Introduction

The Surviving Sepsis Campaign defines sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection.¹ In veterinary medicine, sepsis is still most commonly defined as systemic inflammation in response to infection.² Sepsis in human and veterinary medicine is associated with high morbidity and mortality rates.^1–3 In dogs with septic peritonitis, mortality rates range from 21% to 68%.⁴ Septic shock is associated with poorer prognosis, in which patients are at a higher risk of decompensation and organ failure; in dogs with septic shock, mortality rates approach 70% to 80%.^2,5 Early recognition of septic patients is important so aggressive and targeted intervention can be implemented immediately. Readily available biomarkers may be a helpful tool for diagnosis, prognostication, and therapy. Biomarkers are used in human patients with sepsis, and their use has been evaluated in veterinary studies.^6–11 However, information remains limited and inconsistent on the use of prognostic biomarkers in septic veterinary patients.^11–13

The lactate to albumin ratio (L:A) has been proposed in human medicine as a more effective early prognostic indicator for mortality than either lactate or albumin alone.^3,6,14,15 High L:A is associated with higher rates of organ failure and mortality rates in humans with sepsis and septic shock.^16–18

Increased lactate and lactate clearance have been extensively researched as biomarkers for critically ill human and veterinary patients.¹⁹ Hyperlactatemia in sepsis is multifactorial and occurs due to an increase in lactate production in combination with reduced clearance of lactate by the liver. Type A hyperlactatemia occurs due to tissue hypoperfusion and hypoxia leading to anaerobic glycolysis. Type B hyperlactatemia occurs in the absence of hypoxia. Mechanisms in septic patients include mitochondrial dysfunction and decreased lactate clearance by the liver.^5,19–21 In recent years, the belief that hyperlactatemia is solely due to hypoperfusion and anaerobic glycolysis has been challenged.^19,22–24 Hyperlactatemia in sepsis may also result from sympathetic nervous system stimulation of adrenergic receptors.²³

Albumin is a negative acute-phase protein in sepsis. Inflammatory cytokines in sepsis cause a down-regulation of hepatic albumin production.^25,26 Hypoalbuminemia may also be due to increased vascular permeability, hepatic dysfunction, gastrointestinal or renal losses, or decreased intake.²⁷ A decrease in serum albumin has been associated with increased morbidity and mortality in both critically ill humans and companion animals.^4,25,28–30

This study investigated whether L:A calculated from routine blood work is a viable clinical prognostic marker in septic dogs for early identification of high-risk patients. We hypothesized that more critically ill patients at risk of death or decision to euthanize would show an increased L:A compared to survivors.

Materials and Methods

This study was a single-center, retrospective case-control study conducted at an academic referral veterinary hospital. Records were de-identified but assigned a unique identifier code for analysis.

Data were extracted from the hospital electronic medical record system (IDEXX Cornerstone Practice Management Software) for all dogs diagnosed with sepsis between October 2012 (earliest system record date) and August 2021. Search terms used to identify these dogs were bacterial sepsis, septic peritonitis, septic shock, septicemia, and sepsis (unclassified). The source of sepsis was identified based on a review of cytology and/or culture results, necropsy results when available, or clinical diagnoses documented in the medical record. Controls were dogs with sepsis who survived to hospital discharge, and cases in which dogs died before discharge (cardiac arrest, euthanasia). Subject records were included in this study if at least one venous blood gas analysis and blood biochemistry panel were performed on presentation or by the referring veterinarian immediately before referral. If multiple values were recorded for lactate or albumin, then the first recorded value was used in the study. Subject records were excluded if laboratory readouts did not include either albumin or lactate, if there were concurrent diseases other than sepsis, if there were multiple surgeries before admission, if no diagnostics were performed, if no vital signs were recorded, or if no blood work results were recorded.

Statistical Analyses

The study design was a case-control comparison of in-hospital mortality status and association with the L:A ratio. There were two mortality categories: nonsurvivor (cardiac arrest, euthanized; “cases”), and survivors (survival to discharge, “controls”). Sample size was determined by availability of records meeting inclusion criteria (Figure 1). Primary outcome variables were lactate, albumin, and the L:A ratio. All other laboratory readouts and calculated metrics were secondary outcomes.

View Figure

• Download as Powerpoint
• Download Figure

Analyses were performed in SAS version 9.4.^a All variables were summarized by descriptive statistics. Categorical data were summarized as counts and percentages (%) and continuous data as medians (interquartile range [IQR], range minimum, and range maximum) in SAS proc means.

Differences between mortality categories for the primary outcomes (lactate, albumin, L:A ratio) were estimated by generalized linear models (SAS proc glm). Records with missing data were deleted, leaving only cases including both lactate and albumin for analysis. Potential confounding of differences between survivors and nonsurvivors were accounted for by one-to-one matching on age, sex, body weight, and SIRS score. Matching was performed in SAS proc psmatch using a “greedy” 1:1 matching algorithm without replacement and caliper 0.5.^37–39 Data were ln-transformed before analysis to correct for skew, then back-transformed to the original units. Both unadjusted and matched adjusted results are reported as means and 95% confidence intervals (SAS proc glm).

Robustness of results was assessed by sensitivity analyses on missing data patterns and mortality classification. Two bias patterns resulting from missing data were assessed: the probability that a missing value depends on a specific variable type (clinical, hematology, biochemical) and the probability that one mortality category is more likely than the other to have missing values across any or all variables. Observed and missing values for all response variables were coded as 1 and 0, respectively, then statistically significant patterns of variable association assessed by likelihood χ² tests.⁴⁰ Sensitivity to mortality classification (cardiac arrest vs. euthanasia) was assessed by deleting cardiac arrest cases and re-analyzing the reduced data set.⁴¹

Results

The medical record search identified 190 records, and records for 143 dogs met inclusion criteria (Figure 1). Signalment and clinical presentation data are shown in Table 1. Seventy dogs survived to discharge (49%), 53 dogs were euthanized (37%), and 20 dogs had in-hospital cardiac arrest (14%). The median age was 8 yr (IQR 4–11 yr; range: <1–17 yr) and median weight was 21 kg (IQR 8–40 kg; range 1–70 kg). Approximately 26% (37/143) were mixed-breed dogs, over half were female (74/143; 52%), and 76% (109/143) were castrated or spayed.

TABLE 1 Descriptive Statistics for Signalment and Clinical Signs at Presentation for 78 Dogs Classified by Clinical Outcome Status (Nonsurvivors, Survived to Discharge) Following Diagnosis with One or More Sepsis Conditions at Admission

View Fullscreen

The most commonly identified sources of sepsis were septic peritonitis (55%; 78/143), unclassified sepsis (20%; 29/143), and sepsis secondary to wounds or dermatological conditions (10%; 15/143). The remainder (15%; 21/143) encompassed a variety of different sources such as septic pyometra, urosepsis, septicemia, and neoplasia. Median length of stay was 48 hr (IQR 24–96 hr) for dogs who arrested, 24 hr (IQR 0–24 hr) for dogs who were euthanized, and 96 hr (IQR 72–120 hr) for dogs who survived to hospital discharge. There were 135 records that reported all four SIRS criteria variables (temperature, HR, RR, WBC). Of these, 126 (93%) met the requisite number of two or more SIRS criteria, indicating clinical suspicion for presence of sepsis.^27,33

Hematological and biochemistry profile results are shown in Supplementary Table 1. APPLE_fast scores were available for only 68/143 (48%) of patients. Twelve out of 32 nonsurvivors, and 30 out of 36 survivors had APPLE scores ≤25. SPI2 was 0.55 in nonsurvivors and 0.73 in survivors. SI was 1.71 in nonsurvivors and 1.36 in survivors.

Seventy-eight case records (54%) were complete for both lactate and albumin. In this subset, there were neither clinically nor statistically significant differences between nonsurvivors versus dogs who survived to discharge for lactate, albumin, and L:A ratio (Table 2). Pooled (n = 78) median values were 2.80 mmol/L (IQR 1.6–5.5 mmol/L; range 0.6 to 113 mmol/L) for lactate, and 2.6 g/dL (IQR 2.1–2.9 g/dL; range 1.1–4.1 g/dL) for albumin. Median L:A was 1.22 (IQR 0.69–1.88; range 0.25–8.56).

TABLE 2 Lactate, Albumin, and Lactate to Albumin Ratios (Unadjusted and Adjusted Means on Matched Cases, 95% Confidence Intervals) for Dogs Surviving to Hospital Discharge Versus Nonsurvivors (Cases Euthanized or In-Hospital Cardiac Arrest)

View Fullscreen

Patterns of missing data are shown in Supplementary Table 2. Missingness of lactate and albumin data was not associated with mortality category (χ² = 2.13, P = 0.14) but was strongly associated with whether or not hematology (χ² = 20.75, P < 0.0001) and biochemistry panels (χ² = 42.42, P < 0.0001) were obtained at all. Missingness of SBP was unrelated to mortality category (χ² = 2.41, P = 0.12) but was strongly associated with mentation status (χ² = 17.09, P < 0.0001), with dogs with MENT ≥2 (somewhat to severely obtunded) three to four times more likely to have SBP recorded.

Discussion

In this retrospective case-control study of septic dogs, L:A was not found to be a clinically useful marker to predict mortality in dogs diagnosed with sepsis. Lactate and albumin concentrations and the L:A ratio did not differ statistically or clinically between survivors and nonsurvivors. In human medicine, lactate has been used as a prognostic marker in patients with sepsis, with increased lactate concentrations associated with higher mortality.^3,42,43 Similarly, in veterinary medicine, increased lactate concentration at admission in dogs with septic peritonitis is usually associated with a higher mortality, and, in one study, a lactate >4 mmol/L was 92% specific for nonsurvival.⁴⁴ However it has been suggested that lactate clearance, rather than absolute lactate levels may be a better predictor of survival.^19,44-47 One pitfall of lactate or lactate clearance as a prognostic marker in septic patients is that these processes are not specific for sepsis but can occur in any state of shock.^3,48

There is very little data on albumin as a prognostic indicator in sepsis in veterinary medicine. One study in dogs with septic peritonitis demonstrated that higher serum concentrations of albumin before surgery were associated with better survival.⁴ A different study in septic dogs (both surgical and nonsurgical sepsis) showed no association between albumin concentration and survival. This could be due to the patient being admitted at different stages of disease progression.²⁹

The L:A has been proposed in human medicine as a more reliable prognostic marker in sepsis than either lactate or albumin alone. L:A was a better predictor of mortality for human adult and pediatric intensive care patients compared to lactate or albumin alone, with a higher L:A for nonsurvivors.^49,50 However L:A apparently had little prognostic value for adult septic shock patients seen in an emergency admissions.¹⁵ One possible reason for inconsistencies with L:A is that lactate and albumin follow different clinical trajectories and timelines and therefore changes in each marker may be reflected at different time points.

Limitations of this study include the potential for selection bias, spectrum bias, and the number of records missing important clinical data. Sampled patient data were based on availability and completeness of records at our institution, a tertiary referral center, which may not be representative of the canine sepsis population as a whole. Although all patients were classified as “septic” in their medical record, patients presented in different degrees of acuity, and six patients met less than two SIRS criteria. Using SIRS to identify septic patients can be prone to error because it is neither specific nor sensitive for diagnosing sepsis.^51,52 Because there is no gold standard for the diagnosis of sepsis in veterinary patients, this study relied on the clinical diagnosis within the medical record. This may have reduced the number of patients in this study if the attending veterinarian did not use the prespecified diagnostic codes in the medical record. A further limitation of this study (common to all observational case record reviews) is the lack of consistency in clinical record documentation of vital signs and laboratory readings at patient presentation. These omissions meant that we were unable to calculate disease severity scores (APPLE) and SIRS criteria for a large subset of eligible records. Values of SI and SPI2 cannot be safely interpreted, because there were too few complete records (especially for SBP and laboratory readouts) for valid index calculations. Missingness of specific variables did not appear to be related to patient survival but was probably associated more with decisions of individual clinicians. This indicates that standardization of practice with respect to triage documentation, diagnostic tests, and laboratory work for all patients presenting for emergency admission is important because it can improve triaging and clinical decision-making, minimize missed diagnoses, and improve observational research.

Conclusion

This observational case record review found no clinically or statistically significant differences in lactate, albumin, or L:A between survivors and nonsurvivors of dogs diagnosed with sepsis. Larger multicenter studies may be needed to explore the viability of these as biomarkers in sepsis. Prospective studies could be useful to assess the clinical value of L:A as a predictor for mortality in septic intensive care unit patients over time.