Retrospective investigation of the neutrophil-to-lymphocyte ratio in dogs with pneumonia: 49 cases (2011–2016)

Objective: To assess the utility of the neutrophil-to-lymphocyte ratio (NLR) in predicting outcome in canine pneumonia compared with routine hematological parameters and systemic inflammatory response syndrome (SIRS) scores. Design: Retrospective study. Setting: University teaching hospital. Animals: Forty-nine client-owned dogs. Interventions: None Measurements and Main Results: Medical records were reviewed to identify dogs with a diagnosis of pneumonia from July 2011 to December 2016. Signalment, clinical without SIRS. Receiver operating characteristic analysis identified a %-bands cutoff of 2.5% or higher had an 83% sensitivity and 79% specificity for nonsurvival. Conclusions: Unlike in human medicine, neither NLR nor SIRS scores predicted outcome in this cohort of dogs with pneumonia. However, survivors had a lower %-bands and higher lymphocyte count than nonsurvivors, which may be helpful prognostically in clinical cases.

without SIRS. Receiver operating characteristic analysis identified a %-bands cutoff of 2.5% or higher had an 83% sensitivity and 79% specificity for nonsurvival.
Conclusions: Unlike in human medicine, neither NLR nor SIRS scores predicted outcome in this cohort of dogs with pneumonia. However, survivors had a lower %-bands and higher lymphocyte count than nonsurvivors, which may be helpful prognostically in clinical cases.

K E Y W O R D S
aspiration pneumonia, NLR, prognostic factors, SIRS INTRODUCTION Pneumonia is a common cause of respiratory disease in dogs and has a variety of etiologies, including infectious pneumonia and aspiration pneumonia. In both human and canine medicine, pneumonia can vary in severity but is a common cause of sepsis and the systemic inflammatory response syndrome (SIRS). Reported mortality rates for pneumonia in veterinary medicine are approximately 20%, 1,2 whereas in people the mortality rate varies widely, from as low as 1% to as high as 59% depending on the patient population studied. 3,4 In human medicine, there are a number of clinical scoring systems to stratify patients with pneumonia by mortality risk, including the British Thoracic Society Score, North American Pneumonia Severity Index, and the CURB-65 score. 3,5,6 Biomarkers such as C-reactive protein (CRP) also have some prognostic use in various subsets of pneumonia patients. Although there are a number of veterinary scoring systems for predicting prognosis in acute ICU admissions such as the acute patient physiologic and laboratory evaluation (APPLE) score and SIRS score, these are diagnosis independent scores. 7,8 There are few studies looking at prognostic indicators in canine pneumonia and no pneumonia-specific scores. In one study, the number of lung lobes affected radiographically was significantly higher in nonsurvivors than survivors, but no other clinicopathological differences were reported. 1 Although CRP at presentation has some use in differentiating bronchopneumonia from other respiratory diseases, 9 in a subsequent study it had no correlation with duration of hospitalization or other markers of disease severity; however, its association with survival was not evaluated. 10 In populations of dogs with SIRS resulting from mixed underlying conditions, CRP was not prognostic for survival. 11,12 In human medicine, there has been recent interest in developing less expensive and more easily accessible biomarkers to help predict prognosis in individuals with pneumonia. Studies have shown the neutrophil-to-lymphocyte ratio (NLR), calculated from the standard hematological peripheral blood total neutrophil and lymphocyte counts, is as good a predictor of severity and outcome as the Pneumonia Severity Index and CURB-65 scores and a better predictor than some classical biomarkers such as CRP. 13,14 In veterinary medicine, the NLR has diagnostic and prognostic relevance in various canine tumors. 15,16 And although the NLR is not associated with duration of hospitalization, APPLE scores, or mortality in dogs with septic peritonitis, 17 no studies have evaluated its utility in canine pneumonia.
The aims of this study were to assess whether the NLR is a useful prognostic marker in canine pneumonia and to compare the prognostic value of the NLR with other routine hematological parameters, CRP and SIRS scores.

MATERIALS AND METHODS
A retrospective study of canine pneumonia cases between July 2011 and December 2016 was performed by searching the hospital imaging records for the keyword "pneumonia" as a differential diagnosis.
Full case records were then reviewed for suitability. to determine the reason for euthanasia. These were divided into the categories of "severity of pneumonia," "cost," "long-term prognosis," "not recorded," or a combination of those categories. Cases were classified as being euthanized due to severity of pneumonia alone if the notes recorded worsening of tachypnea, dyspnea, respiratory effort, pulse oximetry readings <90%, or cyanosis, plus recorded recommendation of euthanasia by the case clinician, with no other records of concerns over cost or long-term prognosis. Otherwise, cases were classified as having a combination of reasons for euthanasia.
SIRS scores at admission were calculated for all cases for which there were sufficient data available. Dogs were classified as having SIRS if they fulfilled 2 or more of the following criteria: body temperature <38.1 • C or >39.2 • C; heart rate >120/min; respiratory rate >20/min; total WBC count <6.0 × 10 9 /L (<6 × 10 3 /µL) or >16 × 10 9 /L (>16 × 10 3 /µL); and percentage of bands >3% of the total WBC count, as per the previously reported criteria. 8 Descriptive statistics for total WBC count, total neutrophil count (segmented and band neutrophils), lymphocyte count, band neutrophil count, total neutrophil percent of total WBC count, band neutrophil percent of total WBC count (%-bands), lymphocyte percent of total WBC count, NLR, and SIRS score were generated and compared between dogs that did and did not survive to discharge. Only dogs that either died naturally or were euthanized solely due to the severity of pneumonia were included in the nonsurvivors group.

Statistical methods
All data were tested for normality using a Shapiro-Wilk test. Nonparametric data were then compared between groups using a Mann-Whitney U-test. Receiver operating characteristic (ROC) curve analysis was used to identify cutoff levels for sensitivity and specificity for nonsurvival to discharge for those parameters showing a statistically significant difference between survivors and nonsurvivors. The proportion of dogs with SIRS was compared between survivors and nonsurvivors using a chi-squared test. The data were also evaluated for correlations with hospitalization time in the survivors using Spearman's rank correlation. A P-value < 0.05 was considered statistically significant. All statistical analyses were performed using commercially available statistical software. †

RESULTS
Two hundred records were retrieved, of which 49 cases fulfilled the inclusion criteria. The median age was 5 years (range, 2 wk to 13 y).
Six dogs were mixed breeds, and 43 were pure breeds, with 28 different breeds represented. Weimaraners (n = 4) were the most common pure breed, followed by Pugs and Great Danes (n = 3), then Border Collies, Boston Terriers, Boxers, Bulldogs, Scottish Terriers, Shih Tzus, Staffordshire Bull Terriers, and Yorkshire Terriers (all n = 2). The median hospitalization time for all dogs was 3.5 days, ranging from 0 to 25 days. Of the final diagnoses, 36 of the 49 patients had a diagnosis of "aspiration pneumonia," whereas the remainder were diagnosed with "pneumonia" or "bronchopneumonia." Dogs with a final diagnosis of aspiration pneumonia had a mortality rate of 36%, compared to 23% for other pneumonia subtypes; this was not significantly different Thirty-four of the dogs had 1 or more concurrent conditions reported to predispose to aspiration pneumonia. 18  Of the entire group of 49 dogs, 33 (67%) survived to discharge. Of all the nonsurvivors, 7 died, whereas 9 were euthanized. Of those 9 euthanized, 5 could be retrospectively categorized as being euthanized due to the severity of pneumonia alone, 2 due to the poor long-term prognosis, and 2 due to a combination of the severity of pneumonia and the long-term prognosis. The 4 dogs euthanized for reasons other than pneumonia severity alone were excluded from the statistical analysis.
The median hematological parameters for the 45 dogs are shown in  Figure 1). There was also no significant difference between the groups for total WBC count, neutrophil count, total neutrophil percent of total WBC count, or lymphocyte percent of total WBC count. Nonsurvivors had a significantly higher

DISCUSSION
Unlike in human medicine, 13  Lymphocyte count (× 10 9 /L) (× 10 3 /µL) <0.74 75% 70% Lymphocyte count (× 10 9 /L) (× 10 3 /µL) <0.28 50% 94% Note: Sensitivity and specificity for nonsurvival for client-owned dogs with pneumonia for the band neutrophil percent of total WBC count at levels above 3 selected cutoffs, and for nonsurvival at levels below 3 selected cutoffs for total lymphocyte count based on ROC curve analysis. with a DLS at presentation to a veterinary teaching hospital, pneumonia was the most common underlying disease, diagnosed in 19% of cases. Notably, however, when the significance of a DLS was analyzed in individual diseases, there was no difference in survival between pneumonia patients with and without a DLS. 19 In the current study, no dogs showed a DLS; however, there was a significantly higher %-bands in the nonsurvivors than in the survivors. Based on ROC curve analysis, the %-bands also showed a good prognostic value for survival (AUC = 0.87). A %-bands cutoff of 2.5% or greater had 83% sensitivity and 79% specificity for nonsurvival to discharge. Although in the DLS study, the severity of the left shift was assessed using the ratio of mature to immature neutrophils, 19 unlike in the current study, this ratio does not take into account the total number of WBC, so whether a similar association existed between increasing proportion of bands and survival to that found here is unknown. Equally, given involvement of a greater number of lung lobes is associated with worse prognosis in canine pneumonia, 1 it is possible that respiratory compromise from impaired gas exchange leads to death in some dogs with pneumonia before marrow neutrophil exhaustion is reached.
A recent prospective study compared various WBC parameters with some of the newer acute phase biomarkers in human communityacquired pneumonia (CAP) patients. They found both a higher NLR in patients dying within 30 days compared to survivors, and a higher neutrophil count percentage and lower lymphocyte count percentage, but no statistical difference in the absolute neutrophil or absolute lymphocyte counts between the groups. 20 Similar to the current study's finding of the %-bands having the higher ROC AUC, however, the human CAP study found the neutrophil count percentage had a higher ROC AUC than the NLR (0.87 vs. 0.79), and in multivariate analysis only the admission neutrophil count percentage retained statistical significance for mortality. 20 They did not report the percentage of bands, however, so whether this parameter may be similarly significant in people, as in the canine population in the current study, is unknown. 20 Overall, it is possible that differences in WBC kinetics between people and dogs contributed to this differing significance of absolute cell counts compared to count percentages.
Reasons for the difference in the NLR's significance between people and dogs with pneumonia are likely multifactorial but may in part be due to the option of euthanasia in veterinary patients. As with all veterinary studies investigating predictors of survival, euthanasia of patients due to costs, perceived poor prognosis, or other owner-or case-specific factors is likely to have a significant confounding effect. In the current study, attempts were made to address this by only including cases where euthanasia was recommended due to the severity of the pneumonia and an actively deteriorating clinical picture such that death was considered inevitable. However, it is important to note that this retrospective classification of reasons for euthanasia has significant limitations, in that only information recorded in the case records can be evaluated, and subtle impacts of factors such as cost may have played a role in either the veterinarian's or owners' decision to euthanize a patient, even if they were not verbalized or formally recorded in the records. Although the difference in the median NLR between survivors and nonsurvivors was not statistically significant, it was similar in magnitude to that seen in larger human studies evaluating the NLR. 13,20 As such, it is possible that inclusion of a larger number of dogs may have detected a significant difference, and the present study has not because of a type II error. Future prospective studies that are larger in size and that exclude all euthanized patients may be warranted to further assess the utility of the NLR.
Differences in pneumonia type may also play a role in the different significance of the NLR between human and canine patients. Although in human studies the reported proportion of patients with CAP with the subtype of aspiration pneumonia varies between 7% and 24%, 4,21 in the current study, 73% of dogs had a final recorded diagnosis of aspiration pneumonia. Despite many instances of aspiration involving true bacterial colonization of the lower respiratory tract, either from aspiration of large particulate gastric contents or secondary opportunistic infection, a proportion of cases are better described as a pneumonitis due to chemical injury and sterile inflammation within the airways. 22 As many pneumonia patients are insufficiently stable for anesthesia and sampling of the lower airways for bacterial isolation, combined with the low positive culture rates seen in true bacterial pneumonia, it is difficult to achieve a definitive differentiation between pneumonitis and bacterial pneumonia in clinical cases, both in human and veterinary medicine. 1,4,23,24 In human medicine, although the NLR has a good positive predictive value for bacteremia in patients with a variety of conditions presenting for emergency care, 25 routine WBC counts have not been shown to differ between ventilated patients with confirmed pneumonitis and those with bacterial pneumonia. 26 Whether the NLR may differ between canine patients with bacteremia or sepsis compared to those with nonseptic inflammation is unknown. A recent meta-analysis of mortality risk in human aspiration patients compared to other CAP patients found that those with a risk factor for aspiration and, therefore, presumed aspiration pneumonia, had a more than 3-fold increased risk of in-hospital and 30-day mortality. 4 In the current study, dogs with a final diagnosis of aspiration pneumonia had a slightly higher mortality rate than those with other pneumonia subtypes, although this did not reach statistical significance. Thus, it is possible that relatively high mortality rates in the current study due to overrepresentation of aspiration cases affected the prognostic use of the NLR. Future investigation into whether there is a survival difference between canine pneumonia patients with known aspiration compared to those with bacterial pneumonia secondary to other etiologies, and whether the NLR has any prognostic use in the latter population, would be useful. were not available for most cases in the current study.
To the authors' knowledge, there are no previous studies evaluating the prognostic value of WBC percentages in canine pneumonia. The current study found that although the NLR was not useful prognostically, the number of band neutrophils as a percentage of the total WBC count (%-bands) was significantly higher in nonsurvivors than survivors and showed a weak correlation with longer duration of hospitalization in survivors. Although a %-bands of greater than 2.5% had a specificity of 79% for death in hospital, further studies evaluating this in a larger population of dogs presenting to both referral and primary care practices would be of interest. Additionally, larger studies including additional clinicopathological variables such as biochemical data and CRP may allow development of a better performing multiparameter disease-specific scoring system for canine pneumonia, similar to those used in human medicine.