J Vet Intern Med 2006;20:20 31 Toxic Neutrophils in Cats: Clinical and Clinicopathologic Features, and Disease Prevalence and Outcome A Retrospective Case Control Study Gilad Segev, Eyal Klement, and Itamar Aroch Toxic neutrophils exhibit a variety of nuclear and cytoplasmic abnormalities in Romanowsky-stained blood smears, and are associated with inflammation and infection. The purpose of the retrospective study reported here was to investigate the association of toxic neutrophils with clinicopathologic characteristics, diseases, and prognosis in cats. Cats with toxic neutrophils (n 5 150) were compared with negative-control cats (n 5 150). Statistical analyses included Fisher exact, independent t-, nonparametric Mann-Whitney, and x 2 tests. Cats with toxic neutrophils had significantly (P,.05) higher prevalence of fever, icterus, vomiting, diarrhea, depression, dehydration, weakness, and cachexia, as well as leukocytosis, neutrophilia, left shift, neutropenia, anemia, hypokalemia, and hypocalcemia. The prevalence of shock, sepsis, panleukopenia, peritonitis, pneumonia, and upper respiratory tract diseases was significantly higher among these cats, as were infectious (viral and bacterial) and metabolic disorders. Control cats had a significantly higher prevalence of feline asthma, as well as allergic, idiopathic, and vascular disorders. Hospitalization duration and treatment cost were significantly (P,.001) higher in cats with toxic neutrophils. In 53 and 47% of the cats with toxic neutrophils, the leukocyte and neutrophil counts were normal, respectively, whereas in 43%, both abnormalities and left shift were absent, and toxic neutrophils were the only hematologic evidence of inflammation or infection. In conclusion, toxic neutrophils were found to be associated with certain clinicopathologic abnormalities, and when present, may aid in the diagnosis, as well as the assessment of hospitalization duration and cost. The evaluation of blood smears for toxic neutrophils provided useful clinical information. Key words: Hematology; Hospitalization duration; Inflammation; Leukocytes. The term toxic neutrophil refers to a neutrophil with certain specific morphologic abnormalities observed on examination of Romanowsky-stained peripheral blood smears. These changes occur during the maturation process in the bone marrow under certain conditions or in association with certain diseases. 1 Most of these changes are cytoplasmic, but nuclear changes and changes in cell size or shape also may occur. 2,3 Cytoplasmic changes are most prevalent and important, and include Döhle bodies, increased basophilia, toxic granulation, and vacuolation. 2,3 Döhle bodies are grayish-to-blue cytoplasmic inclusions that are the result of lamellar retention and aggregation of rough endoplasmatic reticulum. 4,5 Remnants of RNA and ribosomes lead to increased basophilia that appears as a bluish-gray to dark-blue cytoplasm, as opposed to the normal neutral-staining cytoplasm of the cell. Toxic granulation refers to the presence of azurophilic granules in the neutrophil s cytoplasm, and is attributed to acid mucopolysaccharide retention, and increased permeability of primary granules to Romanowsky stains. 4,5 Toxic granulation is uncommon in cats, but its prevalence is unknown, and it must be differentiated from the eosinophilic granules present in some Birman cats, granules in animals with certain lysosomal storage disorders, and other congenital or infectious disease inclusions. 6 Vacuolation appears as mildly reticulated From the School of Veterinary Medicine (Segev, Klement, Aroch); and the Section of Epidemiology (Klement), School of Veterinary Medicine, The Hebrew University of Jerusalem, Israel. Reprint requests: Itamar Aroch, School of Veterinary Medicine, The Hebrew University of Jerusalem, Rehovot 76100 Israel; e-mail: aroch@agri.huji.ac.il. Submitted November 30, 2004; Revised March 8, 2005; Accepted July 25, 2005. Copyright E 2006 by the American College of Veterinary Internal Medicine 0891-6640/06/2001-0003/$3.00/0 cytoplasm, with loss of granule uniformity up to an intensively foamy cytoplasm, and is thought to be a result of autodigestion or disruption of cell membrane integrity, or both. 2,3 Vacuolation may occur as a consequence of potassium ethylenediaminetetracetic acid (EDTA) storage artifact, but it rarely occurs in freshly obtained samples. Nuclear toxic changes include vacuolation, polyploidy, hyposegmentation and ring formation, karyorhexis, and karyolysis. Formation of giant neutrophils represents additional toxic change, and is a result of skipped cellular divisions during the maturation process. 2,3 Observation of toxic neutrophils in peripheral blood may precede changes in the leukogram. 7 Therefore, their presence may serve as an early sensitive indicator of disease, and aid in the prediction of disease course and outcome. Evaluation of toxic neutrophils is simple and cost effective; a Romanowsky staining solution and light microscope are the only necessary pieces of equipment. This procedure can be performed quickly in any facility, and the results may be obtained before a CBC is available. Evaluation of blood smears for toxic neutrophils potentially may provide the clinician with valuable information concerning the components of the differential diagnosis to be considered, as well as severity of the disease, and it may be useful in the assessment of prognosis. Evaluation of neutrophil morphology provides additional, complementary data to that of the CBC. The morphology of toxic neutrophils is well documented, under light and electron microscopy, 7,8 but was only described in several isolated feline case reports. 9 To our knowledge, studies of its clinical relevance and association with disease conditions in cats have not been published. The objective of the study reported here was to investigate the association of toxic neutrophils in cats with clinical findings and disease prevalence and outcome.
Feline Toxic Neutrophils 21 Materials and Methods Selection of Cases and Collection of Data Clinical records of cats presented to the Department of Small Animal Internal Medicine at the Hebrew University Veterinary Teaching Hospital (HUVTH) were reviewed retrospectively. Onehundred fifty cats with evidence of toxic neutrophils (eg, Döhle bodies, basophilia, vacuolation or foaminess, toxic granulation, giant toxic neutrophils) on examination of Mäy-Grünwald-Giemsa (MGG)-stained peripheral blood smears were included in the study (defined as cats with toxic neutrophils). These cats were compared with a randomly selected equal number of cats from the same period (the next cat within a 14-day interval after a cat with toxic neutrophils was presented), admitted to the Department of Small Animal Internal Medicine at the HUVTH that presented with no evidence of toxic neutrophils on examination of MGG-stained peripheral blood smears. The examination of all blood smears was performed by a single clinician (IA) before case selection for the study. The data collected from hospital records included signalment, history, physical examination findings, clinicopathologic findings, diagnosis, hospitalization duration, treatment cost, and 30-day survival. Nonsurvivor cats included those that died naturally or were euthanized within the hospitalization period or within 30 days from discharge. The total number of diagnoses in each group exceeded the number of cats, because a cat may have had more than 1 diagnosis during a single visit. Specific diseases were classified into the following categories: allergic, anatomic, degenerative, developmental, infectious-bacterial, infectious-viral, infectious-parasitic, inflammatory (presumed noninfectious), immune-mediated, idiopathic, iatrogenic, metabolic, neoplastic, nutritional, traumatic, toxic, and vascular. In all cats of this study, hypertrophic cardiomyopathy (HCM) and dilated cardiomyopathy (DCM) were classified as idiopathic, because the history excluded familial causes (for both diseases) or taurine deficiency (for DCM). Obstructive and nonobstructive feline lower urinary tract diseases (FLUTD) also were classified as idiopathic, when urinary tract infection (classified as infectious), and crystalluria or urolithiasis (both classified as metabolic) were excluded. Feline asthma was classified as allergic, whereas acute renal failure (ARF) was regarded as a metabolic disorder. High rise syndrome, bite wounds, head trauma, and road traffic car accident were classified as mechanical trauma. Laboratory Tests On admission, blood samples for a CBC were collected in potassium EDTA-containing tubes, and analyses were performed within 15 minutes of sample collection (to avoid EDTA storage artifacts) by automatic impedance cell analyzers, a,b calibrated for feline blood. Blood smears for differential leukocyte counting and morphologic evaluation were prepared within 30 minutes of sample collection. The differential count was obtained by manually counting 100 leukocytes in MGG-stained blood smears. Hematologic variables in the automated CBC included red blood cell (RBC) count, hemoglobin concentration, hematocrit, mean corpuscular volume (MCV), mean corpuscular hemoglobin (MCH), mean corpuscular hemoglobin concentration (MCHC), and white blood cell (WBC) count. Nucleated red blood cells (NRBC) were counted manually (as NRBC/100 WBC) when observed, and the WBC count was corrected for them. 10 Blood for biochemical analysis, when performed, was collected in plain tubes, and serum was separated by centrifugation of the sample within 1 hour after collection. Sera were stored at 4uC pending analysis, which was performed within 24 hours of sample collection by means of a wet chemistry autoanalyzer. c The measured biochemical variables included alanine transaminase Fig 1. Cytoplasmic toxic morphologic changes in feline neutrophils. (a) Toxic neutrophils from a cat with Mycoplasma hemofelis infection. Neutrophils contain Döhle bodies (solid arrows) and moderate cytoplasmic basophilia. Notice band neutrophil (open arrowheads) with mild cytoplasmic basophilia (solid arrowhead). (b) Large toxic neutrophil with marked cytoplasmic basophilia from the cat in a. (c) Neutrophil from a cat with acute pancreatitis, Notice ring-shaped nucleus and moderate cytoplasmic basophilia. (d) Two neutrophils with moderate cytoplasmic basophilia (open arrowheads), a large Döhle body (solid arrow), and a normal monocyte (asterisk) from the cat in a. (e) Neutrophils from the cat in a, with mild (solid arrow), moderate (open arrowhead), and marked (white arrowhead) cytoplasmic basophilia. The last was defined as a toxic giant band neutrophil. Giemsa stain; bars 5 10 mm (original magnification, 6003). (ALT), alkaline phosphatase (ALP), albumin, amylase, aspartate transaminase (AST), calcium, chloride, creatinine, creatine kinase (CK), c-glutamyltranferase (c-gt), glucose, lactate dehydrogenase (LDH), phosphorus, potassium, sodium, total protein (TP), total bilirubin, triglycerides, and urea. Ionized calcium and ionized magnesium concentrations were measured by an electrolyte analyzer. d Definition of Toxic Neutrophils The severity of toxic neutrophil presence was assessed subjectively and semiquantitatively on the basis of cytoplasmic toxic changes (eg, foaminess, basophilia, Döhle bodies, toxic granulation, giant toxic neutrophils) on examination of a single MGGstained blood smear for each case (Figs 1, 2). Each individual type of toxic change was assigned 1 of 3 final grade scores of morphologic abnormality: mild, moderate, or marked (scores 1, 2, and 3, respectively). The final grade of each type of toxic change was a combination of a quantitative assessment of the percentage of affected neutrophils (,10% 5 mild, 10 30% 5 moderate,.30% 5 marked), and a qualitative grade of the intensity of each of the individual toxic scores for each morphologic change (Table 1).
22 Segev, Klement, and Aroch Table 1. Grade of individual toxic changes in neutrophils. a Morphologic change intensity b Cells affected in peripheral blood smear (%),10 10 30.30 Döhle bodies Mild 1 1 1 Moderate 1 1 2 Marked 2 2 3 Cytoplasmic basophilia Mild 1 1 2 Moderate 2 2 3 Marked 2 3 3 Cytoplasmic vacuolation Mild 1 1 2 Moderate 2 2 3 Marked 2 3 3 Giant toxic neutrophils 3 3 3 a Overall toxic score is the sum of all individual toxic grades. b See text for details. Mild cytoplasmic basophilia was defined as presence of a nonuniform grayish to light-blue cytoplasm in the neutrophils (Figs 1a, 2b), whereas a uniformly light-blue cytoplasm was judged as a moderately abnormal change (Figs 1a, c, d, e). The presence of a uniformly blue to dark-blue cytoplasm was considered a marked abnormality (Figs 1a, b, e); presence of 1 2 small Döhle bodies/cell was judged as a mild abnormality (Fig 1a), whereas presence of 3 4 Döhle bodies was ranked as a moderate abnormality (Fig 2a); and a large prominent Döhle body (Figure 1d), more than 4 Döhle bodies per cell (Fig 2a), or both were judged as a marked abnormality. The presence of cytoplasmic vacuolation was classified into 3 categories. Mild vacuolation was defined as loss in cytoplasm clarity and neutral-stained granules, whereas a moderate change was defined as observation of small cytoplasmic vacuoles. Appearance of intense vacuolation with grayish reticulation was classified as severe cytoplasmic vacuolation. The presence of giant toxic neutrophils was considered a marked abnormality (Fig 1e). For assessment of overall neutrophil toxic change in a smear, the sum of all scores of the individual morphologic abnormalities in neutrophils was calculated. Mild overall neutrophil toxic change included a sum of scores of 1 6. Moderate toxic change included a total score of 7 12, whereas a total score.12 was classified as marked toxic change. Fig 2. Cytoplasmic toxic morphologic changes in feline neutrophils. (a) Toxic neutrophils from a cat with septicemia. Notice two (solid arrow) and multiple (solid arrowhead) Döhle bodies and mild cytoplasmic basophilia. (b) Toxic neutrophils from a cat with panleukopenia. Notice mild basophilia (solid arrowhead) and toxic granulation (open arrowheads). (c) Normal feline Statistical Analysis Normality of interval variables was assessed with P-P plots. Descriptive statistics are presented as median with interquartile range (IQR) for all variables. The prevalence for each variable was calculated separately within the study and the control groups. Prevalence ratio with its 95% confidence interval was calculated for each variable. Results for nominal variables from the 2 groups were compared by means of the x 2 (clinicopathologic data) and Fisher exact (all other data) tests. Interval variables were compared by an independent t-test (if they were normally distributed) or by the nonparametric Mann-Whitney test (if they were not normally r neutrophils. Giemsa stain; bars 5 10 mm (original magnification, 6003).
Feline Toxic Neutrophils 23 Table 2. Prevalence of selected clinical signs of disease in cats with toxic neutrophils and in controls. Clinical sign Toxic a Control b PR CI 95% P-value Tachypnea (.30 breaths/min) 80 88 0.91 0.74 1.11.42 Anorexia 49 35 1.40 0.97 2.03.09 Depression * 34 19 1.78 1.07 2.99.03 Dyspnea 30 31 0.97 0.62 1.51 1.00 Vomiting * 30 13 2.31 1.25 4.25.008 Hypothermia #37.5uC) 29 27 1.07 0.67 1.72.88 Weakness * 28 13 2.15 1.16 3.99.02 Dehydration * 24 10 2.4 1.19 4.84.02 Diarrhea * 21 9 2.33 1.11 4.93.03 Fever ($39.5uC) * 21 9 2.33 1.11 4.93.03 Pale mucus membranes 19 13 1.46 0.75 2.85.35 Salivation * 10 1 10.00 1.30 77.15.01 Icterus * 9 1 9.00 1.15 70.16.02 Increased lung sounds 9 6 1.50 0.55 4.11.60 Cachexia * 8 1 8.00 1.01 93.18.04 Skeletal fracture 8 2 4.00 0.86 18.53.10 Polyuria/polydipsia 8 11 0.73 0.30 1.76.64 Abdominal mass 7 3 2.33 0.61 8.85.34 Cough 7 5 1.40 0.45 4.31.77 Abdominal pain 6 2 3.00 0.62 14.63.28 Constipation 5 3 1.67 0.41 6.85.73 Lymphadenopathy 5 2 2.50 0.49 12.69.45 Hematuria 4 5 0.80 0.22 2.92 1.00 Stranguria 3 8 0.38 0.10 1.39.22 Heart murmur 2 6 0.33 0.07 1.63.28 Dysuria * 0 12 NA NA,.001 PR, Prevalence ratio; CI 95%, 95% confidence interval; NA, not applicable. a Number of cats with toxic neutrophils. b Number of control cats. * Significant difference by the Fisher exact test (P,.05). distributed). Differences in hospitalization duration were evaluated for all surviving cats in both groups by survival analysis, and were tested for statistical significance by the Gehan test. 11 The same comparison was performed separately for each specific diagnosis, when at least 5 surviving cats in each group were present (eg, enteritis, feline immunodeficiency virus infection, mechanical trauma, pneumonia). Multivariate analysis was performed with the Cox proportional hazards model. This model was applied to control for disease effect on hospitalization duration. For all tests applied, differences were considered statistically significant when P #.05. Statistical analysis was performed by means of 2 statistical software programs. e,f Results Differences in age and sex between cats with toxic neutrophils and control cats were not found. Differences were not found in mean body temperature, pulse, or respiratory rate between groups (38.3uC versus 38.1uC, 175 beats/min [bpm] versus 176 bpm, and 48 bpm versus 51 bpm, respectively). However, cats with toxic neutrophils had significantly (P,.05) higher prevalence of increased body temperature (.39.5uC, 17.2% versus 8.3%). Depression, dehydration, weakness, cachexia, diarrhea, icterus, salivation, and vomiting were significantly (P,.05) more prevalent in cats with toxic neutrophils. Dysuria was the only clinical sign of disease that was significantly more prevalent in the control group (Table 2). Cats with toxic neutrophils had significantly (P,.05) lower RBC count, hemoglobin concentration, and hematocrit (6.82 3 10 6 /ml versus 7.72 3 10 6 /ml, 10.45 g/dl versus 11.75 g/dl, and 31.7% versus 35.2%, respectively), and higher prevalence of anemia (hematocrit,24%, 22.8% versus 7.7%, P 5.0353 [Table 3]). Cats with toxic neutrophils also had significantly (P,.05) higher prevalence of macrocytosis (MCV.55 fl, 6.71% versus 1.3%, P,.001), although there was no significant difference in mean MCV between groups. Cats with toxic neutrophils had significantly (P,.05) higher WBC and absolute neutrophil counts (18.52 3 10 3 /ml versus 13.16 3 10 3 /ml, and 14.08 3 10 3 /ml versus 9.85 3 10 3 /ml, respectively). The prevalence of neutrophilia and neutropenia also was significantly (P 5.001) higher in cats with toxic neutrophils, compared with controls (45.0% versus 25.5% and 10.1% versus 2.7%, respectively). Normal leukocyte and neutrophil counts were observed in 53 and 47% of cats with toxic neutrophils, respectively, whereas leukocytosis with neutrophilia and left shift were absent in 43% of cats with toxic neutrophils. In 45% of affected cats, leukocytosis and left shift or neutrophilia and left shift were absent. Mean absolute band neutrophil counts tended to be higher in cats with toxic neutrophils, but did not achieve statistical significance (P 5.082), and prevalence of left shift was significantly higher in those cats (28.9% versus 7.4%, P,.001).
24 Segev, Klement, and Aroch Table 3. CBC results for cats with toxic neutrophils and for controls. Cats with toxic neutrophils Control cats Parameter Median IQR %,RI %.RI Median IQR %,RI %.RI RI WBC count (10 3 /ml) * 14.0 14.1 8.05 38.9 11.5 7.0 2.0 22.8 5.0 16.0 Corrected WBC count (10 3 /ml) * 13.5 13.3 8.05 38.3 11.4 6.8 2.0 22.8 5.0 16.0 RBC count (10 6 /ml) * 7.05 3.09 22.8 6.0 7.67 2.45 4.1 13.5 5.0 10.0 Hemoglobin (g/dl) * 10.6 5.1 27.5 8.1 11.8 3.9 8.7 13.4 8.0 15.0 Hematocrit (%) * 33.1 12.7 22.8 7.4 35.4 9.4 6.7 10.7 24.0 45.0 MCV (fl) 47.0 10.0 2.01 6.7 46.0 5.0 0.0 1.3 39.0 55.0 MCH (pg) 15.2 2.9 6.7 30.9 15.2 2.5 6.7 11.4 12.5 17.5 MCHC (g/dl) 32.3 3.5 20.1 10.1 32.5 3.0 12.2 13.5 30.0 36.0 Neutrophils (10 3 /ml) * 10.07 12.21 10.1 45.0 8.33 6.41 2.7 25.5 3.0 11.5 Band neutrophils (10 3 /ml) 0.00 0.50 0.0 28.9 0.00 0.00 0.0 2.0 0.0 0.3 Nucleated RBC (10 3 /ml) 0.00 4.00 0.0 16.0 0.00 0.00 0.0 9.3 0.0 0.0 Monocytes (10 3 /ml) 0.44 0.56 12.8 8.1 0.39 0.43 17.4 3.4 0.15 1.3 Lymphocytes (10 3 /ml) 1.19 1.45 38.9 6.0 1.58 1.86 30.2 6.7 1.0 4.8 Eosinophils (10 3 /ml) 0.18 0.60 38.9 10.7 0.40 0.81 24.8 12.1 0.10 1.25 Basophils (10 3 /ml) 0.00 0.00 0.0 0.0 0.00 0.00 0.0 0.00 0.00 0.24 RI, reference interval; IQR, interquartile range; RBC, red blood cell; MCV, mean cell volume; MCH, mean cell hemoglobin; MCHC, mean cell hemoglobin concentration. * Significant difference by the x 2 test (P,.05). Cats with toxic neutrophils had significantly higher mean serum activity of AST and significantly (P,.05) lower mean serum activity of ALP, as well as significantly lower mean concentration of total calcium and potassium. Compared with control cats, those with toxic neutrophils tended to have higher mean serum CK and LDH activities, but the differences did not achieve statistical significance (P 5.056 and P 5.071, respectively). Mean sodium concentration tended to be lower in cats with toxic neutrophils, but the value did not achieve statistical significance (P 5.072 [Table 4]). Cats with toxic neutrophils also had a significantly (P,.05) higher prevalence of hyperbilirubinemia (P 5.002), hypocalcemia (P 5.038), hypokalemia (P 5.012), and Table 4. Serum biochemical values in cats with toxic neutrophils and in controls. Cats with neutrophil toxicity Control cats Parameter n Median IQR %,RI %.RI n Median IQR %,RI %.RI RI Albumin (g/dl) 45 3.10 0.75 17.8 0.0 42 3.10 0.53 7.1 0.0 2.6 4.0 ALP (U/L) * 44 28 35 9.1 0.0 42 44 44 9.5 4.8 13 140 ALT (U/L) 73 45 51 2.7 42.5 69 49 56 1.4 40.4 10 50 Amylase (U/L) 43 608 399 11.6 23.3 41 607 403 4.9 31.7 340 800 AST (U/L) * 45 45 63.5 4.4 44.4 42 30 39 2.4 12.9 14 50 Chloride (meq/l) 38 113.4 9.1 2.6 28.9 38 115.2 5.6 5.3 31.6 102 117 Cholesterol (mg/dl) 44 127 93 38.6 15.9 42 153 87 23.8 4.8 120 260 Creatine kinase (U/L) 42 476 1339 2.4 81.0 42 218 540 0.0 66.7 13 100 Creatinine (mg/dl) 76 1.29 1.7 2.7 37.3 80 1.36 1.0 0.0 38.7 0.5 1.6 Total calcium (mg/dl) * 43 9.2 1.2 25.6 2.3 42 9.7 1.1 7.1 2.4 8.7 11.8 c-gt (U/L) 44 1.8 1.8 22.7 0.0 41 1.7 1.9 29.3 0.3 1.0 10.0 Globulin (g/dl) 44 4.3 1.0 0.0 79.5 41 3.8 1.2 4.8 64.3 1.9 3.5 Glucose (mg/dl) 37 122.0 60 8.1 73.0 38 143.0 98 5.3 65.8 70 110 LDH (U/L) 43 738 918 0.0 82.1 41 516 634 0.0 61.0 34 360 Potassium (meq/l) * 69 3.91 0.93 42.9 1.4 62 4.24 0.91 22.6 11.3 3.8 5.6 Total protein (g/dl) 44 7.3 1.6 2.3 40.9 41 7.0 1.4 4.9 29.3 5.5 7.5 Phosphate (mg/l) 42 4.10 7.00 16.7 14.3 38 4.72 6.00 2.6 21.1 2.5 6.2 Sodium (meq/l) 66 148.0 9.7 9.0 28.4 58 151.60 5.57 3.4 24.1 140.0 154.0 Total bilirubin (mg/dl) 47 0.39 0.92 2.1 40.4 39 0.24 0.18 2.6 10.3 0.10 0.60 Triglycerides (mg/dl) 38 86 110 13.2 31.6 35 57 68 14.3 25.7 40 100 Urea (mg/dl) 86 55.3 59.2 1.2 71.4 90 57.6 56.0 2.2 80.0 21 40 Ionized calcium (mmol/l) 43 1.17 0.18 30.2 2.3 30 1.16 0.16 26.7 3.3 1.1 1.4 Ionized magnesium (mmol/l) 41 0.54 0.17 9.8 31.7 30 0.52 0.18 13.3 33.3 0.4 0.6 IQR, Interquartile range; RI, reference interval; ALP, alkaline phosphatase; ALT, alanine transaminase; ASP, aspartate transaminase; c- GT, c-glutamyltransferase; LDH, lactate dehydrogenase. * Significant difference by the x 2 test (P,.05).
Feline Toxic Neutrophils 25 Table 5. Prevalence of selected diagnoses in cats with toxic neutrophils and in controls. Diagnosis Toxic a Control b PR CI 95% P-value Mechanical trauma 19 23 0.83 0.47 1.45.63 Pneumonia * 16 6 2.67 1.07 6.63.04 Chronic renal failure 14 13 1.08 0.52 2.21 1.00 Enteritis 12 8 1.50 0.63 3.56.49 Skeletal fracture 12 9 1.33 0.58 3.07.65 Feline immunodeficiency virus infection 11 7 1.57 0.63 3.94.47 Hypertrophic cardiomyopathy 10 12 0.83 0.37 1.87.82 Upper respiratory tract disease * 10 1 10.00 1.30 77.15.01 Bite wounds 7 5 1.40 0.45 4.31.77 Hepatic lipidosis 7 2 3.50 0.74 16.57.17 Diabetic ketoacidosis 6 2 3.00 0.62 14.63.28 Gastritis 6 3 2.00 0.51 7.85.50 Gingivitis 6 6 1.00 0.33 3.03 1.00 Intestinal foreign body 6 2 3.00 0.62 14.63.28 Peritonitis * 6 0 NA NA.03 Pleural effusion 6 1 6.00 0.73 49.24.12 Acute renal failure 5 1 5.00 0.59 42.29.21 Congestive heart failure 5 9 0.56 0.19 1.62.41 Lymphoma 5 3 1.67 0.41 6.85.72 Panleukopenia 5 0 NA NA.06 Sepsis 5 0 NA NA.06 Shock 5 0 NA NA.06 Nonobstructive FLUTD 4 9 0.44 0.14 1.41.26 Obstructive FLUTD 4 12 0.33 0.11 1.01.07 Pancreatitis 4 0 NA NA.12 Pneumothorax 4 4 1.00 0.25 3.92 1.00 Poisoning 4 3 1.33 0.30 5.86 1.00 Eosinophilic granuloma 3 4 0.75 0.17 3.29 1.00 Diabetes mellitus 0 5 NA NA.06 Feline asthma * 0 6 NA NA.03 Lung contusion 0 5 NA NA.06 Thromboembolism 0 5 NA NA.06 PR, prevalence ratio; CI 95%, confidence interval; NA, not applicable; FLUTD, feline lower urinary tract disease. a Number of cats with toxic neutrophils. b Number of control cats. * Significant difference by the Fisher exact test (P,.05). increased LDH activity (P 5.038), whereas control cats had a higher prevalence of hyperkalemia (P 5.019). Cats with toxic neutrophils had a significantly (P,.05) higher prevalence of pneumonia, sepsis, shock, and upper respiratory tract infections. Panleukopenia tended (P 5.06) to be more prevalent in cats with toxic neutrophils, but the value did not achieve statistical significance. Controls had a significantly (P 5.03) higher prevalence of feline asthma, and tended to have higher prevalence of thromboembolism and lung contusions, but this difference did not achieve statistical significance (P 5.06 [Table 5]). Cats with toxic neutrophils had a significantly higher prevalence of metabolic disorders (P 5.05) and bacterial (P,.001) and viral (P 5.003) infections, whereas controls had a significantly (P #.05) higher prevalence of allergic, idiopathic, and vascular diseases (Table 6). Differences in mortality were not found between cats with toxic neutrophils and controls (20.0% versus 15.3%, respectively, P 5.29). However, there was a significant (P,.001) difference between cats with toxic neutrophils and controls in median hospitalization duration for surviving cats (3.0 days versus 1.1 days, respectively, Fig 3), and treatment cost ($357.1 versus $252.8, respectively). This difference in hospitalization duration also was highly significant when controlled for disease type by means of the Cox proportional hazards model (adjusted hazards ratio 5.32, P,.001). Comparison of hospitalization duration between cats with toxic neutrophils and controls also was done for specific diseases, if the number of cases in each group was $5. These diseases included pneumonia (median 2.75 days versus 0.83 days, respectively, P 5.011), enteritis (median 3.00 days versus 0.70 days, respectively, P 5.002), feline immunodeficiency virus infection (median 3.17 days versus 0.75 days, respectively, P 5.013), and mechanical trauma (median 3.33 days versus 1.25 days, respectively, P,.001). Differences in mortality and treatment cost between cats with moderate-to-marked (total toxic score.6), compared with mild (total toxic score,6) neutrophil toxic changes, were not found. However, significantly (P 5.007) longer hospitalization duration was found in cats with moderate-to-marked (total toxic score.6), compared with mild (total toxic score,6) neutrophil toxic changes (median toxic score, 3.75 versus 2.61).
26 Segev, Klement, and Aroch Table 6. Disease category prevalence in cats with toxic neutrophils and in controls. Disease category Toxic (n) a Control (n) b PR CI 95% P-value Allergic * 3 11 0.27 0.08 0.96.05 Degenerative 13 7 1.86 0.76 4.52.25 Developmental 2 3 0.67 0.11 3.93 1.00 Metabolic * 20 9 2.22 1.05 4.72.05 Neoplasia 10 10 1.00 0.43 2.33 1.00 Nutritional 0 4 NA NA.12 Infectious bacterial * 67 30 2.23 1.55 3.22,.001 Infectious viral * 28 10 2.80 1.41 5.56.003 Infectious parasitic 4 5 0.80 0.22 2.92 1.00 Inflammatory 14 11 1.27 0.60 2.71.67 Immune 0 0 NA NA 1.00 Idiopathic * 17 35 0.49 0.28 0.83.009 Iatrogenic 0 1 NA NA 1.00 Traumatic 27 27 1.00 0.62 1.62 1.00 Anatomic 9 7 1.29 0.49 3.36.80 Toxic 5 4 1.25 0.34 4.56 1.00 Vascular * 0 6 NA NA.03 PR, prevalence ratio; CI 95%, 95% confidence interval; NA, not applicable. a Cats with toxic neutrophils. b control cats. * Significant difference by the Fisher exact test (P,.05). Discussion Fig 3. Comparison of hospitalization duration between cats with toxic neutrophils ( ) and controls (- - - - -). Rate of cats discharge from the hospital is depicted as a function of days after admission. Observation of toxic neutrophils has been a well known phenomenon in humans and animals, and its morphologic characteristics are well described. 2,3,7,8 However, its association with specific disease conditions has been described mostly in dogs, in a limited number of clinical case reports or series, and in animals with experimentally induced inflammation. 8,12 18 Recently, toxic neutrophils in dogs were investigated in a large retrospective study. 20 In recent reports, toxic neutrophils were associated with chloramphenicol toxicosis and tularemia in cats. 9,15 Toxic neutrophils have been associated with systemic, rather than localized infection and inflammatory conditions. 2,3 The results of the study reported here support this observation, because cats with toxic neutrophils had a significantly (P,.05) higher prevalence of systemic infectious conditions, such as peritonitis and pneumonia, as well as a higher tendency, however insignificant (P 5.06), toward panleukopenia and sepsis, compared with controls. Although some of these diseases may sometimes be localized, in most such cases admitted to the HUVTH, the animals are systemically ill, requiring intensive care and hospitalization. Some of these diseases are characterized by high tissue demands for neutrophils attributable to inflammation, leading to accelerated neutrophil production and maturation in the bone marrow, which may predispose the cells to disturbances and manifestation of toxic changes. 19 Alternatively, inflammatory mediators and cytokines may influence the neutrophil maturation process in the bone marrow, resulting in toxic changes. Cats with toxic neutrophils had more severe clinical signs of disease and a wider variety of signs on presentation, compared with controls, probably because the former had more severe and systemic disorders. This was manifested by a significantly (P,.05) higher prevalence of cachexia, dehydration, depression, diarrhea, increased body temperature, icterus, salivation, vomiting, weakness, and signs of shock (Table 2). Increased body temperature and icterus were also significantly more prevalent in dogs with toxic neutrophils, compared with controls (Table 7). 20 Compared with controls, cats with toxic neutrophils had a significantly (P 5.01) higher prevalence of icterus. It is likely that some of the same disorders that led to icterus also induced formation of toxic neutophils, but the possibility that increased bilirubin concentration actually is a cause of toxic neutrophil formation cannot be ruled out. Bilirubin, especially in the nonconjugated
Feline Toxic Neutrophils 27 Fig 4. Comparison of hospitalization duration for specific disease entities between cats with toxic neutrophils ( ) and controls (- - - - -). form, may lead to cellular disturbances in various body systems. 21 24 Although there was a significantly higher prevalence of hyperbilirubinemia (P 5.002) and clinical icterus (P 5.01) in cats with toxic neutrophils, mean total bilirubin concentration only tended to be different between groups (P 5.07), probably because hyperbilirubinemia in cats with toxic neutrophils was not severe (mean, 1.05 mg/dl; maximum, 8.76 mg/dl; reference interval, 0.1 0.6 mg/dl). The traditional observation that toxic neutrophils are associated with infection and inflammation 2 is further supported by the hematologic results of the study reported here. Cats with toxic neutrophils had a significantly higher prevalence of most hematologic markers of inflammation, such as leukocytosis, neutrophilia, neutropenia, and left shift (Table 3), probably because of a significantly higher prevalence of pneumonia, peritonitis, and upper respiratory tract infections, and higher tendency for sepsis in this group (Table 5). In contrast, control cats had a lower prevalence of these hematologic markers of inflammation. They tended (P 5.06) to present with noninflammatory and, presumed, noninfectious diseases, but this association was not significant. These conditions included thromboembolism, obstructive FLUTD, diabetes mellitus, and lung contusions. The same applies to feline asthma, which was significantly (P 5.03) more prevalent in control cats (Table 5).
28 Segev, Klement, and Aroch Table 7. Results in cats and dogs with toxic neutrophils, compared with their respective controls. * Dogs with toxic neutrophils versus controls (n 5 248) Cats with toxic neutrophils versus controls (n 5 150) Clinical signs of disease seen more frequently in animals with toxic neutrophils Fever, icterus, pale mucous membranes, vaginal discharge, abdominal organomegaly, melena Fever, icterus, vomiting, diarrhea, depression, dehydration, weakness, cachexia, Prevalence of hematologic abnormalities in animals with toxic neutrophils, compared with controls Leukocytosis q Leukocytosis q Leukopenia q Neutrophilia q Neutrophilia q Neutropenia q Neutropenia q Left shift q Left shift q Monocytosis q Anemia q Anemia q Macrocytosis q Macrocytosis q Hypochromia q Prevalence of biochemical abnormalities in animals with toxic neutrophils, compared with controls Hyperbilirubinemia q Hyperbilirubinemia q Increased LDH activity q Increased LDH activity q Hypocalcemia q Hypocalcemia q Hypoalbuminemia q Hypoproteinemia q Hypokalemia q Hypokalemiaq Hyponatremia q Hyperkalemia Q Increased ALP activity q Increased ALT activity q Increased c-gt activity q Increased creatinine concentration q Increased urea concentration q Increased triglycerides concentration q Prevalence of diseases in animals with toxic neutrophils, compared with controls Sepsis q Sepsis q Parvovirus infection q Panleukopenia q Pancreatitis q Peritonitis q Peritonitis q Acute renal failure q Pneumonia q Immune-mediated hemolytic anemia q Proximal respiratory tract disease q Disseminated intravascular coagulationq Shock q Pyometra q Pyoderma Q Thromboembolism Q Intervertebral disk disease Q Diabetes mellitus Q Feline asthma Q Lung contusion Q Prevalence of disease categories in animals with toxic neutrophils, compared with controls Neoplasia q Metabolic q Infectious bacterial q Infectious bacterial q Developmental Q Infectious viral q Degenerative Q Idiopathic Q Allergic Q Nutritional Q q, Higher prevalence; Q, lower prevalence; LDH, lactate dehydrogenase; ALP, alkaline phosphatase; ALT, alanine transaminase; c-gt, c-glutamyltransferase. * Included are categories with prevalence ratio $2 and P #.06. Although cats with toxic neutrophils had a significantly higher prevalence of neutrophilia and neutropenia, compared with controls, 53 and 47% of these cats presented with normal leukocyte and neutrophil counts, respectively. A combination of normal neutrophil and WBC counts and no left shift was present in 43% of cats with toxic neutrophils, whereas in 45% of those cats, leukocytosis, left shift, or neutrophilia and left shift were absent. Furthermore, difference was not found in mean band neutrophil count between cats with toxic neutrophils and controls. These findings indicate that presence of toxic neutrophils may not necessarily correlate with abnormalities in the absolute numbers of leukocytes, neutrophils, and band cells. Thus, in certain cases,
Feline Toxic Neutrophils 29 presence of toxic neutrophils may serve as the only hematologic marker of inflammation and infection. This finding emphasizes the benefit of assessing neutrophil morphology. Cats with toxic neutrophils also were found to be significantly more anemic, compared with controls, similar to what was reported in dogs. 20 Unfortunately, reticulocyte counts, bone marrow cytologic examination findings, and quantitative analysis of polychromasia were not available for the cats of this retrospective study. Thus, it is impossible to provide a definitive explanation for the higher prevalence of anemia in cats with toxic neutrophils. In contrast to dogs with toxic neutrophils, in which immune-mediated hemolytic anemia was more prevalent, compared with that in controls, we found no significant difference in the prevalence of hemolysis between the two groups of cats. Inflammation and blood loss are 2 possible mechanisms that might have led to a higher prevalence of anemia in cats with toxic neutrophils. Although anemia of inflammation is normocytic in most cases, 25 some cats with toxic neutrophils might have experienced other kinds of anemia, because macrocytosis, a possible marker of erythroid regeneration, was significantly (P 5.0002) more prevalent in cats with toxic neutrophils, compared with controls (6.71% versus 1.3%, respectively), although no significant (P 5.09) difference in mean MCV was found between groups. There were 4 of 23 significant mean serum biochemical variable differences between cat groups (Table 4). When the prevalence of deviations from the reference interval for 5 variables was compared between groups, it was statistically significant (Table 7). Hypocalcemia and mean lower total calcium concentration in cats with toxic neutrophils probably was of little clinical relevance, because differences in mean ionized calcium concentration between groups were not observed. We have no reasonable explanation for the observed higher mean serum ALP activity in the control cats, but there was no difference in the prevalence of increased serum ALP activity between groups. Data of disease categories in both groups (Table 6) indicate that cats with toxic neutrophils had a significantly higher prevalence of bacterial and viral infections. Possibly, the presence of viruses alone did not lead to toxic neutrophil formation, but rather, secondary bacterial infections complicated some of these viral diseases (eg, panleukopenia, upper respiratory tract infection), and played a major role in such changes. Compared with controls, cats with toxic neutrophils had a significantly higher prevalence of metabolic disorders (Table 6), as was also reported in dogs (Table 7). 20 This probably was a result of a cumulative effect of the higher, although nonsignificant, prevalence of acute renal failure, diabetic ketoacidosis, hepatic lipidosis, and other metabolic conditions, more commonly observed in cats with toxic neutrophils (Table 5). Idiopathic conditions were more prevalent in the control group because of the higher prevalence of FLUTD (obstructive and nonobstructive). The significantly higher prevalence of feline asthma in the controls probably accounted for the higher prevalence of allergic conditions in this group. Differences in the prevalence of toxicologic conditions between cat groups were not found, similar to what was reported for dogs, 20 although these conditions were previously linked with presence of toxic neutrophils. 2,9,17,26 This inconsistency could result from differences in the type and nature of toxicoses between the present study and previous reports. The results of this study indicated that the presence of toxic neutrophils in cats is associated with a higher prevalence of systemic and more severe diseases, compared with that in controls. This was reflected by marked clinical signs of disease and hematologic abnormalities, as well as a significantly longer hospitalization period and higher treatment cost in cats with toxic neutrophils. The hospitalization period was consistently longer when cats with toxic neutrophils were compared with controls that had the same diseases (Fig 4). Thus, for a given disease, when toxic neutrophils are present, a longer hospitalization period should be expected. There are some differences between cats and dogs with toxic neutrophils, compared with their respective control, but some similarities can be pointed out (Table 7). 20 The prevalence of many of the clinical signs of disease was higher in dogs and cats with toxic neutrophils, compared with that for their respective controls, indicating that, in both species, presence of toxic neutrophils was associated with more severe illness. In contrast, dogs with toxic neutrophils had considerably more serum biochemical and hematologic abnormalities (13 and 12, respectively) and mean serum biochemical and hematologic variable differences (13 and 12, respectively) than did controls. However, in this study, cats with toxic neutrophils had significantly fewer serum biochemical and hematologic abnormalities (5 and 6, respectively), and fewer mean serum biochemical and hematologic variable differences (4 and 6, respectively) than did their controls. 20 Thus, in cats, toxic neutrophils possibly may appear in milder disorders or earlier in the disease course, and may be a more sensitive indicator of illness in them, compared with dogs. Nevertheless, certain diseases were found to be significantly more prevalent in dogs and cats with toxic neutrophils, compared with their respective controls (eg, sepsis, peritonitis, canine parvovirus, panleukopenia). In both species, the infectious bacterial disease category was the most prevalent category in animals with toxic neutrophils. In dogs and in cats, presence of toxic neutrophils was associated with longer hospitalization duration (3.9 and 2.7 times longer, respectively) and higher treatment cost (2.0 and 1.4-fold, respectively), compared with their respective controls. However, only in dogs, was it also associated with significantly higher mortality. These differences further support the suggestion that presence of toxic neutrophils in cats is associated with milder diseases, compared with those in dogs. The comparison between dogs and cats has its limitations because of the different disease prevalence and bone marrow characteristics between species, and
30 Segev, Klement, and Aroch differences in the number of cases between studies (248 dogs versus 150 cats). Possibly, some variables that had only a tendency for higher prevalence in cats with toxic neutrophils, compared with controls, could potentially have become significant if a higher number of feline cases had been included in the study. This study has a limitation that relates mainly to the problem of multiple comparisons. We performed over 100 comparisons throughout the study. Thus, it is probable that some of the associations were erroneously found to be statistically significant. However, as outlined in a recent similar study conducted in dogs, 20 correction of this problem involves a major reduction in the statistical power of the study. 27 Thus, we preferred to maintain the P 5.05 value as the statistical significance cut-off, with the risk of rejecting some spurious null hypotheses. Additional research is warranted to strengthen the associations found in this study, and to assess the sensitivity and specificity of such changes for diagnosis of various diseases and conditions. In conclusion, evaluation of blood smears for neutrophil toxic changes is an inexpensive, quick, simple, and readily available process that was found to be a marker of infectious and metabolic disease processes in cats. In some cats, such changes were present when abnormal results of other tests, such as CBC and serum biochemical analysis, were minimal or absent, and along with the clinical signs, were the only indicators of disease. The presence of toxic neutrophils also was associated with longer hospitalization and higher treatment cost. Observation of toxic neutrophils was found to be an important diagnostic finding, as well as an aid in assessment of the patient, disease course, hospitalization duration, and therapeutic planning. In cats, unlike that in dogs, toxic neutrophils were not associated with higher mortality; therefore, they may not necessarily indicate the same severity of illness as they do in dogs. Additional research is warranted to strengthen the associations found in this study, to shed light on the mechanisms that induce toxic neutrophil production, and to assess the sensitivity and specificity of such changes for diagnosis of various diseases and conditions. Footnotes a Minos, ST-Vet, Montpellier, France b Abacus, Diatron, Vienna, Austria c Kone Progress Selective Chemistry Analyzer, Kone Corporation Instrument Group, Espoo, Finland d Nova 8, Nova Biomedical, Waltham, MA e SPSS 10.0 for Windows, SPSS Inc, Chicago, IL f PEPI 4.0, Abramson JH, Gahlinger PM. Computer programs for epidemiologists: PEPI version 4.0. 2001. Salt Lake City, UT; Sagebrush press References 1. Gay JC, Athens JW. Variations of leukocytes in disease. In: Lee GR, Foerster J, Lukens J, Paraskevas F, Greer JP, Rodgers GM, eds. Wintrobe s Clinical Hematology, 10 th ed. Baltimore, MD: Williams & Wilkins; 1998:1836 1861. 2. Smith GS. Neutrophils. In: Feldman BF, Zinkl JG, Jain NC, eds. Schalm s Veterinary Hematology, 5 th ed. Philadelphia, PA: Lippincott Williams, Wilkins; 2000:281 296. 3. Tyler RD, Cowell RI, Clinckenbreard KD, McAllister CG. Hematologic values in horses and interpretation of hematologic data. Vet Clin North Am Equine Pract 1987;3:461 484. 4. Jain NC. The neutrophils. In: Jain NC, ed. Schalm s Veterinary Hematology, 4 th ed. Philadelphia, PA: Lea & Febiger; 1986:714 717. 5. Schutze AE. Interpretation of canine leukocytosis responses. In: Feldman BF, Zinkl JG, Jain NC, eds. Schalm s Veterinary Hematology, 5 th ed. Philadelphia, PA: Lippincott Williams, Wilkins; 2000:366 381. 6. Harvey JW. Atlas of Veterinary Hematology: Blood and Bone Marrow of Domestic Animals. Philadelphia, PA: WB Saunders; 2001:51. 7. Gossett KA, MacWilliams PS, Cleghorn B. Sequential morphological and quantitative changes in blood and bone marrow neutrophils in dogs with acute inflammation. Can J Comp Med 1985;49:291 297. 8. Gossett KA, MacWilliams PS. Ultrastructure of canine toxic neutrophils. Am J Vet Res 1982;43:1634 1637. 9. Watson ADJ, Middleton DJ. Chloramphenicol toxicosis in cats. Am J Vet Res 1978;39:1199 1203. 10. Tvedten H. Referral and in-office laboratories. In: Willard MD, Tvedten H, Turnwald GH, eds. Small Animal Clinical Diagnosis by Laboratory Methods, 2 nd ed. Philadelphia, PA: WB Saunders; 1994:17.. 11. Parmar MKB, Machin D. Survival Analysis, A Practical Approach. West Sussex, UK: Wiley; 1995:65 96. 12. Hirsh D, Spencer S, Jang BS, Biberstein EL. Blood culture of the canine patient. J Am Vet Med Assoc 1984;184: 175 178. 13. Chickering WR, Brown J, Prasse KW, Dawe DL. Effects of heterologous antineutrophil antibody in the cat. Am J Vet Res 1985;46:1815 1819. 14. Salisbury SK, Lantz GC, Nelson RW, Kazacos EA. Pancreatic abscess in dogs: Six cases (1978 1986). J Am Vet Med Assoc 1988;193:1104 1108. 15. Woods JP, Crystal MA, Morton RJ, Panciera RJ. Tularemia in two cats. J Am Vet Med Assoc 1998;1:81 183. 16. Marchevsky AM, Read RA. Bacterial septic arthritis in 19 dogs. Aust Vet J 1999;77:233 237. 17. Geiger TL, Correa SS, Taboada J, et al. Phenol poisoning in three dogs. J Am Anim Hosp Assoc 2000;36:317 321. 18. Gasser AM, Birkenheuer AJ, Breitschwerdt EB. Canine Rocky Mountain spotted fever: A retrospective study of 30 cases. J Am Anim Hosp Assoc 2001;37:41 48. 19. Kociba GJ. Leukocyte changes in disease. In: Ettinger SJ, Feldman EG, eds. Veterinary Internal Medicine, 5 th ed. Philadelphia, PA: WB Saunders; 2000:1842 1857. 20. Aroch I, Klement E, Segev G. Clinical, biochemical, and hematological characteristics, disease prevalence, and prognosis of dogs presenting with neutrophil cytoplasmic toxicity. J Vet Intern Med 2005;19:64 73. 21. Hansen TWR. The pathophysiology of bilirubin toxicity. In: Maisels MJ, Watchko JF, eds. Neonatal Jaundice. London, UK: Harwood Academic Publishers; 2000:89 104. 22. Hansen TWR. Kernicterus: An international perspective. Semin Neonatol. 2002;7:103 109. 23. Morphis L, Constantopoulos A, Matsaniotis N, Papaphilis A. Bilirubin-induced modulation of cerebral protein phosphorylation in neonate rabbits in vivo. Science 1982;218: 156 158.
Feline Toxic Neutrophils 31 24. Rodrigues CMP, Solá S, Castro RE, et al. Perturbation of membrane dynamics in nerve cells as an early event during bilirubin-induced apoptosis. J Lipid Res 2002;43:885 94. 25. Waner T, Harrus S. Anemia of inflammation. In: Feldman BF, Zinkl JG, Jain NC, eds. Schalm s Veterinary Hematology, 5 th ed. Philadelphia, PA: Lippincott Williams, Wilkins; 2000: 205 209. 26. Bloom JC, Lewis UB, Sellers TS, Deldar A, Morgan DG. The hematopathology of cefonicid- and cefazadone-induced blood dyscrasias in the dog. Toxic Appl Pharmacol 1987;90:143 155. 27. Rothman KJ, Greenland S. Fundamentals of epidemiologic data analysis. In: Rothman KJ, Greenland S, eds. Modern Epidemiology, 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998:201 229.