In total, 697 ENKTL patients (492 males and 205 females), with a median age of 43 years (range 10–82 years), met the inclusion criteria. The median ages for blood type O, A, B, and AB groups were 44 years (range 10–80 years), 43 years (range 14–82 years), 42 years (range 16–76 years), and 49 years (range 21–76 years), respectively. The clinical characteristics of the 697 patients are listed in Table 1.
Most of the patients (680, 97.6%) displayed a favorable performance status (ECOG PS 0–1). Three hundred and twenty-four patients (46.5%) presented with B symptoms. Elevated LDH levels were observed in 194 (27.8%) patients. Sixty-five patients (9.3%) had a mass ≥5 cm, and only 15 (2.2%) displayed bone marrow involvement. One hundred and seventy-one patients (24.5%) displayed regional lymph node involvement, and 77 (11.0%) displayed at least two extranodal involvement sites. Most of the patients (619, 88.8%) had localized disease (stage I/II). According to the IPI, 597 cases (85.7%) were classified as low-risk disease (IPI = 0–1), and 100 (14.3%) were categorized as high-risk disease (IPI = 2–5). The number of patients with KPI = 0–1 was higher than those with KPI = 2–4 [478 (68.6%) vs. 219 (31.4%)]. The baseline CRP levels were available in 228 patients (range 0.16–154.92 mg/L; median value 7.00 mg/L), and the baseline plasma EBV-DNA data were available in 171 patients (range 0–48,500,000 copies/mL; median value 1530 copies/mL).
The distribution of ABO blood types in the ENKTL group was blood type O in 255 (36.6%) patients, blood type A in 195 (28.0%) patients, blood type B in 188 (27.0%) patients, and blood type AB in 59 (8.5%) patients. According to the previous matching method, 697 patients with nonmalignant disease were randomly selected as the control group. The distribution of ABO blood types in the control group was blood type O in 279 (40.0%) patients, blood type A in 175 (25.1%) patients, blood type B in 162 (23.2%) patients, and blood type AB in 81 (11.6%) patients. There was no significant difference in blood type distribution between the ENKTL and control groups (P = 0.056). ABO blood type was not associated with patient age, gender, ECOG PS, B symptoms, LDH levels, tumor size, number of extranodal sites, regional lymph node involvement, baseline EBV-DNA copies, Ann Arbor stage, IPI score, or KPI score (all P > 0.1, Table 1). However, we found that patients with blood type non-O had a higher percentage of elevated CRP serum level compared with those with blood type O (P = 0.038, Table 1).
Treatment modalities and response
The primary treatment modalities were as follows: 436 (62.6%) patients received chemotherapy combined with radiotherapy, 171 (24.5%) received chemotherapy alone, 72 (10.3%) received radiotherapy alone, and 18 (2.6%) received only best supportive care. The treatment details and responses are listed in Table 2. No significant difference was found in treatment modalities when dividing patient into either four blood type groups (O vs. A vs. B vs. AB, P = 0.701) or two groups (O vs. non-O, P = 0.690). After the initial treatment, 509 (75.0%) of the 679 treated patients displayed a complete response (CR) or CR unconfirmed (CRu). The rate of CR to initial treatment was significantly higher in the blood type O group than in the blood type non-O group (79.2% vs. 69.5%, P = 0.005).
The effect of ABO blood type on survival of patients with ENKTL
There were 302 deaths (43.3%) during a median follow-up of 41 months (range 1–214 months). The deaths were due to tumor progression (n = 289), treatment-related toxicities (n = 3), cardiovascular diseases (n = 2), and unknown causes (n = 8). The estimated 5-year PFS and OS rates for all 697 patients were 47.3% and 53.6%. The 5-year PFS rates for blood type A, B, AB, and O groups were 43.1%, 39.7%, 53.3%, and 55.0%, respectively (P = 0.002, Fig. 1a). The 5-year OS rates were 49.6%, 45.5%, 58.0%, and 62.0%, respectively (P = 0.007, Fig. 1b). Because the ENKTL patients with blood type O had longer PFS and OS compared with those with blood type A, B, or AB, we therefore divided these patients into blood type O and non-O (A, B, and AB) groups. The patients with blood type O had significantly higher 5-year PFS rate (55.0% vs. 42.9%, P < 0.001, Fig. 1c) and 5-year OS rate (62.0% vs. 48.9%, P = 0.001, Fig. 1d) compared with those with blood type non-O. Since patients with blood type AB and O have similar PFS and OS, we next examined the effect of ABO blood type on survival by comparing type O/AB versus type A/B. We found that patients with blood type O/AB had significantly higher 5-year PFS rate (54.6% vs. 41.5%, P < 0.001, Fig. 1e) and 5-year OS rate (61.3% vs. 47.6%, P = 0.001, Fig. 1f) compared with those with blood type A/B.
Blood type non-O was significantly associated with shorter OS in patients with Ann Arbor stage I/II disease (P = 0.002), but not in advanced cases (P = 0.151). For patients receiving chemotherapy plus radiotherapy, blood type non-O was also significantly associated with shorter OS (P = 0.003), whereas among patients receiving chemotherapy or radiotherapy alone, blood type non-O did not significantly affect the survival (P = 0.158 and P = 0.352, respectively). Blood type non-O was also associated with a worse outcome among patients receiving anthracyclines-containing chemotherapy (P = 0.010) or l-asparaginase-containing chemotherapy (P = 0.011). Table 3 displays the detailed data of prognostic significance of ABO blood type (O vs. non-O) in different subgroups.
Univariate and multivariate Cox regression analysis
Table 4 displays the results of the univariate and multivariate analysis of potential predictors of PFS and OS. Multivariate analysis using the forward conditional Cox region model identified tumor size ≥5 cm (risk ratio [RR] = 1.555, 95% confidence interval [CI] 1.130–2.139, P = 0.007), regional lymph node involvement (RR = 1.323, 95% CI 1.041–1.680, P = 0.022), blood type non-O (RR = 1.539, 95% CI 1.266–1.932, P < 0.001), and an IPI score ≥2 (RR = 2.285, 95% CI 1.751–2.983, P < 0.001) as adverse factors for PFS. In the multivariate analysis for OS, age >60 years (RR = 1.904, 95% CI 1.434–2.527, P < 0.001), tumor size ≥5 cm (RR = 1.720, 95% CI 1.233–2.400, P = 0.001), stage III/IV (RR = 2.114, 95% CI 1.554–2.875, P < 0.001), elevated LDH levels (RR = 1.514, 95% CI 1.183–1.936, P = 0.001), and blood type non-O (RR = 1.491, 95% CI 1.166–1.906, P = 0.001) were significant independent predictors of OS.
The comparison of prognostic value between ABO blood type and IPI and KPI models
Using the IPI predictive model, we identified 3 categories of patients with different survival outcomes: 598 (85.8%) patients in the low-risk (IPI = 0–1) group, 87 (12.5%) in the intermediate-risk (IPI = 2–3) group, and 12 (1.7%) in the high-risk (IPI = 4–5) group. The 5-year OS rate was 58.6% for the low-risk group, 25.3% for the intermediate-risk group, and 22.2% for the high-risk group (P < 0.001, Fig. 2a). Significant differences in survival were also found between the low-risk and intermediate-risk groups (P < 0.001) and between the intermediate-risk and high-risk groups (P = 0.024). However, based on the IPI data, 85.8% of the patients were disproportionately grouped into the low-risk group, and the IPI score was unable to identify patients with different survival statuses within the low-risk group, whereas ABO blood type (O vs. non-O) efficiently categorized patients in the low-risk IPI group into two subgroups with different survival outcomes (P = 0.010, Fig. 2b).
The KPI model balanced the distribution of patients in different risk categories more efficiently than the IPI model (score 0:233 cases, 33.4%; score 1:245 cases, 35.2%; score 2:145 cases, 20.8%; and score 3–4:74 cases, 10.6%), and it was able to discriminate between patients with different survival outcomes. The 5-year OS rate was 63.8% for the KPI = 0 group, 54.2% for the KPI = 1 group, 50.0% for the KPI = 2 group, and 27.1% for the KPI = 3–4 group (P < 0.001). Moreover, the KPI model significantly distinguished between the low- and intermediate-to-low-risk groups (KPI = 0 vs. KPI = 1, P = 0.010, Fig. 3a), but not between the intermediate-to-low- and high-to-intermediate-risk groups (KPI = 1 vs. KPI = 2, P = 0.321, Fig. 3b). The KPI model also significantly distinguished between the high-to-intermediate- and high-risk groups (KPI = 2 vs. KPI = 3–4, P = 0.002, Fig. 3c). In contrast, ABO blood type (O vs. non-O) was efficient at discriminating patients with a KPI score = 0–1 (P = 0.020, Fig. 3d), 1–2 (P = 0.015, Fig. 3e), or 2–4 (P = 0.019, Fig. 3f).