Efficacy of neutrophil-lymphocyte ratio, serum lactate, and lactate clearance in predicting mortality in patients with sepsis admitted to a tertiary care hospital: A prospective observational study

INTRODUCTION

Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to an infection.^[1] Other than contributing to significant morbidity and mortality, it has brought forth a substantial health care and economic burden worldwide during the past decade.^[2] In hospital, the mortality rate in patients with septic shock is approximated to 30%, making sepsis the second most common cause of death.^[1,3,4]

Biomarkers have different diagnostic and prognostic implications in predicting outcomes in patients with sepsis and septic shock.^[5] Among different biomarkers, serum procalcitonin (PCT) and C-reactive protein (CRP) have been used since ages in sepsis for different purposes, denoting disease severity, prediction of outcome/death, antibiotic de-escalation, etc. However, both the above parameters have been shown to be raised due to non-infective inflammatory causes; thus, their reliability in predicting disease outcome has been questioned.^[6] In postoperative sepsis patients, though serum PCT levels decreased significantly among survivors, CRP levels did not differ.^[7] Moreover, due to significant inter-individual variation, no standard cutoff values could be proposed.^[7]

The neutrophil-to-lymphocyte ratio (NLR) is a rapid and simple parameter that can be used to predict severity and outcomes in sepsis patients.^[6,8] Underlying infection leads to an increase in neutrophil count, whereas inflammatory stress leads to lymphopenia.^[9] These factors along with sepsis-induced lymphocyte apoptosis increase NLR, which helps to stratify patients as per disease severity even before a change in total leucocyte count (TLC). Serum lactate levels and lactate clearance (LC) are often used as a surrogate marker for tissue hypoperfusion and also serve to prognosticate patients with sepsis. Presentation at advanced stage of sepsis leads to death within 24 h or may have protracted hospital course, performance of these biomarkers in these clinical scenarios is questionable. Different cutoff values of LC have been proposed, as to denote adequate response to therapy.^[10-12] Studies found that persistently elevated lactate and lactate non-clearance are ominous signs associated with worse outcomes in sepsis.

In patients with sepsis admitted to intensive care unit (ICU), both NLR (≥14.08 ng/mL) and serum lactate (≥2.99 mmol/L) predicted 28-day mortality with moderate certainty [area under receiver operating curve (AUROC) ~0.6]; however, their combination performed better (AUROC ~0.7).^[13] While individual markers may predict major health-related outcomes with varying reliability across different patient population, their combined values may be more accurate. Rapid bedside prognostic markers are of paramount importance in the emergency department (ED), helping to triage patients and tailor treatment accordingly. Therefore, we aim to evaluate the individual as well as combined efficacy of NLR, serum lactate, and LC, measured on presentation to the ED, in predicting 28-day mortality due to sepsis.

MATERIALS AND METHODS

RESULTS

Demographic parameters, vitals, and interventions

A total of 87 patients were screened sequentially, throughout the study period and after applying inclusion/exclusion criteria, 50 patients were analyzed [Figure 1]. The mean age of the population was 52.82 (±12.73) (mean [±SD]), 46% were male. Based on 28-day mortality, 64% were survivors.

Close

Figure 1:

Flow chart of the study. AKI: Acute kidney injury.

Export to PPT

During interim analysis, the AUROCs of variables showed moderate to fair predictive accuracy. Post hoc power analysis was performed, considering survivors (n = 33) and non-survivors (n = 17) as separate groups, based on all three parameters, that is, NLR, LC, and initial lactate, the power analysis showed 97%, 55.6%, and 84% power with respect to NLR, LC, and lactate, respectively. Therefore, no further patients were included after this analysis.

Baseline demographic characteristics as well as status of co morbidities were similar between the two groups [Table 1]. However, survivor group had higher SBP, DBP, MAP, similar RR and GCS, and lower HR than non-survivors, which were both clinically and statistically significant.

Table 1: Demographic characteristics, vitals, and interventions of enrolled patients.

Parameters	Survivors (n=33 [66%])	Non-survivors (n=17 [34%])	P-value
Age, (mean±SD)	51.06±12.95	56.24±11.93	0.176
Male gender, n (%)	14 (42.42)	9 (52.90)	0.490
Ideal body weight (kg), (mean±SD)	65.09±12.17	68.18±13.86	0.422
Cardiovascular disease, n (%)	11 (33.33)	8 (26.09)	0.354
COPD, n (%)	6 (18.18)	2 (11.76)	0.567
Diabetes, n (%)	10 (30.30)	3 (15.38)	0.152
Others (Depression, BPH), n(%)	2 (6.06)	0 (0)	0.059
Systolic blood pressure (mm Hg), (mean±SD)	103.03±23.69	84.12±13.77	0.001**
Diastolic blood pressure (mm Hg), (mean±SD)	70.48±24.75	48.71±9.48	0.001**
Heart rate (beats/min), (mean±SD)	114.09±12.77	126.94±8.25	0.011**
Respiratory rate (breaths/min), (mean±SD)	32.12±4.78	37±4.80	0.437
GCS, (mean±SD)	13.21±2.32	11.94±2.60	0.086
q-SOFA, median (IQR)	2 (2–2)	3 (2–3)	0.02**
SOFA score baseline, media n (IQR)	5 (4–7)	11 (6–12)	0.001**
Fluid infused in first 6 h (mL), (mean±SD)	1432.12±236.09	1501.76±273.61	0.354
Vasopressors used within 6 h, n (%)	8 (24.24)	11 (64.70)	0.015**
Need for mechanical ventilation, n (%)	36.36	76.47	0.006**

**P<0.05. SD: Standard deviation, GCS: Glasgow coma scale, q-SOFA: Quick sequential organ failure assessment, COPD: Chronic obstructive pulmonary disease, IQR: Interquartile range, BPH: Benign prostatic hyperplasia

Expectedly, the q-SOFA scores were higher for non-survivors as compared to survivors. The survivor group had lesser requirement of vasopressors and mechanical ventilation, had significantly lower SOFA score as compared to non-survivors [Table 1].

Relationship between NLR, LC, initial lactate, and mortality

NLR, serum lactate, and LC predicted mortality with moderate to good certainty; however, combination of NLR and LC did not increase predictive accuracy. NLR, combination of NLR and LC, and LC had AUROC in descending order [Table 3, Figures 2 and 3]. After adjusting for confounding variables, that is, DBP, SI, and SOFA score, multistep multivariate regression analysis revealed that, both LC and NLR were able to predict mortality independently [Table 3]. NLR, initial lactate, and LC showed weak to modest correlation between them (NLR vs. LC-spearman coefficient, r = 0.358 [P = 0.01] and NLR vs. initial serum lactate-r = 0.313 [P = 0.03]) [Figures 4 and 5].

Table 3: Univariate and multivariate analysis.

Biomarker	AUROC	Optimal cut-off	Sensitivity (%)	Specificity (%)	P-value
NLR	0.87	11.13	76.47	69.70	0.002**
Lactate clearance (%)	0.73	17.07	78.79	76.47	0.001**
Initial lactate mmol/L	0.77	3.30	70.59	72.73	0.015**
Multivariate analysis1
Biomarker	Odds ratio	Standard error	Z score	P-value	95% CI
NLR	5.1	3.82	2.17	0.03**	1.17–22.16
Lactate clearance (%)	7.4	5.49	2.7	0.007**	1.79–31.67

**P<0.05. ¹Multivariate analysis against diastolic blood pressure, shock index, and sequential organ failure assessment score. AUROC: Area under receiver operating curve, NLR: Neutrophil-to-lymphocyte ratio, CI: Confidence interval

Close

Figure 2:

Receiver operating curve (ROC) of neutrophil-tolymphocyte ratio.

Export to PPT

Close

Figure 3:

Receiver operating curve of lactate clearance. ROC: Receiver operating curve.

Export to PPT

Close

Figure 4:

Correlation between lactate clearance and neutrophil-tolymphocyte ratio (NLR).

Export to PPT

Close

Figure 5:

Correlation between initial lactate and neutrophil-tolymphocyte ratio (NLR).

Export to PPT

DISCUSSION

This single center and prospective observational study revealed that NLR and LC predicted 28-day mortality with moderate reliability. Univariate analysis with optimum cutoff value of ≥11.13 for NLR, mortality could be predicted with nearly 71% sensitivity and 72% specificity (AUROC 0.87). Similarly, with a cutoff level of ≥17.07%, LC had 75.8% sensitivity and 76.5% specificity in predicting mortality (AUROC 0.73). However, their combination in multivariate analysis did not offer extra predictive accuracy (AUROC 0.82). NLR and LC were also modestly correlated among themselves. Our study had almost 34% mortality, which was similar to those reported in two other studies.^[13,14]

NLR is a simple and rapid tool used to stratify patients according to their disease severity. Due to any physiological stress, for example, sepsis, trauma, catecholamines, and steroids can lead to neutrophilia, both absolute (increased bone marrow production) and relative (peripheral to central migration). Simultaneously, cortisol-induced lymphopenia and sepsis-induced lymphocyte apoptosis cause ALC to fall. Due to the above factors, the NLR rises and the degree of rise correlates with the severity of underlying infection.^[8] Higher NLR was associated with worse functional outcome, in patients presenting with intracerebral hemorrhage in the ED, indicating that higher degrees of stress lead to higher NLR (AUROC ~0.8).^[15]

One previous study reported that, NLR ≥14.08 was able to predict survival with almost 78% sensitivity and 50% specificity (AUROC 0.634) in sepsis patients.^[13] Another study reported higher NLR cutoff ≥23.8 (AUROC 0.695) for predicting mortality. Our study revealed similar finding but with lower cutoff levels (Univariate analysis). Heterogeneity of study population and disease severity could have accounted for such different levels. Despite a lower cutoff, our study revealed similar sensitivity ~71% and even higher specificity ~72%, as compared to Liu et al.^[13] Except NLR, difference in age and disease severity also accounted for worse outcome in sepsis patients.^[16]

Similarly, in our study, non-survivors had greater disease severity, requirement of mechanical ventilation, and vasopressors support; however, they had similar age distribution. NLR remained an independent predictor of mortality, even after adjustment.

Serum lactate is still used as a surrogate marker of tissue hypoperfusion. Tissue hypoxia leads to anaerobic metabolism and, thereby, increased serum lactate: Pyruvate ratio. Therefore, raised serum lactate and its reduced utilization (lower LC) are ominous signs in patients with sepsis, as it predicts mortality. Elevated serum lactate ≥2.99 mmol/L levels were nearly 83% sensitive and 55% specific in predicting sepsis-related mortality (AUROC 0.7). Raised serum lactate ≥4 mmol/L predicted mortality in patients presenting to ED as well as in ICU.^[17,18] The importance of raised serum lactate was highlighted in the earlier definition of “severe sepsis,” where serum lactate ≥4 mmol/L predicted worse outcomes.^[19] In our study, non-survivors had significantly higher serum lactate. The cutoff used in our study was somewhere in between the above-mentioned values, that is, 3.3 mmol/L (AUROC 0.77) [Figure 6]. However, a single-elevated serum lactate value can be erroneously interpreted as abnormal, as endogenous catecholamines and liver dysfunction can lead to falsely elevated serum lactate. Therefore, the trend of serum lactate and LC carry somewhat better prognostic significance. While, higher LC was associated with better survival, different cutoff levels have been proposed till date, ranging from 10 to 50%.^[10,20] One study reported complete LC, that is, 100% LC as the best predictor of survival.^[12] Our study revealed that LC ≤17.07% was best as well as an independent predictor of mortality (AUROC 0.73). A modest correlation existed between NLR versus LC (r = 0.358) and NLR versus serum lactate (r = 0.313), which was similar to the study by Liu et al.^[13]

Close

Figure 6:

Receiver operating curve (ROC) of initial lactate. ROC: Receiver operating curve.

Export to PPT

Liu et al.,^[12] also reported better predictive power with combination of NLR and serum lactate.^[13] However, our study failed to demonstrate any benefit of such combination. As such, we used a lower cutoff for NLR, which resulted in greater sensitivity as compared to Liu’s study. Second, we used LC instead of serum lactate for combination with NLR.

Other than being an observational study which is fraught with risk of biases, our study had limited sample size. Despite a small sample size, our study showed fair reliability of various markers in predicting mortality, and better post hoc power (>80%) as compared to Liu et al. (post hoc power 5.6%).^[13] Second, we accounted for major confounders related to mortality. It is intuitive to think that other confounders could be present, which if included, could change the result. We evaluated all-cause mortality and not disease or time (early vs. late) specific mortality. Third, data on culture positivity and use of antibiotics were not included. Finally, combined efficacy of NLR and serum lactate in predicting mortality was not evaluated. Indeed, this was a small feasibility study not aimed to detect causality.

CONCLUSION

In patients presenting with features of sepsis in the ED, NLR, serum lactate, and LC were all associated with 28-day survival. Among these, NLR (with cutoff ≥11.13) performed better (AUROC 0.87) than serum lactate (cutoff ≥3.3 mmol/L, AUROC 0.77) and LC (cutoff ≥17.07%, AUROC 0.73) and after adjusting for confounders, NLR and LC independently predicted mortality. Therefore, the above simple and yet rapid markers can predict outcome with fair reliability, which can be used to triage patients and tailor better treatments.