3.1 Patient characteristics
The patient characteristics table is shown in Table 2. A total of 75 patients were included in this study, after 5 who were lost to follow-up were excluded. The mean age was 52.69 (18-84), and 58(77.3%) patients were male. The median hospital stay was 17 days, and the mean BMI was 25.67, 64 (85.3%) patients underwent surgery after admission. Among the 75 patients, 31 (41.3%) smoked, and 23 (30.7%) drank alcohol. The comorbidities at admission were hypertension in 10 patients (13.3%), diabetes in 2 patients (2.7%), and coronary heart disease in 1 patient (1.3 %), 2 cases of COPD (2.7%), 2 cases of cerebral infarction (2.7%). A total of 3 cases (4.0%) had infection after operation, and they were cured after symptomatic treatment with antibiotics. At the last follow-up, a total of 3 patients (4.0%) died, including 1 patient who did not receive surgical treatment (AIS B, died 6 months after injury), and 2 patients who received surgical treatment (AIS A died 12 months after surgery; AIS D died 2 months after surgery). On admission, the AIS grades of the patients were grade A in 12 cases (16.0%), grade B in 10 cases (13.3%), grade C in 15 cases (20.0%), and grade D in 38 cases (50.7%), and 53 patients’ (70.7%) injured segment was cervical, 10 cases (13.3%) were thoracic, and 12 cases (16.0%) were lumbar.
3.2 Determination of cutoff value
According to the receiver operating characteristic (ROC) curve of the 75 tSCI patients at the last follow-up, the area under the ROC curve (AUC) of PNI and COUNT scores were 0.752 (P<0.001) and 0.766 (P<0.001), respectively. According to the Youden index, the optimal cutoff values of PNI and COUNT scores were 45.05 (95%CI 0.641-0.862) and 3.5 (95%CI 0.654-0.879), respectively (Figure 1). According to cutoff value, the patients were divided into high PNI group (≥45.05, 28 cases, 37.3%) and low PNI group (<45.05, 47 cases, 62.7%), high COUNT group (≥3.5,44例,58.7%) and low COUNT group (<3.5,31例,41.3%).
3.3 Results of PNI and COUNT subgroups
There was no significant difference in clinical characteristics between the high PNI group and the low PNI group (Table 3). While patients in the low CONUT group had a shorter hospital stay (P=0.042), a higher BMI index (P=0.034), and a higher motor score on admission (P=0.007) than those in the high COUNT group (Table 4). Statistical analysis showed the motor score, Barthel index score, light touch score, and improvement rate of AIS grade in the high PNI group at the last follow-up was better than that in the low PNI group, and the difference was statistically significant (P<0.05). The Barthel index score, light touch score and improvement rate of AIS grade in the low COUNT group was better than that in the high COUNT group, and the difference was statistically significant (P<0.05), and there was no significant difference in the changes in motor score (P=0.075), for details see Tables 5-6.
3.4 Relationship between CONUT score, PNI and AIS grade
A total of 51 (68%) patients improved at last follow-up (AIS grade ≥1), of which 28 (90.3%) were in the low COUNT group and 26 (92.9%) in the high PNI group. After controlling for hospitalization days and postoperative infection factors, Spearman correlation analysis showed that PNI and COUNT scores were strongly correlated with AIS grades (r=0.629, P<0.001; r=-0.620, P<0.001). At the same time, there was also a strong correlation between PNI and COUNT scores (r=-0.855, P<0.001), and the patients in the low PNI group and the high COUNT group had a large overlap (Table 7).
3.5 Impact of CONUT score and PNI on outcome
As shown in Table 8, compared with the improvement group, more patients with COUNT ≥ 3.5 and PNI<45.05 were observed in the non-improvement group (P<0.001). To explore the relationship between the last follow-up outcome and clinical characteristics, univariate analysis showed that compared with the improvement group, the patients in the non-improvement group had longer hospitalization days and higher postoperative infection rates (P<0.05). The clinical characteristic factors with P<0.2 were included in the multivariate Logistics regression equation. Because PNI and COUNT score had a strong correlation and were found to be complete intermediary variables after being included in the model, the models were established separately. The hospitalization days, drinking, COPD, postoperative infection, PNI or COUNT scores were included to construct multivariate Logistic regression analysis. The results showed that both PNI and COUNT scores were independent predictors of acute tSCI improvement (Tables 9-10), with adjusted odds ratios of 1.396 (95%CI: 1.141-1.709, P=0.001) and 0.284 (95%CI: 0.136-0.0.594, P=0.001).
The receiver operating characteristic (ROC) curve of improvement at last follow-up showed that PNI and COUNT scores had predictive value. The AUC of PNI and COUNT scores were 0.752 (95%CI 0.641-0.862, P<0.001) and 0.766 (95%CI 0.654-0.879, P<0.001), respectively, and the PNI cutoff value was 51.0% sensitive and 91.7% specific, the COUNT score was 54.9% sensitive and 87.5% specific (Figure 1). There was no statistical difference in the AUC comparing PNI and COUNT scores by Delong test (P=0.690).