Patients’ characteristics and survival
The clinical characteristics are shown in Table 1. All patients had primary disease in the UADT sites. The male-to-female ratio was 2.7:1. The median age was 46 years (range, 6-85). The majority of patients originated from the nasal cavity (80.1%), presented with early-stage disease (94.3%), and had PTI (71.6%) and good performance status (ECOG score 0-1; 94.3%). With a median follow-up time of 50 months for surviving patients, the 5-year OS and PFS rates were 74.9% and 64.9% for all patients, and 76.8% and 67.9% for early-stage patients.
Table 1
Patient characteristics stratified by the MRI radiomics signature in patients with all-stages and early-stage ENKTCL.
|
|
MRI radiomics signature
|
|
Total
|
Type I
|
Type II
|
|
Characteristic
|
No. (%)
|
No. (%)
|
No. (%)
|
P
|
All Stages (n = 176)
|
176
|
65
|
111
|
|
Gender, male
|
128 (72.7)
|
46 (70.8)
|
82 (73.9)
|
0.655
|
Age, > 60 years
|
26 (14.8)
|
10 (15.4)
|
16 (14.4)
|
0.861
|
B symptoms
|
82 (46.6)
|
22 (33.8)
|
60 (54.1)
|
0.009
|
ECOG score ≥2
|
10 (5.7)
|
1 (1.5)
|
9 (8.1)
|
0.094
|
Stage I–II
|
166 (94.3)
|
63 (96.9)
|
103 (92.8)
|
<0.001
|
PTI
|
126 (71.6)
|
23 (35.4)
|
103 (92.8)
|
<0.001
|
Elevated LDH
|
46 (26.1)
|
9 (13.8)
|
37 (33.3)
|
0.005
|
Primary site, nasal cavity
|
141 (80.1)
|
58 (89.2)
|
83 (74.8)
|
0.020
|
Stage I–II (n = 166)
|
166
|
63
|
103
|
|
Gender, male
|
121 (72.9)
|
45 (71.4)
|
76 (73.8)
|
0.740
|
Age, > 60 years
|
25 (15.1)
|
10 (15.9)
|
15 (14.6)
|
0.819
|
B symptoms
|
77 (46.4)
|
21 (33.3)
|
56 (54.4)
|
0.008
|
ECOG score ≥ 2
|
10 (6.0)
|
1 (1.6)
|
9 (8.7)
|
0.091
|
Stage II
|
65 (39.2)
|
3 (4.8)
|
62 (60.2)
|
<0.001
|
Elevated LDH
|
43 (25.9)
|
9 (14.3)
|
34 (33.0)
|
0.008
|
PTI
|
116 (69.9)
|
21 (33.3)
|
95 (92.2)
|
<0.001
|
Primary site, nasal cavity
|
134 (80.7)
|
56 (88.9)
|
78 (75.7)
|
0.037
|
Abbreviations: ENKTCL extranodal nasal-type NK/T-cell lymphoma, MRI magnetic resonance imaging, ECOG Eastern Cooperative Oncology Group, LDH lactate dehydrogenase, PTI primary tumor invasion
Risk-stratified subgroups by radiomics signatures
A total of 777 features with statistical significance (P < 0.05) were preliminarily selected from the 3144 radiomics features in the whole group. A radiomics signature was further constructed based on the spectrum cluster analysis of unsupervised learning. The heat map of cluster analysis showed the final classification results (Fig. 3A, Supplementary Table 1). In type I group, 1-528 and 641-705 are dark green, with low eigenvalues, and 529-640 and 706-777 are red, yellow, and light green, with high eigenvalues. In type II group, 1-528 and 641-705 are red, yellow, and light green, with high eigenvalues, and 529-640 and 706-777 are dark green, with low eigenvalues.
Patients with type II had significantly higher adverse prognostic factors, including B symptoms, ECOG score ≥2, elevated LDH, advanced-stage disease and PTI, than those with type I (Table 1). The 5-year OS was 87.2% in type I, significantly higher than 67.3% in type II (Hazard Ratio [HR] 3.12, 95% CI 1.45-6.72; P = 0.002; Fig. 3B). Similar results between type I and II were observed in early-stage patients (88.8% vs. 69.2%; HR 3.17, 95% CI 1.39-7.22; P = 0.003; Fig. 3C).
Validation of risk-stratified groups based on MRI radiomics classifier
The AUC and Harrell's C index of MRI radiomics classifier for predicting 5-year OS were 0.664 and 0.623 (95% CI: 0.566-0.681) for all patients, and 0.667 and 0.622 (95% CI: 0.559-0.686) for early-stage patients, respectively. The 5-year OS predicted by MRI radiomics classifier was 87.8% for type I and 66.7% for type II in the whole group, and 88.9% for type I and 69.0% for type II in the early-stage group. The predicted OS was comparable to the observed OS (Table 2). The calibration curve for the probability of 5-year OS showed good correlation between the actual observation and the radiomics signature prediction in the whole group and early-stage patients.
Table 2
The observed and predicted 5-year OS of the MRI-based radiomics signature and NRI-M in patients with all-stages and early-stage ENKTCL.
Risk group
|
No. (%)
|
Observed
5-year OS (%)
|
P
|
|
Predicted
5-year OS (%)
|
All Stages (n = 176)
|
|
|
|
|
|
MRI radiomics signature
|
|
|
0.002
|
|
|
Type I (Favorable)
|
65 (36.9)
|
87.2
|
|
|
87.8
|
Type II (Unfavorable)
|
111 (63.1)
|
67.3
|
|
|
66.7
|
NRI-M
|
|
|
<0.001
|
|
|
Low-risk (0-1)
|
53 (30.1)
|
90.5
|
|
|
90.9
|
Intermediate-low-risk (2)
|
37 (21.0)
|
80.8
|
|
|
83.5
|
Intermediate-high-risk (3)
|
47 (26.7)
|
69.6
|
|
|
70.9
|
High-risk (4)
|
28 (15.9)
|
63.4
|
|
|
52.0
|
Very high-risk (≥5)
|
11 (6.3)
|
22.7
|
|
|
28.8
|
Stage I–II (n = 166)
|
|
|
|
|
|
MRI radiomics signature
|
|
|
0.004
|
|
|
Type I (Favorable)
|
63 (38.0)
|
88.8
|
|
|
88.9
|
Type II (Unfavorable)
|
103 (62.0)
|
69.2
|
|
|
69.0
|
NRI-M
|
|
|
<0.001
|
|
|
Low-risk (0-1)
|
53 (31.9)
|
90.5
|
|
|
90.4
|
Intermediate-low-risk (2)
|
37 (22.3)
|
80.8
|
|
|
82.9
|
Intermediate-high-risk (3)
|
45 (27.1)
|
71.5
|
|
|
70.8
|
High-risk (≥4)
|
31 (18.7)
|
54.8
|
|
|
52.8
|
Abbreviations: OS overall survival, MRI magnetic resonance imaging, NRI nomogram-revised risk index, NRI-M MRI radiomics-based NRI, ENKTCL extranodal nasal-type NK/T-cell lymphoma
Construction and validation of NRI-M model
The NRI-M model integrates MRI radiomics classifier into the NRI-defined clinical prognostic factors [19], and assigns one point-each to the type II MRI radiomics. Based on the NRI-M, the 5-year OS rates in the entire cohort were 90.5% for low-risk, 80.8% for intermediate-low-risk, 69.6% for intermediate-high-risk, 63.4% for high-risk, and 22.7% for very high-risk groups (P <0.001, Fig. 4A). The corresponding OS rates in early-stage patients were 90.5% for low-risk, 80.8% for intermediate-low-risk, 71.5% for intermediate-high-risk, and 54.8% for high-risk group, respectively (P <0.001, Fig. 4B).
The 5-year OS rates predicted by the NRI-M for the whole group in the low-, intermediate-low-, intermediate-high-, high-, and very high-risk groups were 90.9%, 83.5%, 70.9%, 52.0%, and 28.8%, respectively. The predicted 5-year OS for early-stage patients in the low-, intermediate-low-, intermediate-high-, and high-risk groups was 90.4%, 82.9%, 70.8%, and 52.8%, respectively (Table 2). The calibration plot for the probability of 5-year OS showed a good correlation between the actual observed outcome and the prediction by the NRI-M for all-stages (Fig. 4C) and early-stage patients (Fig. 4D).
Evaluation of NRI-M Model
The NRI-M model was evaluated by predictive accuracy, discrimination and predictive error. Compared with the NRI and the MRI radiomics classifier, the NRI-M model had better levels of accuracy for predicting OS. The AUC of NRI-M for predicting 5-year OS (0.748, 95%CI 0.654-0.842) for all patients was higher than that of the NRI (0.736, 95%CI 0.641-0.832) or the MRI radiomics classifier (0.664, 95%CI 0.575-0.752; P = 0.013, Fig. 5A). Similarly, for early-stage patients, the AUC of NRI-M (0.717, 95%CI 0.616-0.819) for predicting 5-year OS was higher than that of the NRI (0.699, 95%CI 0.597-0.802) or radiomics classifier (0.667, 95%CI 0.577-0.758; Fig. 5B). Moreover, the tAUC of the NRI-M model between 12 and 84 months was consistently higher than the MRI radiomics classifier and the NRI model in the whole group (Fig. 5C) and in the early-stage patients (Fig. 5D). Moreover, the Harrell's C-index of the NRI-M of the whole group (0.740, 95%CI: 0.667-0.814) and early-stage patients (0.729, 95%CI: 0.649-0.810) was higher than that of the NRI (0.737, 95%CI: 0.664-0.810; 0.727, 95% CI: 0.648-0.807) and MRI radiomics classifier (0.623, 95%CI: 0.566-0.681; 0.622, 95%CI: 0.559-0.686).
The performance of NRI-M model was assessed by calculating prediction error over time in the entire and early-stage patients. In the whole group, the NRI-M IBS (0.142) of the 5-year OS was lower than that of the NRI (0.144) and MRI radiomics classifier (0.156). Similarly, in the early-stage patients, the IBS (0.140) of the NRI-M was also lower than that of the NRI (0.142) and the MRI radiomics classifier (0.146). The corresponding prediction error curves of all models were shown in the whole group (Fig. 5E) and early-stage patients (Fig. 5F). The results suggest that the NRI-M have better discrimination and accuracy for all-stages and early-stage patients.