This single-center, prospective, observational cross-sectional study with a cohort design was conducted in the Department of Medicine, Jawaharlal Nehru Medical College and its attached hospital Acharya Vinoba Bhave Rural Hospital (AVBRH), Sawangi (Meghe), Wardha. The suspicious sepsis patients were admitted to theICU, a 40-bed unit of the hospital under the Medicine Department of Acharya Vinoba Bhave Hospital, Sawangi (Meghe) associated to Datta Meghe Institute of Medical Sciences, Wardha.

The entire study was conducted as per the principles of the Declaration of Helsinki and local regulations. Relevant regulations and Good Clinical Practices and were followed throughout the study duration by each study team member. The present study was conducted after the clearance of the Institutional Ethical Committee (IEC number, DMIMS (DU)/IEC/2020-21/9300) of Acharya Vinoba Bhave Rural Hospital (AVBRH). Patients who showed their active interest to follow the projected research study design after a detailed written informed consent process were enrolled in the study by signing the informed consent.

This mid-term study was conducted during November 2020 to April 2022 as we found that peaks of infectious diseases leading to sepsis were at peak level during this time.

Patient enrollment was conducted at the Department of Medicine, Jawaharlal Nehru Medical College and its attached hospital Acharya Vinoba Bhave Rural Hospital (AVBRH), Sawangi (Meghe), Wardha. Subjects were screened in the ICU between Nov 2020 to Apr 2022 were eligible with a diagnosis of sepsis. Sepsis was diagnosed on SEPSIS-3 criteria.

1)

Age of the patients ≥18 years admitted to the ICU of the hospital.

1)

Patients with either hepatic dysfunction, chronic kidney disease.

See Figure 1 for a plan of the study.

Medical residents under the direction of surgeons and clinicians from several medical disciplines made up the ICU's medical staff. Fellows with specialized training in detecting potential sepsis patients as per SOFA criteria went four times a day to the ICU to look for people who met the inclusionary criteria. Without being aware of the patients' subsequent hospital courses, we prospectively identified the patients and studied the ICU documents. In order to locate patients, we looked through the daily ICU log for admissions with an infection-related diagnosis (such as pneumonia, shortness of breath, etc). To confirm a suspect of infection based on ICU presentation as recorded in the medical decision-making section of the chart, all records of patients with infection or a probable infection underwent a confirmatory review.

According to institutional protocol, individuals with suspected of sepsis with SOFA score >2 had their venous lactate levels checked. The Glasgow Coma Score was assessed and recorded in the patient's record. Several other parameters like significant vitals, blood cultures, antibiotics, fluid therapy, resuscitation, length of hospital stay, complication, if any were also taken into account. The hospital ICU team's choices were unaffected by the researcher. We followed the study participants from initial admission to the hospital ICU until discharge or death.

The first venous serum lactate levels (mmol/L) obtained in the ICU were collected, together with the basic chemistry panel, estimated AG ([Na mEq/L] [Cl mEq/L] [HCO ₃ mEq/L]). The lactate levels were tested using a chemical analyzer from serum taken in a heparinized tube. According to normal reference range established by our institutional laboratory, we defined clinically relevant lactate to be >4 mmol/L and an anion gap of >12 was regarded as the highest limit of normal. Survival to hospital was referred to as a secondary outcome of in-hospital mortality.

The medical record number, baseline vital signs (BP, HR, RR), AG, and serum lactate, collected by venous sample, were obtained and documented by the ICU doctors in collaboration with the study team members upon ICU presentation and fulfilment of inclusion criteria. The following formula is used by to determine the anion gap: AG = Na + K – Cl + HCO 3 .

Serum Lactate levels were measured by Lactate was measured by an enzymatic lactate oxidase assay using the Cobas 6000 ^® analyzer, (Roche Diagnostics, IN, USA).

Sample size calculation:

Minimum sample size required n = Z 1 − α 2 2 Sens 1 − Sens d 2 X Prev ∝ Alpha = Type I error = 0.05

Estimated specificity of anion gap=0.84

Prevalence of positive character (Prev)=0.283

Estimation Error 5%=7%

n=148

Minimum sample size required is 148, considering the drop out 10% sample size required was 163.3 samples drop out from the study, as per protocol analysis total 160 samples taken for observations.

The sampling method was consecutive sampling with a sample size of 160 subjects.

The data from ICU admitted sepsis patient inclusionary of SOFA Score was entered into the Microsoft Excel sheet version 2016. Data were analyzed using SPSS software version 15 (SPSS IBM statistics, Chicago, USA), we can recommend R-studio software as a proprietary free alternative. We presented categorical variables as frequency and percentage. The continuous variables were presented as mean and standard deviation. In order to accurately characterize the results, demographic data were expressed as means with standard deviation, medians (interquartile range), or percentages. The student’s t-test was used to compare the means of two independent sets of quantitative variables. The Friedman and the Mann-Whitney test were used to gauge the non-symmetrical distributions of the variables. The Chi-square test was used to make proportional comparisons. Receiver operating curves were obtained to determine the area under curve with serum lactate and AG in prediction of mortality, P value was considered significant if p<0.05.

In this study, we included 160 patients with sepsis with SOFA score >2.

Mean age of the population was 53.1±17.0 and 61.9% were males. Hypertension (41.3%) and diabetes (30%) were major comorbidities. On admission, mean HR was 123.1±20.8 bpm, mean RR was 29.8±9.2 per min, mean SBP was 82.3±28.7 mmHg, and O ₂ saturation level was 84±12%. The mean SOFA score was 9.5±5.3. 53.8%, patients had alerted mental status. 58.8% required ventilator whereas 63.1% needed inotrope support. Among laboratory parameters, mean Hb was 10.5±2.8 gm/dl, mean serum albumin was 3.0±0.8 mg/dl and mean serum creatinine was 2.6±2.6 mg/dl.

Mean serum lactate was 5.1±1.2 mmol/L and it was ≥4 mmol/L in 90% of patients.

Mean AG was 14.0±3.9 and it was ≥12 in 75.6% of patients. During a mean hospital stay of 10.6±8.8 days, mortality occurred in 45.6% of patients. Mean lactate level was significantly higher in patients with high AG (5.6±1.1 vs. 4.1±0.8 mmol/L, p<0.0001). Similarly, the proportion of patients who had lactate levels ≥4 mmol/L was higher in those with AG ≥12 than AG <12 (95.9% vs. 71.8%, p<0.0001). Among non-survivors, serum lactate level was significantly higher than survivors (5.8±1.1 vs. 4.5±0.9 mmol/L, odds ratio 4.55 (95% CI 2.68-7.70), p<0.0001). All non-survivors had lactate levels ≥4 mmol/L. Among non-survivors, mean AG was significantly higher than survivors (15.8±3.7 vs. 11.9±3.4, odds ratio 1.39 (95% CI 1.23-1.56) p<0.0001). AG ≥12 was seen in significantly higher proportion non-survivors than survivors (89% vs. 64.4%, p<0.0001). The area under curve of receiver operating characteristic (AUC of ROC) in predicting mortality was significant for both serum lactate (AUC 0.797, p<0.0001) and AG (AUC 0.835, p<0.0001).

Among various other parameters, significant predictors of mortality were elevated HR, RR, SOFA score and lower levels of SBP, O ₂ saturation, PiO ₂/FiO ₂ ratio. Also, need of ventilator and inotropes predicted mortality. Hospital stay was shorter in non-survivors than survivors.

Sepsis is one the commonly encountered diagnoses in ICU setting. In an Indian setting, beside the Gram-negative bacterial infections, tropical illnesses like dengue, malaria, leptospirosis, enteric fever, and tuberculosis are also significant causes of severe sepsis/septic shock. ⁵ In the sepsis diagnosis and predicting the outcome of sepsis, multiple scores and biomarkers are used across the world. CRP, procalcitonin, proinflammatory cytokines, complement system biomarkers, and organ dysfunction biomarkers are employed. ²³ Elevation of serum lactate levels has been identified as important marker for predicting the outcome of sepsis. ²⁴ An elevated corrected AG usually reflects the presence of metabolic acidosis caused by the overproduction or decreased excretion of organic acids. In addition, elevated AG has been reported as a predictor of mortality in critically ill patients. ²⁵ We estimated the serum lactate levels and AG in sepsis patients and evaluated their role in predicting the in-hospital outcome (death or discharge). Below we discuss the findings from our study.

Age and gender

In our study, mean age was 53.1±17.0 years and 38.1% were in age group pf 41 to 60 years and 36.9% were above 60 years.

A recent Sepsis in India Prevalence Study (SIPS) from Hammond et al. reported median age of 60 years among patients admitted in ICU. ²⁶

Chatterjee et al. reported a mean age of 59.4±17.9 years in their study. ²⁷

A study from Mohamed et al. observed 44% of patients of sepsis being over the age of 60 years. Thus, sepsis is more common in middle-age to elderly population than younger population. ²⁸

Males were affected with slightly greater frequency. In our study, the proportion of males was 61.9%. Chatterjee et al. reported proportion of males affected to be 56.8%. ²⁷

Mohamed et al. also reported 71.25% of males in their study diagnosed with sepsis. These data indicate males are a slightly higher risk of sepsis than female. ²⁸

Comorbidities

Hypertension (41.3%) and diabetes (30%) were major comorbidities in our study.

Hammond et al. in their SIPS study reported diabetes (44.0%) and chronic renal failure (11.6%) as common comorbidities. ²⁶

Mohamed et al. reported type 2 diabetes mellitus and systemic hypertension 46.25% of patients each. This is similar to our observations. ²⁸

Sepsis etiology

We observed respiratory etiologies as most common cause of sepsis in our study followed by gastrointestinal and urinary etiologies.

Similar to our findings, Mohamed et al. reported respiratory tract as the suspected source of sepsis in 66.25% cases. ²⁸

Another study from Chatterjee et al reported similar results with respiratory tract (53.3%) as the most frequent site of infection followed by abdomen (14.9%), blood stream (14.3%) and urinary tract (12.9%). ²⁷

A study from Abu-Humaidan et al. from Jordan identified gastrointestinal cause (37.8%) as most frequent one leading to sepsis followed by respiratory causes (24.4%), genitourinary etiology (24.4%), skin and soft tissue infections (13.3%) and others (17.8%). These results indicate that respiratory, GI and urinary infections are the most frequent causes of sepsis. ²⁹

In-hospital outcome

In our study, during a median hospital stay of 9 days, mortality occurred in 45.6% of cases. A study from West Bengal by Chatterjee et al. reported in-hospital mortality rate of 63.6% whereas the ICU and 28-day mortality rates were 56% and 62.8% respectively. ²⁷ A study in Brazilian cohort of patients with sepsis by Sogayar et al. reported mortality rate of 49.1% and 36.7% in public and private hospitals. ³⁰ Another study from North India by Nasa et al. reported ICU mortality ranging 45.6% to 60.7% and 78.9% among younger (<60 years), old (60 to 80 years) and very old (>80 years) patients who were diagnosed with sepsis. ³¹

These data support our observation and indicate that sepsis is associated with significantly higher risk of mortality.

Serum lactate levels and anion gap estimates

Mean serum lactate levels in our study were 5.1±1.2 mg/dL and 90% had lactate levels of 4 mmol/L. A study from Mikkelsen et al. reported median serum lactate levels of 2.9 mmol/L (interquartile range: 2.0 – 4.4). The proportion of patients who had serum lactate levels >4 mmol/L was 24.7% and 50% among patients who presented without and with septic shock respectively. ³² Another study from Shapiro et al. reported serum lactate levels of >4 mmol/L in 10.5% of cases. ³³ A study from Singh et al reported mean serum lactate level at admission to be 2.87 ± 1.25 mmol/L. ³⁴ The relatively higher on admission lactate levels in our study population can be due to majority of patient presenting in septic shock as was observed by Mikkelsen et al. ³²

In our study, mean AG was 14.0±3.9. High AG (AG ≥12) was seen in 75.6% of patients. A study from Berkman et al. involving 1419 adult patients with sepsis, the mean AG was 11.8±3.6 whereas the mean lactate level was 2.1±1.3 mmol/L. ²² A study from Mitra et al. involved 441 patients with sepsis. The initial median AG was 15.2 (13.8–17.4) and the median lactate level was 2.1 (1.5–2.9) mml/L. ³⁵ L-lactic acidosis and ketoacidosis are the 2 most prevalent endogenous causes of high-AG metabolic acidosis (HAGMA). Type A and Type B L-lactic acidosis are the two subtypes into which L-lactic acidosis is typically subdivided. Type A L-lactic acidosis, which can be brought on by hypovolemia, heart failure, sepsis, acute severe anemia, convulsions, or cardiopulmonary arrest, is the outcome of marked tissue hypoperfusion. In patients with shock brought on by sepsis, the initial serum lactate level and the rate at which lactate levels recover are indicators of survival. If tissue perfusion cannot be quickly restored, the prognosis is typically quite poor. ³⁶ The high anion gap in our study indicates significant acidosis induced by hyperlactatemia.

Serum lactate levels and anion gap association with mortality

We first studied the association of lactate levels with AG. The mean serum lactate level was significantly higher in patients with high AG (p<0.0001). Also, 95.9% of patients with high AG (≥12) had serum lactate levels ≥4 mmol/L. Berkman et al. reported 80% sensitivity and 69% specificity of elevated AG (>12) that predicted lactate levels >4 mmol/L. They also observed a 7.3-fold increased risk of having a lactate >4 mmol/L for patients who had high AG in the emergency department. It indicates that sepsis patients may have anion gap abnormalities that is related to increased lactic acid levels. It is therefore advisable that high anion gap should be considered in the assessment of sepsis. ²²

The lactate level was significantly elevated in non-survivors compared to survivors (p<0.0001) and all non-survivors had elevated lactic acid levels. Similarly, the AG was significantly higher in non-survivors than survivors (p<0.0001). High AG (≥12) was seen in 89% of non-survivors than 64.4% of survivors (p<0.0001). Serum lactate levels had odds ratio of 4.55 and AG had odds ratio of 1.39 suggesting higher risk of mortality with both markers. Multiple studies have reported similar outcomes. Shapiro et al. reported higher mortality with serum lactate levels of >4 mmol/L (4.9%) compared to 2.5 mmol/L (9.0%) and 2.5 to 3.9 mmol/L (28.4%). ³³ In a retrospective cohort study of severe sepsis patients, Mikkelsen et al. found that in hemodynamically stable individuals with intermediate lactate levels (2.0-4.0 mmol/L) were also at substantial risk, as opposed to patients with levels <2.0 mmol/L. They used lactate cut-off of 4.0 mmol/L as a screening tool for sepsis. ³² Another retrospective cohort study from Wacharasint et al. has confirmed this finding, by observing a significantly higher mortality in those with lactate levels >4 mmol/L than among patients with high normal lactate levels (1.5–2.3 mmol/L) and intermediate lactate levels (2.3–4.0 mmol/L). ³⁷ A study from Khosravani et al. reported 1.94–10.89-fold higher fatality rate with on admission serum lactate levels >2 mmol/L when compared to values below 2 mmol/L. ³⁸ Claridge et al. reported that not only the admission lactate levels but also the persistent hyperlactatemia predicts in-hospital mortality. ³⁹ Mitra et al. reported mortality rates with elevated lactate and AG levels. They reported in-hospital mortality rate of 11.7% among 231 patients with lactate >2 mmol/L and 14.9% among 221 patients with elevated AG (>16). ³⁵ Berkman et al. reported significantly higher mortality with elevated AG (>12) than lower AG (<12) (9.5% vs. 3.8%, relative risk 2.50). They also stated that compared to AG threshold of >16 and >20, AG>12 performed the best in predicting mortality. ²² Adams et al. in a retrospective cohort analysis identified lactic acidosis with a sensitivity of 58.2%, specificity of 81.0%, and negative predictive value of 89.7% by using AG cutoff 12. Thus, it can be concluded that AG > 12 might be considered as one of the important predictors of in-hospital mortality in sepsis. This is further substantiated by our observations of AUC of ROC being 0.797 (95% confidence interval 0.728-0.865, p<0.0001) for AG and 0.835 (95% confidence interval 0.773 – 0.896, p<0.0001) for lactate in predicting mortality. ⁴⁰ Shapiro et al. reported that ROC AUC for lactate level as a predictor of death was 0.670. This finding is nearly similar to our results. ³³ Also, Mitra et al. reported AUC of 0.630 and 0.6800 for serum lactate and AG for predicting in-hospital mortality. Thus, both parameters have good sensitivity in predicting the mortality in sepsis. ³⁵

Other predictors of mortality

We observed that higher heart rate, higher respiratory rate, lower systolic BP, lower O ₂ saturation, need of invasive ventilation, inotrope requirements, higher SOFA scores, lower PaO ₂/FiO ₂ and lower hospital stay were associated with in-hospital mortality in sepsis patients. A study from Kerala by Mohamed et al. reported elevated CRP (>100), APACHE II score >25, and need of invasive ventilation as predictors of mortality in adult patients with severe sepsis. A higher heart rate and lower mean arterial pressure at the time of admission to the ICU were also predictors of mortality. ²⁸ Vincent et al. reported significantly higher mortality in patients undergoing mechanical ventilation. This imply that need of invasive mechanical ventilation could be a predictor of mortality in severe sepsis. ⁴¹ A systematic review by Minne et al. identified SOFA score in patients at presentation and at 48 hours are significant predictors of mortality. ⁴² Similarly, various other studies have identified predictors of mortality. Bale et al. reported SOFA score, Shrestha et al. reported anemia, SOFA score, SAPS II and III scores and Mohan et al. reported SAPS III, anemia, and SOFA scores as predictors of mortality. ^{43 – 45} Beside these, thrombocytopenia, and acute renal failure are also identified as predictors of mortality in other studies. ^{46 , 47}

This was one of the few studies that evaluated AG as parameter for predicting mortality in sepsis. Higher AG can be associated with increased mortality. However, our study has certain limitations. We did not assess other commonly used biomarkers such as CRP, procalcitonin. Comparison of lactate and AG with these parameters might provide greater insights as to which parameters can be best predictor of mortality. We did not assess APACHE II score to determine the severity of illness as APACHE II score has been identified as predictor of mortality. Also, we assessed in-hospital mortality only. Assessing 28-day or 90-day mortality can provide different picture in associating AG with such short-term mortality figures.

The present study was conducted after the clearance of the Institutional Ethical Committee (IEC number, DMIMS (DU)/IEC/2020-21/9300) of Datta meghe institute of medical sciences.