3.1 Patient characteristics

A total of 86 patients were included in this study, including 45 men and 41 women with a mean age (57.8 ± 12.2 years). The initial chemotherapy regimens for all patients were R-CHOP or R-CHOP-like. Table 1 presents the clinical characteristics of the patients. The median follow-up was 36.4 months (95% CI 15.825–20.515), with 34 (39.4%) patients experiencing progression or recurrence and the death of 9 (10.4%) patients during follow-up. The median PFS was 22.5 months, and the 1- and 3-year PFS rates were 68.4% and 31.6%, respectively.

Table 1 Clinical characteristics of 86 patients with diffuse large B-cell lymphoma3.2 Univariate and multivariate regression analyses of patient parameters

Univariate Cox regression analysis showed that the International Prognostic Index (IPI), Ann Arbor staging, and DS score were clinical risk factors for PFS in patients, with hazard ratios (HR) of 3.403 (95% confidence interval [CI] 1.290–8.974, P = 0.013), 2.915 (95% CI 1.075–7.907, P = 0.036), and 4.566 (95% CI 1.780–11.710, P = 0.002), respectively. Furthermore, ΔSUVmax%, ΔMTV%, ΔTLG%, and ΔDmax% were metabolic and diffusion risk factors affecting the PFS of patients with DLBCL, with HRs of 6.213 (95% CI 2.368–16.300, P < 0.001), 12.516 (95% CI 4.169–40.645, P < 0.001), 12.516 (95% CI 4.169–40.645, P < 0.001), and 13.430 (95% CI 4.925–36.622, P < 0.001), respectively.

As ΔMTV% and ΔTLG% were highly correlated, ΔMTV% and ΔTLG% were performed separately in Cox multifactorial regression analyses in order to avoid the influence of multicollinearity on the study. Clinical parameters (IPI, Ann Arbor staging, and DS), metabolic parameters (ΔSUVmax%, ΔMTV%, or ΔTLG%), and diffusion parameters (ΔDmax%) that were statistically significant in the univariate regression analyses were included in the multifactorial Cox regression analyses. The results indicated ΔMTV% and ΔDmax% as independent predictors of PFS, with an HR of 10.727 (95% CI 1.928–56.672, P = 0.007) and 7.178 (95% CI 1.514–34.041, P = 0.013), respectively (Table 2).

Table 2 Cox analysis results of PFS in 86 Patients with DLBCL3.3 PFS survival curves of patients grouped according to different parameters

The Kaplan–Meier survival curves of patients who underwent 18F-FDG PET/CT after mid-treatment prediction of PFS for EN-DLBCL using each assessment metric grouped according to different clinical profiles are shown in Fig. 1. The median PFS time for patients with pre-treatment IPI < 4 was 37 months, whereas that for patients with IPI ≥ 4 was 16 months. The median PFS for patients with Ann Arbor stages I and II was 39 months, whereas that for patients with Ann Arbor stages III and IV was 17 months. The median PFS time was 34 months for patients with DS scores of 1–3, and 12 months for patients with DS scores of 4–5. We used the Youden index method in the ROC curve analysis to determine the best cut-off value of the dichotomous parameters. The results showed that the optimal cutoff values of ΔMTV% and ΔDmax% for PFS were 99.10% (AUC = 0.849, P < 0.001) and 96.47% (AUC = 0.786, P = 0.001), respectively. The Chinese Society of Nuclear Medicine's Clinical Guidelines for Clinical Application of 18F-FDG PET/CT Imaging in Lymphoma (2021 edition) recommends 66% as the cut-off value of ΔSUVmax%, and this was included in the analysis. The median PFS duration for patients with ΔSUVmax% < 66% and ΔSUVmax% ≥ 66% was 9 and 34 months, respectively. The median PFS time for patients with ΔMTVmax% < 99.10% and ΔMTVmax% ≥ 99.10% was 12 and 37 months, respectively. Patients with ΔDmax% < 99.47% had a median PFS of 8.5 months, whereas those with ΔDmax% ≥ 99.47% had a median PFS time of 35 months.

Fig. 1

PFS survival curves of patients grouped by different clinical data. A IPI score curve; B Ann Arbor staging curve; C DS score curve; D ΔSUVmax% curve; E ΔMTV% curve; F ΔDmax% curve

3.4 Predictive modeling of PFS

We used the Youden index method in the ROC curve analysis to determine the best cut-off value of the dichotomous parameters. The results showed that the optimal cutoff values of ΔDmax% and ΔMTV% for PFS were 96.47% and 99.10%. Kaplan–Meier survival curve analysis showed that PFS was better in the group with ΔDmax% ≥ 96.47% than in the group with ΔDmax% < 96.47% (Fig. 2); PFS was better in the group with ΔMTV% ≥ 99.10% than in the group with ΔMTV% < 99.10% (Fig. 3), suggesting that patients with low ΔDmax% and ΔMTV% after treatment had a poorer prognosis and were more likely to recur or progress. A new model was established by combining the parameters of ΔDmax% and ΔMTV%.

Fig. 2

Kaplan–Meier survival analysis of PFS according to ΔDmax%

Fig. 3

Kaplan–Meier survival analysis of PFS according to ΔMTV%

The patients were categorized into three groups as follows: the high-risk group was ΔMTV% ≥ 99.10% + ΔDmax% ≥ 96.47%, the low-risk group was ΔMTV% < 99.10% + ΔDmax% < 96.47%, and the rest of the combinations were intermediate-risk groups.Kaplan–Meier survival curve showed that PFS was statistically different between all three groups (P < 0.001), PFS in the low-risk group was significantly higher than in the medium- and high-risk groups (Fig. 4). The results of the ROC curves of ΔMTV%, ΔDmax%, and ΔMTV% + ΔDmax% for predicting the PFS of patients showed that the AUCs, sensitivity, and specificity of the high-, intermediate- and low-risk groups were 0.706, 51.4%, 81.4%; 0.687, 78.1%, and 57.1%; and 0.778, 57.3%, and 84.8%, respectively. The joint model improved the predictive performance compared with the individual parameters (Fig. 5).

Fig. 4

Kaplan–Meier survival analysis of PFS according to the Combination Model

Fig. 5

ROC curves of ΔMTV%and ΔDmax% and the Combination Model for prediction of PFS

Figure 6 Patient 1,baseline PET maximum intensity projection (MIP) map shows a right perirenal hypermetabolic solid mass and multiple right perirenal hypermetabolic lymph nodes (arrows).Mid-treatment MIP map showing an increase in the size of the right renal lesion (arrows) compared with the previous lesion and a decrease in the size of the surrounding lymph nodes.The patient's ΔMTV% was 27.84%, which was < 99.10% cut-off value,and ΔDmax% was18.43%, which was < 96.47% cut-off value.The PFS was 3 months. Patient 2, The baseline MIP map shows multiple foci of metabolic increase in the abdomen (circle). The mid-treatment MIP map shows no additional positive uptake foci atthe site of the original lesion (circle). The patient's ΔMTV% was 100%, which was greater than the 99.10% cut-off value, and ΔDmax% was 100%, which was greater than the 96.47% cut-off value. The PFS was 39 months

Fig. 6

Representative 18F-FDG PET/CT images of two Typical Cases