Predictive value of the neutrophil to lymphocyte ratio for disease deterioration and serious adverse outcomes in patients with COVID-19: a prospective cohort study

We conducted a prospective cohort study of COVID-19 patients who received medical treatment at the Third People’s Hospital of Yichang, Hubei Province, from January 9 to February 26, 2020. All patients who were confirmed by laboratory tests (positive for RT-PCR of nasal and pharyngeal swab specimens and/or specific IgM/G antibody assay of serum) before admission were recruited for the study. Basic information (e.g., age, sex, smoking, drinking, and history of comorbidities) was collected through face-to-face interviews with the patients or their families using a standardized questionnaire [2]. Routine blood examinations, including complete blood cell counts, coagulation profiles and serum biochemical tests, were conducted at hospital admission. Patients with missing neutrophil and/or lymphocyte records at hospital admission were excluded from the study. The patients were then treated according to the diagnosis and treatment plan for COVID-19 (7th version) issued by the National Health Commission of the People’s Republic of China [10] and were followed until discharge (for those who recovered) or death. The last patient was discharged on March 25, 2020. The changes in disease condition, treatment details and clinical outcomes of the disease were recorded during the follow-up period. The study was approved by the ethics committee of the First Affiliated Hospital of Fujian Medical University. The hospital was in charge of supporting the treatment of patients with COVID-19 in Yichang City.

Patients with an NLR, calculated by the neutrophil and lymphocyte data measured at hospital admission, equal to or higher than the cutoff value (2.6937) were categorized as the exposed group; otherwise, they were classified as the control group. The COVID-19 condition at hospital admission was evaluated and classified into mild/moderate, severe, and critical according to the diagnosis and treatment guidelines [10]. Patients were reclassified if their condition worsened and reached a higher classification criterion.

The primary outcome was disease deterioration, defined as progression of disease severity during hospitalization, including changes from from mild/moderate to severe and/or critical and from severe to critical. The secondary outcomes were shock and death during the treatment.

The NLR was expressed as the median (interquartile range) since the data were nonnormally distributed (P < 0.001 for Kolmogorov-Smirnov test). Categorical data are expressed as n (%) between patients with and without disease deterioration. Differences between the two groups were compared by chi-square or Fisher’s exact test (if applicable) for categorical variables. Box diagrams of the distribution of NLR between patients with and without the outcomes (i.e., disease deterioration, shock, death) are shown and were compared by Mann-Whitney U-test.

The predictive performance of the NLR for disease deterioration, shock and death was assessed by estimating the area under the curve and the corresponding 95% confidence intervals (95% CI) of the receiver operating characteristic curve method. The cutoff value and corresponding sensitivity, specificity, positive predictive value and negative predictive value were estimated.

A total of 352 patients with COVID-19 were included in the cohort and followed until the outcome was observed (average follow-up time of 24.3 days, ranging from 5 to 51 days). Among these subjects, 341 patients (96.9%) had a measured NLR at hospital admission and were included in the analysis. The mean (SD) age of patients in the cohort was 52.1 (18.1) years; 190 patients (54.0%) were men, and 117 patients (33.2%) had a history of at least one comorbidity. At hospital admission, 301 patients (85.5%) were classified as mild/moderate, 48 (13.6%) were classified as severe and 3 (0.9%) were classified as critical. During the follow-up, 51 (14.5%) patients’ conditions deteriorated, 15 patients (4.3%) had complicated septic shock, and 15 patients (4.3%) died (Table S1). Baseline characteristics, except smoking and drinking, were found to be unbalanced between patients with and without deterioration (Table 1).

The median NLR measured at hospital admission was 2.30, with a range from 0.11 to 85.5. The NLR was higher in patients with deterioration than in those without deterioration (median: 5.33 vs. 2.14, P < 0.001), and higher in patients with serious clinical outcomes than in patients without serious clinical outcomes (shock vs. no shock: 6.19 vs. 2.25, P < 0.001; death vs. survival: 7.19 vs. 2.25, P < 0.001) (Table 1, Fig. 1).

The NLR measured at hospital admission had high value in predicting disease deterioration, shock and death, with areas under the curve of 0.801, 0.830, and 0.828, respectively, and a cutoff value of 2.6937 (Fig. 2). The sensitivity of the ≥2.6937 NLR cutoff for predicting the subsequent presence of disease deterioration, shock and death was 82.0% (95% CI, 69.0 to 91.0), 93.3% (68.0 to 100), and 92.9% (66.0 to 100), respectively, and the corresponding negative predictive values were 95.7% (93.0 to 99.2), 99.5% (98.6 to 100) and 99.5% (98.6 to 100), respectively (Table 2).

Compared with patients who had an NLR of < 2.6937, the adjusted hazard ratios for disease deterioration, shock and death were 4.1, 13.1 and 11.3, respectively, among those with an NLR of ≥2.6937 (Table S2). The cumulative risk of the clinical outcome was significantly higher in patients with an NLR of ≥2.6937 than in those with an NLR of < 2.6937 (all P < 0.001, Figure S1).