous or oral contrast scans were performed using a sixteen-slice helical CT with a continuous spiral technique (120 KeV; automatic current regulation adjusted to thickness and density of each patient's body; section thickness of 5 mm). PET scans were obtained for 3 min per frame and were reconstructed using iterative algorithm (Siemens).

Experienced nuclear medicine physicians select the right lobe of the liver away from the portal area and outline a region of the interest (ROI) with a diameter of 3 cm to measure the SUV_max of liver. The bone marrow was measured at lumbar 4–5 vertebral bodies when these lower than liver FDG uptake. The SUV_max of bone marrow was measured when the lesion was focally or diffusely elevated. In quantitative analysis, the ratio of M/L (the SUV_max of BM divided by the SUV_max of the liver) was calculated in order to explore the survival outcomes.

There are no prospectively defined criteria for the interpretation of BMI by PET. The most previously reported studies [2, 14, 18] consider focal FDG uptake exceeding the intensity of uptake in the normal liver as indicative of BMI. This criterion is also reported in the published recommendations [9,19]. In our study, BMI based on PET was determined as focal FDG uptake with or without increased diffuse status exceeding liver FDG uptake by two nuclear medicine doctors. Any other reasons which could cause false positive results, such as bone fractures, were excluded.

2.3 . Bone marrow aspiration and biopsy

Specimens from the unilateral posterior iliac crest occasionally were acquired for initial staging without the guidance of PET/CT results. BMI was defined as the existence of neoplastic lymphoid T-cells by aspiration and biopsy technique, with supportive findings on immunohistochemical staining or flow cytometry.

2.4. Next generation sequencing (NGS)

NGS completely covers the exons, fusion-related intron regions, and alternative splicing regions of 103 genes related to T/NK cell lymphoma based on the authoritative TCGA database, NCCN guidelines and 2016 WHO consensus. Detection platform is illumina Hiseq/MiSeqDx/NextSeq. Relevant gene expression profiling was obtained from the medical record system. During the study period, total of 35 cases were tested by NGS, among which 26 patients underwent PET/CT and 33 patients underwent BMB to evaluate BM status.

2.5. Statistical analysis

When BMB was considered gold standards, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were assessed. The consistency between the two diagnostic methods (BMB vs. PET) was compared by Cohen kappa values. Utilizing the McNemar test to determine whether there is a significant difference between visual and quantitative approach. The endpoint events (OS and PFS) were analyzed by Kaplan-Meier methods; a log-rank test was used for comparison. The optimal cut-off point of M/L was determined using the receiver operating curve (ROC) analysis. The potential factors were analyzed by univariate analysis firstly. Factors (P < 0.1 in univariate analysis) were enrolled in multivariate analysis by the Cox proportional hazard regression model. All data were analyzed using SPSS (version 25) and R (version 4.2.2) software. The reported p-values were two-sided and  a P < 0.05 was considered statistically significant.

3. Results

3.1. Baseline characteristics

Detailed information about characteristics before treatment in eligible patients is summarized in Table 1, including 39 PTCL-NOS, 108 AITL, 29 ALCL, 13 PTCL-TFH, 6 MEITCL, and 6 HSTCL. The median age among the entire patients was 60.6 (range from 18 to 90); 24.4% of patients had Eastern Cooperative Oncology Group Performance Status scores (ECOG PS) ≥ 2, and 87.6% were with stage III/IV. Patients choosing treatment approaches for first-line therapy with azacytidine (a demethylated drug), chidamide (a histone deacetylase inhibitor) or lenalidomid (an immunomodulatory agent) were 13.8%, 3.1% and 2.1%, respectively.

Table 1. Baseline characteristics in all PTCL patients.

Characteristic	Total	PET (-) (n = 170)	PET (+) (n = 31)	P-value
Male, no. (%)	114/201	95 (56.2)	19 (59.4)	0.846
Age >60 years, no. (%)	120/201	102 (60.4)	18 (56.3)	0.697
Stage III/IV, no. (%)	176/201	144 (85.2)	32 (100)	0.017
ENSs ≥2, no. (%)	12/201	9 (5.3)	3 (9.4)	0.411
B symptom, no. (%)	80/201	69 (40.8)	11 (34.4)	0.558
ECOG PS ≥2, no. (%)	49/201	43 (25.4)	6 (18.8)	0.506
LDH >250U/L, no. (%)	109/196	91 (55.5)	18 (56.3)	1.000
B2M > 2.53ug/L, no. (%)	100/135	81 (73.0)	19 (79.2)	0.616
WBC (median ± SD, 10^9 /L)	n = 200	n = 168	n = 32	0.510
		8.05 ± 6.18	7.28 ± 5.64
HB (median ± SD, g/L)	n = 200	n = 168	n = 32	0.038
		113.30 ± 25.02	103.13 ± 26.78
PLT (median ± SD, 10^9 /L)	n = 200	n = 168	n = 32	0.299
		196.30 ± 107.18	174.84 ± 104.30
Albumin (median ± SD, g/L)	n = 199	n = 167	n = 32	0.311
		37.78 ± 7.13	36.38 ± 6.92
CRP (median ± SD, mg/L)	n = 189	n = 157	n = 32	0.006
		18.07 ± 26.34	33.12 ± 35.68
High level of M/L, no. (%)	32/201	7 (21.9)	25 (78.1)	<0.001
BMB (+), no. (%)	37/201	21 (12.4)	16 (50.0)	<0.001
EBER positive, no. (%)	32/182	28 (18.5)	4 (12.9)	0.607
IPI, no. (%)	n = 196	n = 164	n = 32	0.195
0–1	41/196	38 (23.2)	3 (9.4)
2	71/196	55 (33.5)	16 (50.0)
3	60/196	50 (30.5)	10 (31.3)
4–5	24/196	21 (12.8)	3 (9.4)
PIT, no. (%)	n = 196	n = 164	n = 32	0.850
0	26/196	22 (13.4)	4 (12.5)
1	66/196	55 (33.5)	11 (34.4)
2	70/196	57 (34.8)	13 (40.6)
3–4	34/196	30 (18.3)	4 (12.5)
First-line treatment, no. (%)	n = 195	n = 164	n = 31	0.626
CHOP or COP	82/195	72 (43.9)	10 (32.3)
CHOPE	51/195	42 (25.6)	9 (29.0)
CHOP + AZA	27/195	21 (12.8)	6 (19.4)
Combined with brentuximab vedotin	14/195	10 (6.1)	4 (12.9)
Combined with lenalidomide	4/195	3 (1.8)	1 (3.2)
Combined with chidamide	6/195	6 (3.7)	0 (0.0)
Others	11/195	10 (6.1)	1 (3.2)
Transplantation, no. (%)	19/201	16 (9.5)	3 (9.4)	0.987
Subgroup, no. (%)	n = 201	n = 169	n = 32	0.942
PTCL-NOS	39/201	33 (19.5)	6 (18.8)
AITL	108/201	90 (53.3)	18 (56.3)
ALCL	29/201	24 (14.2)	5 (15.6)
PTCL-TFH	13/201	11 (6.5)	2 (6.3)
MEITCL	6/201	6 (3.6)	0 (0)
HSTCL	6/201	5 (3.0)	1 (3.1)

Abbreviations: PTCL, peripheral T-cell lymphoma; PET (+), bone marrow involvement confirmed by PET/CT; PET (-), no bone marrow involvement confirmed by PET/CT; ENSs, extranodal sites; ECOG PS, Eastern Cooperative Oncology Group performance status; LDH, lactate dehydrogenase; B2M, Beta2-microglobulin; WBC, white blood cell; SD, standard deviation; HB, hemoglobin; PLT, platelet count; CRP, C-reactive protein; M/L, the highest SUV of BM divided by the mean SUV of the liver; BMB (+), bone marrow infiltration comfiremd by biopsy; EBER, EBV-encoded RNA; IPI, International Prognostic Index; PIT, prognostic index for PTCL; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; COP, cyclophosphamide, vincristine, and prednisone; CHOPE, etoposide plus CHOP; AZA, azacytidine; PTCL-NOS, peripheral T-cell lymphoma, not otherwise specified; AITL, angioimmunoblastic T-cell lymphoma; PTCL-TFH, nodal T-follicular helper (TFH) cell lymphoma; ALCL, anaplastic large cell lymphoma; HSTCL, hepatosplenic T-cell lymphoma; MEITCL, monomorphic epitheliotropic intestinal T-cell lymphoma.

In addition, 18 patients received autologous hematopoietic cell transplantation (auto-HCT) and 1 patient received allogeneic hematopoietic stem cell transplantation (allo-SCT) within front-line settings. There were 2 PTCL-NOS, 5 AITL, 8 ALCL, 1 PTCL-TFH and 3 MEITCL. The median age among these patients was 45 (range from 22 to 65); 3 patients had an international prognostic index (IPI) score of ≥3 points and 1 patient had a prognostic index for PTCL (PIT) score of ≥3 points. For these patients, the 1-year OS and PFS rates were 82.0% ± 12.2% and 70.2% ± 11.5%, respectively.

There were no significant differences in sex, symptom, elevated lactate dehydrogenase (LDH), white blood cell (WBC) level, platelet count (PLT) level, ENSs, ECOG PS, Beta 2-microglobulin (B2M), EBV-encoded RNA (EBER), IPI, PIT, transplantation or first-line treatment between the PET (+) and PET (-) cohorts (+, i.e. BMI confirmed by PET/CT). Patients with PET (+) have higher C-reactive protein (CRP) level (P = 0.006) and lower hemoglobin (HB) level (P = 0.038) than PET (-) patients. Moreover, patients with BMB (+) or PET (+) had significantly difference compared with the respective counterparts in M/L (P = 0.015 and P < 0.001, respectively). Intriguingly, 25 patients assessed by PET/CT staging (having stage I to II disease) had no evidence of BMI detected by BMB. In other words, patients with stage I/II assessed as PET (-) were also BMB (-) in our study.

3.2. Diagnostic performance

On biopsy, 37 patients (18.4%) were positive for BMI. According to visual examination by PET, 169 patients were PET (-) and 32 patients PET (+). Totally 148 patients had no evidence of BMI by PET or BMB. Sixteen patients were confirmed as PET (+)/BMB (-), while 21 were confirmed as PET (-)/BMB (+). When BMB was used as reference standard, the sensitivity, specificity, PPV, and NPV of PET/CT were calculated as 43.2%, 90.2%, 50.0% and 87.6%, respectively. A comparison of the validity of these 2 ways to detect BMI was done (κ =  0.353, Table 2). M/L failed to replace BMB with 56.8% sensitivity and 62.2% specificity (data not shown). Using McNemar's exact test, a significant difference was found in diagnostic accuracy of visual vs. quantitative approach (P < 0.001).

Table 2. Diagnostic performance in all PTCL patients.

PTCL		BMB		kappa
		(+)	(-)
PET/CT	(+)	16	16	κ = 0.353
	(-)	21	148
Sensitivity = 43.2% Specificity = 90.2%

Abbreviations: PTCL, peripheral T-cell lymphoma; PET/CT, positron emission tomography/computed tomography; BMB, bone marrow biopsy.

Interestingly, 2 patients who confirmed BMB (-) but PET (+) received BMB once again under PET/CT guidance and the results of second BMB were positive, indicating that BMB to some extent had sampling error and that location of the biopsy under PET/CT guidance might be helpful for the evaluation of bone marrow. However, only 2 patients received the second BMB, and the left parts failed to had a confirmation.

3.3. Subgroup analysis

To clarify whether the expression of PET/CT in bone marrow was discrepant among different PTCLs, subgroup analysis was performed. MEITL patients (6 cases) had no BMI, while HSTCL patients had the highest rate of BMI (50%), followed by PTCL-TFH patients (30.8%). The presence of BMI confirmed by PET and BMB among different subtypes is shown in Table S1. There were 18 AITL patients with BMI detected by visual PET analysis, among which 8 were confirmed by BMB (sensitivity of 42.1%, specificity of 88.8%). The sensitivity and specificity of PET/CT among PTCL-NOS were 37.5% and 90.3%, respectively. Sensitivity in PTCL-TFH and ALCL was 50.0% and 66.7%, while specificity was 100% and 88.5%. The kappa reflecting the agreement between BMB and visual analysis in AITL, PTCL-NOS, ALCL, and PTCL-TFH was 0.315, 0.307, 0.426 and 0.581, respectively. More details are shown in Table S1. There is not much difference in the application of PET/CT in different PTCL subtypes.

3.4. Prognostic value

A total of 167 (83.1%) patients were included in survival analysis. The 2-year and 5-year OS in all cohorts were 66.7% ± 3.9% and 52.2% ± 5.2%, respectively, with a median follow-up of 27 months. In a visual examination, PET (+) was neither an independent factor for PFS (P = 0.419, Figure 2A) nor for OS (P = 0.672, Figure 2B). BMB (+) was an independent factor for OS (P = 0.020, Figure 2C) but not for PFS (P = 0.711, Figure 2D). In quantitative analysis, the value of M/L (cut-off value, 1.50) was associated with significantly shorter OS and PFS (P < 0.001, Figure 2E; P = 0.002, Figure 2F). The 2-year OS of low-level and high-level M/L patients was 81.2% ± 4.0% and 65.7% ± 5.7%, respectively. The 2-year PFS of low-level and high-level M/L patients was 51.0% ± 5.3% and 27.7% ± 5.9%, respectively.

Figure 2. OS and PFS in patients with PTCL according to PET status (2A and 2B), BMB finding (2C and 2D) and level of M/L (2E and 2F). Abbreviations: OS, overall survival; PFS, progression-free survival; PTCL, peripheral T-cell lymphoma; PET (+), bone marrow involvement confirmed by PET/CT; PET (-), no bone marrow involvement confirmed by PET/CT; BMB (+), bone marrow involvement confirmed by biopsy; BMB (-), no bone marrow involvement confirmed by biopsy; M/L, the highest SUV of BM divided by the mean SUV of the liver.

In univariate analysis, the following variables were associated with OS: age (P = 0.042), symptom (P = 0.060), BMI (P = 0.023), ECOG PS (P < 0.001), LDH (P = 0.027), HB level (P < 0.001), PLT level (P = 0.005), albumin level (P < 0.001), M/L (P < 0.001) and transplantation (P = 0.054). In multivariate analysis, ECOG PS (hazard ratio [HR] 3.501, 95% confidence interval [CI] 2.083-5.884, P < 0.001), albumin (HR 0.955, 95% CI 0.915-0.996, P = 0.033) and M/L (HR 1.825, 95% CI 1.071-3.110, P = 0.027) were independent prognostic factors for predicting OS in patients with PTCL. The results about OS and PFS are shown at Table 3.

Table 3. Univariate and multivariate statistical analysis of prognostic factors for OS and PFS in PTCL Patients.

Prognostic factors	OS			PFS
	Univariate analysis	Multivariate analysis		Univariate analysis	Multivariate analysis
		HR (95%CI)	P-value		HR (95%CI)	P-value
Age	0.042			0.018
Symptom	0.060			0.302
BMB (+)	0.023			0.712
LDH	0.027			0.004
ECOG PS	<0.001	3.501 (2.083-5.884)	<0.001	<0.001	2.340 (1.543-3.547)	<0.001
HB	<0.001			<0.001
PLT	0.005			0.002
Albumin	<0.001	0.955 (0.915-0.996)	0.033	0.002	0.963 (0.934-30.992)	0.015
M/L	<0.001	1.825 (1.071-3.110)	0.027	0.002
Transplantation	0.054			0.073

Notes: *Some parts are blank because the data is not statistically significant in multivariate analysis.

Abbreviations: OS, overall survival; PFS, progression-free survival; HR, hazard ratio; CI, confidence interval; BMB (+), bone marrow involvement confirmed by biopsy; LDH, lactate dehydrogenase; ECOG PS, Eastern Cooperative Oncology Group performance status; HB, hemoglobin; PLT, platelet count; M/L, the highest SUV of BM divided by the mean SUV of the liver.

3.5. Molecular profile

We analyzed the overall gene expression signatures in 35 cases. Totally, 45.7% patients demonstrated TET2 mutations, 28.5% patients RHOA^G17V mutations and 22.8% patients DNMT3A mutations; TP53 (14.2%) and KMT2D (11.4%) mutations followed next. In subgroup analysis, common gene mutations in AITL include TET2 (68.7%), RHOA^G17V (43.7%), DNMT3A (37.5%), IDH2^R172 (12.5%), NOTCH1 (12.5%) and ATM (12.5%). JAK1, JAK3, SETD2, PRDM1 and CREBBP alterations were not tested in AITL. One case of PTCL-TFH had TET2, KRAS, and PTEN mutations, and the other case of PTCL-TFH only had TP53 mutations. Among 11 PTCL-NOS patients, common gene mutations included TP53 (27.2%), DNMT3A (18.1%), TET2 (36.3%) and RHOA^G17V (27.2%). ALK mutations were only detected in 3 ALK-positive ALCL patients. PRDM1 and TP63 mutations were found in ALK-negative ALCL patient. The frequency of gene mutations was the highest in HPTCL and MEITCL compared with others in our study. HPTCL and MEITCL had distinct molecular profile, like JAK1, JAK3, SETD2 and CREBBP, which occurred at much lower frequency in other subtypes. The upregulation of MYC only happened in MEITCL.

By comparing two groups of patients (n=33) with and without BMI confirmed by BMB, we found no statistical differences at all genetic level. Furthermore, we analyzed 26 patients who underwent PET/CT to evaluate BM status and observed that patients carrying genetic mutations had no significant correlation with focal FDG uptakes. Incredibly, patients without TET2 tended to have focal FDG uptakes (P = 0.051). Among the patients who underwent NGS, 5 patients had inconsistent results of BMI by BMB and PET/CT (Table S2).

4. Discussion

The present study explored the utility of PET/CT in predicting BMI. PET/CT showed a considerable specificity (90.2%) for identifying BMI but low sensitivity (43.2%) in all patients. Kappa was 0.353 for all cohorts, exhibiting poor consistency between PET/CT and BMB for the assessment of BMI. However, 25 patients with stage I to II disease by PET had no evidence of BMI. We recommended that, for patients with early stage (I to II), BMB could probably be absent for these patients, similar to HL. For patients with advanced stage (III to IV), PET/CT failed to substitute BMB due to the low sensitivity. Considering positive results by directed biopsy in accordance with PET/CT in 2 cases, the image-guided biopsy was recommended when PET/CT showed a focal or multifocal FDG uptake outside the iliac crest, which BMB could miss. For PTCL patients who are willing to transplant within front-line settings, classical BMB is unavoidable.

BMI by biopsy was an independent prognostic factor for OS (P = 0.020) while BM finding by PET failed to predict survival in PTCL. The reason probably was false-positive and false-negative rates from PET we will talk later. However, the value of M/L was associated with significantly shorter OS and PFS (P < 0.001, P = 0.002, respectively). With a median follow-up of 27 months, the 2-year OS of low-level and high-level M/L patients was 81.2% ± 4.0% and 65.7% ± 5.7%, respectively. The 2-year PFS of low-level and high-level M/L patients was 51.0% ± 5.3% and 27.7% ± 5.9%, respectively. In a multivariate analysis, M/L remained an independent predictor of inferior OS (HR 1.825, 95% CI 1.071-3.110, P = 0.027). Taken together, our results showed that PET/CT could not completely substitute BMB but had a complementary role in assessing BMI and an ability to predict prognosis in PTCL patients.

Previous retrospective studies have reported partly similar results about prognosis. Abe et al. [14] observed that patients with BMB (+) or PET (+) showed significantly shorter OS and PFS than the counterparts in the setting of inconsistent definition of BMI (Table 4). Koh et al. [16] found PET (+) was not significant for predicting PFS but for OS. Interestingly, M/L (cut-off value, 0.68) was significant for both PFS and OS (P = 0.005 and P < 0.001), resonating well with our findings although the cut off for M/L was inconsistent. It is worthy to note that Koh et al included 63 PTCL together with 46 NK/T-cell lymphoma (NKTCL). Sundaram et al. [2] evaluated 60 patients and demonstrated that BMB (+) was associated with lower PFS (P = 0.038) and OS (P = 0.003) while PET (+) was associated only with OS (P = 0.02). Prognostic differences may stem from inconsistent definitions, inconsistent study populations or sample sizes. It is speculated that quantitative measurement is superior to biopsy and visual analysis because M/L is a continuous parameter and optimal cutoff values can be selected. These aforementioned results should be confirmed in prospective trails with uniform definition and a larger size.

Table 4. Relative retrospective studies about the diagnostic value of PET/CT for the identification of the involvement of bone marrow.

Investigation (year)	Total number	No. of patients BMB+	No. of patients PET+	Sensitivity	Specificity	PPV	NPV	complement
Sundaram et al. (2020) [2]	53 PTCL + 7 NKTCL	15	8	53.3% (95% CI 29.4, 76.1)	100% (95% CI 94.6, 100)	100% (95% CI 73.8, 100)	86.5% (95% CI 75.4, 93.8)
Abe et al. (2019) [14]	83 PTCL	28	25	89.3%	100%	100%	94.8%	BMB (+): BMB specimens positive for PTCL or the lack of an initially abnormal bone marrow uptake on follow-up PET/CT indicative of a successful treatment response.
Koh et al. (2019) [16]	63 PTCL + 46 NKTCL	39	41	61.5%	75.7%	N/A	N/A	PET (+): moderately increased diffuse uptake that is higher than the liver uptake; markedly increased diffuse uptake and multifocal uptakes in the bone marrow
Pham et al. (2017) [35]	16 PTCL	16	N/A	50%	N/A	N/A	N/A
El-Galaly et al. (2015) [15]	125 PTCL	17	13	18% (95% CI 4–43)	90% (95% CI 82–95)	N/A	N/A

Abbreviations: PTCL: peripheral T-cell lymphoma; NKTCL: natural killer/T cell lymphoma; PET/CT: positron emission tomography/computed tomography; BMB+: bone marrow involvement by bone marrow biopsy; PET+: bone marrow involvement by PET/CT; PPV: positive predictive value; NPV: negative predictive value; BMB: bone marrow biopsy; CI: confidence interval; N/A: not applicable.

The diagnostic value of PET/CT for the identification of BMI has been reported by other studies (Table 4). Abe et al. [14] included 83 PTCLs without NKTCL and found PET/CT performed better than BMB based on sensitivity and NPV (89.3% vs. 60.7% and 94.8% vs. 83.3%, respectively) with similar specificity and PPV. However, the definition of BMI in this article that was BMB (+) or the disappearance of initial focal uptake on follow-up PET/CT monitoring was different from general ones (positive only by BMB). Our study was in accordance with other findings. Sundaram et al. [2] found PET/CT had high specificity but relatively low sensitivity. El-Galaly et al. [15] collected 114 cases of PTCL and noted that 17 patients (15%) were BMB (+), with 3 having PET/CT (+), 2 having diffuse uptake higher than liver and 12 having no FDG-uptake. The sensitivity and specificity of PET/CT for the evaluation of BMI were 18% and 90%, respectively. Koh et al. [16] enrolled 63 PTCL and 46 NKTCL patients, calculated the kappa value and concluded that the assessment of BMI by PET/CT in visual as well as quantitative analyses failed to replace BMB (kappa was 0.368 and 0.359, respectively).

There is not much difference in the application of PET/CT in various subtypes (Table S1). Some previous studies tried to investigate the reason behind it. Patients with AITL presented with polymorphous infiltration [20, 21] and inflammatory cells [22], enhancing metabolic phenotypes on PET/CT and leading to overestimation of the involvement of bone marrow [2]. PTCL-NOS was biologically and clinically very heterogeneous [21], so it had a different performance on PET/CT. ALK-negative ALCL had a relatively higher expression of miR-146a [23] than ALK-positive ALCL, which increased glucose consumption and caused differential metabolism, as shown by PET/CT. Some therapies, such as colony-stimulating factor, which were ignored by our investigators, influenced the reactive bone marrow process on PET/CT and eventually our results. We also found that the frequency of BMI in HPTCL subtypes is very high, even reaching 50%. If a patient is clearly diagnosed as this subtype, we may need to pay closer attention to whether there is BMI and use PET/CT to assist in accurate evaluation.

The genetic basis of PTCL with BMI is rarely discussed. By comparing two groups of patients with and without BMI confirmed by PET or BMB, we found no statistical differences at the genetic level. Theoretically, BMI was associated with genes about cell proliferation or immune escape, like DLBCL [24]. These insignificant results may be due to small sample sizes or to genetic mutations that are more strongly associated with something else, for instance, T cell lymphoma’s subtypes. TET2, DNMT3A, RHOA^G17V and IDH2^R172 mutations had a correlation with the development of AITL as previous studies reported [25, 26]. ALK was muted only in 3 ALK-positive ALCL patients in the present study. HPTCL and MEITCL had unique genetic alterations, which resonated well with other findings [27–29]. Based on the fact patients without TET2 mutations usually have focal FDG uptakes (P = 0.051), we speculated boldly that TET2 mutations may affect bone marrow imaging programs. Considering the higher frequency of TET2 mutations in PTCL, is the above phenomenon a chance event or is it as inferred? Further exploration and verification of large samples are needed.

Diffusely increased FDG uptake of bone marrow without clear definition is considered as feedback of blood picture or reactive myelopoiesis [30, 31]. This theory was in keeping with our observation. In this study, PET (+) patients tend to have lower HB level (p = 0.038) compared with PET (-) patients. The use of drugs before the implementation of PET/CT, such as hematopoiesis-stimulating drugs [10], steroids, and the beta-blocker propranolol [32], probably influenced the results to a certain extent. The picture of diffusely even no increased uptake could be seen in patients with and without BMI. We hypothesized that aggressive lymphoid cells, which overexpress cyclin D1, such as mantle cell lymphoma (MCL) [33] and multiple myeloma (MM) [34], exhibited low metabolic activity on PET/CT and acted like normal cells, causing false negative PET/CT pattern in PTCLs.

There are some limitations in the present study. First, this retrospective and single-center study may lead to bias. Moreover, limited sample size and heterogeneous treatment modalities partly influenced the statistical outcomes. Therefore, an additional validation cohort was necessary. In addition, only 2 patients received directed BMB with focal lesions on PET/CT. Although these image-guided biopsies illustrated the reason for the false negative by BMB to a certain extent, the sample size was too small, causing the strength of the obtained results. Lastly, the assumption about subtype analysis and false negatives was not proved, and follow-up should be extended better to explain the prognostic value of PET/CT. Also, we did not discuss the role of PET/CT in assessing the clearance of bone marrow after treatment because of unacceptable false-positive and false-negative rates [10, 35, 36] after treatment.

To the best of our knowledge, this is the largest study on patients with PTCL, explored the differences in distinct subgroups and exhibited the molecular profile about PTCL. As for PTCL patients in early-stage disease, the very low incidence of BMI based on 25 cases allowed sparing BMB if PET/CT was performed. PET/CT was insufficiently sensitive for BMI but had an auxiliary role in identifying BM lesions, and directed biopsy was recommended once focal uptake on PET/CT was found. The value of M/L did provide prognostic information. There were no statistical differences at the genetic level about BMI confirmed by PET or BMB. Molecular characteristics were likely to be associated with lymphoma subtypes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Medical Science and Technology Project of Zhejiang Province [Grant Number Project No. 2021KY143]; Natural Science Foundation of Zhejiang Province [Grant Number LY22H080004].