A prospective cohort study to identify clinical diagnostic and prognostic markers of primary immune thrombocytopenia in dogs

Abstract

Background

Primary immune thrombocytopenia (pITP) in dogs presents a diagnostic challenge, and clinical markers of severity are lacking.

Objectives

Identify clinicopathologic features that differentiate pITP from secondary ITP (sITP) and markers related to bleeding severity, transfusion, and survival of dogs with pITP.

Animals

Ninety‐eight thrombocytopenic dogs (58 pITP and 40 sITP).

Methods

Client‐owned dogs with platelet counts <50 000/μL were enrolled in a prospective, multi‐institution cohort study. History and treatment information, through a maximum of 7 days, was recorded on standard data forms. Bleeding severity was scored daily using a bleeding assessment tool (DOGiBAT). At‐admission blood samples were collected for CBC, biochemistry, C‐reactive protein concentration, and coagulation panels, and to measure platelet surface‐associated immunoglobulin G (PSAIg) and expression of platelet membrane proteins and phospholipids. Dogs with evidence of coincident disease were classified as sITP.

Results

No definitive pITP diagnostic test was found. However, pITP cases were characterized by lower platelet counts, D dimer concentrations, and platelet membrane protein expression than sITP cases. Differentiation between pITP and sITP was further enhanced using logistic regression modeling combining patient sex, coagulation profile, platelet count, D dimer, and PSAIg. A second model of pITP severity indicated that low hematocrit and high BUN concentration were associated with non‐survival. Low hematocrit at admission, but not platelet count or DOGiBAT score, was associated with transfusion.

Conclusions and Clinical Importance

Pending validation studies, models constructed from at‐admission clinicopathologic findings may improve differentiation of pITP from sITP and identify the most severe pITP cases at the time of presentation.

Abbreviations

APTT

activated partial thromboplastin time

AT

antithrombin

CBC

complete blood count

CRP

C‐reactive protein

DOGIBAT

bleeding assessment tool

FSC‐A/SSC‐a

forward vs side scatter

ITP

immune thrombocytopenia

PBS

phosphate buffered saline

pITP

primary ITP

PRP

platelet rich plasma

PSAIg

platelet surface‐associated IgG

PT

prothrombin time

REDCAP

research electronic data capture

sITP

secondary ITP

VWF:Ag

von Willebrand factor antigen concentration

1. INTRODUCTION

Immune thrombocytopenia (ITP), in the absence of other identifiable disease, is among the most common causes of severe thrombocytopenia in dogs. ¹ , ² , ³ , ⁴ The disease counterpart in humans, now defined as primary immune thrombocytopenia, is an autoimmune disease targeting platelets with no obvious initiating or underlying cause. ⁵ , ⁶ , ⁷ , ⁸ The disease is managed using immunosuppressive drugs, and both species often experience a chronic form with relapses after cessation of treatment. In dogs and people, primary ITP (pITP) is considered a diagnosis of exclusion with no pathognomonic sign or definitive diagnostic test. ⁷ , ⁸ The term secondary ITP has been proposed as a broad term for patients with a comorbidity or coincident disease that might influence or contribute to immune‐mediated platelet destruction. ⁷ Included among conditions associated with secondary ITP (sITP) in people are infectious agents, drug exposure, transfusion, endocrinopathies, other autoimmune syndromes, lymphoproliferative disorders, and other cancers. ⁶ , ⁹ Many of these conditions also have been associated with ITP in dogs. ¹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ Although antiplatelet autoantibodies are implicated in pITP pathogenesis, their detection is not included in diagnostic guidelines for humans, ⁵ , ¹⁶ because circulating antiplatelet antibodies are absent in some pITP patients. Thrombocytopenia in these cases is believed to result from T‐cell mediated platelet destruction and impaired megakaryocyte function. Monitoring changes in platelet autoantibodies over time, however, may predict treatment response and relapse in humans ¹⁷ , ¹⁸ and dogs. ¹⁰

The clinical course of pITP in dogs is variable, ¹⁹ , ²⁰ , ²¹ and platelet counts at admission are poorly predictive of bleeding severity or response to treatment. ¹³ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ Some dogs develop life‐threatening hemorrhage necessitating extensive transfusion support, whereas others with similar platelet counts do not bleed. Glucocorticoids typically are considered first‐line treatment for dogs with pITP and are sometimes combined with other immunosuppressive drugs. Diagnostic uncertainties and poorly defined prognostic criteria often result in empirical and long‐term immunosuppressive drug regimens that can cause adverse effects and secondary infections. Managing pITP represents a considerable financial burden for owners because of the costs of diagnostic evaluation, immunosuppressive medications, and the required monitoring. ¹⁹ , ²⁷ Identifying clinical features and laboratory biomarkers denoting bleeding risk would enable individualized care for dogs with pITP by facilitating early diagnosis and improving severity predictions.

In humans, inflammatory biomarkers and qualitative platelet assays correlate with ITP severity. ²² , ²⁸ , ²⁹ The acute phase reactant, C‐reactive protein (CRP) contributes to antibody‐mediated platelet destruction and enhances platelet phagocytosis in mouse ITP models. ³⁰ Moreover, CRP concentrations correlate with platelet count and bleeding tendency in human pediatric ITP patients and predict platelet count recovery. ³⁰ Flow cytometric assays are well‐suited to assess platelet activation, membrane injury, and apoptosis in patients with low platelet counts. ³¹ , ³² The alpha granule protein, P‐selectin, is mobilized to the outer platelet membrane of activated platelets. Cytometric detection of platelet membrane P‐selectin (CD62P) denotes degranulation and is the most commonly used marker of basal platelet activation status and reactivity in response to agonist stimulation. The platelet fibrinogen receptor complex and tetraspanin are constitutive transmembrane proteins, present at high density on platelets. Membrane microparticles that express the fibrinogen receptor (CD41/CD61) or tetraspanin (CD9) are considered markers of platelet activation and injury. Apoptotic platelets and activated, procoagulant platelets externalize phosphatidylserine (PS), a negatively charged phospholipid restricted to the inner membrane leaflet of resting, viable cells. Outer membrane PS expression can be detected using protein markers, annexin V and lactadherin, that bind specifically to negatively‐charged phospholipids. Externalized platelet membrane PS facilitates assembly of active tenase and prothrombinase complexes as described in the cell‐based model of coagulation. ³³ Several studies of pITP in humans have examined the relationship of various cytometric markers to bleeding severity. Low bleeding scores indicating fewer bleeding events were found in patients that had low basal P‐selectin expression, yet retained a high stimulated P‐selectin response. ³⁴ In other studies, high antiplatelet antibody concentrations and increased clinical severity were associated with PS expression and other markers of apoptosis. ²⁹ Loss of constitutive platelet membrane antigens was found in acute and chronic childhood ITP patients, ²² and bleeding severity has been associated with resting and agonist‐induced platelet activation status. ²⁹ , ³⁴

Similar studies have yet to be performed in dogs with isolated ITP, comparable to pITP in people. We therefore aimed to evaluate the association of inflammatory and cytometric markers with bleeding severity, transfusion requirements, duration of hospitalization and survival to discharge in dogs with pITP, and to assess the relationship between bleeding severity and platelet count. We also aimed to compare novel clinicopathologic features of pITP from those of dogs with ITP and a coincident disease (sITP) to identify underlying differences inapparent from routine evaluation. We hypothesized that clinical and clinicopathologic features, including acute phase protein concentration and qualitative platelet abnormalities correlate with outcome measures and bleeding severity, and that at‐admission bleeding severity score, anatomic site(s) of hemorrhage, or changes in procoagulant and inflammatory proteins, or differences in platelet membrane markers, including platelet membrane‐associated antibodies might discriminate pITP from sITP in dogs.

2. MATERIALS AND METHODS

2.1. Study design and selection of animals

Ours was a multicenter prospective cohort study conducted across 8 veterinary academic or referral centers located in the Northeast (n = 4), Midwest (n = 2), Southwest (n = 1), and Western (n = 1) United States between December, 2018 and February, 2022. Dogs were eligible for the study if their initial platelet count at admission was ≤50000/μL and body weight was >5.5 kg. Dogs were excluded if they had been treated with glucocorticoids for >48 hours, received any other immunosuppressive or immunomodulating drugs, or had a neoplastic or inflammatory disease process known to be associated with consumptive coagulopathy and laboratory abnormalities that satisfied criteria of overt disseminated intravascular coagulation (DIC). ³⁵ , ³⁶ A priori sample size calculations using data from a previous study, ³⁷ suggested that enrollment of 100 eligible dogs would yield at least 60 cases of primary ITP, with an 80% chance of identifying significant (P < .05) correlations between clinicopathologic variables at admission and duration of hospitalization. All study procedures were reviewed and approved by the local Institutional Animal Care and Use Committee (Cornell University Protocol #2017‐0117) and studies were conducted with written, informed client consent.

2.2. Clinical evaluation and sample collection

Clinicians collected pretreatment blood samples and entered history and demographic information into a data collection form (Data S1). All dogs had onsite CBC and serum biochemistry panels performed for diagnostic evaluation at enrollment. Additional testing, including diagnostic imaging, was performed at the clinicians' discretion. Clinicians used a standard bleeding assessment tool (DOGiBAT) to evaluate bleeding severity (Data S2) after completing a DOGiBAT training course and quiz. ³⁷ The day of enrollment was defined as Day 1 of scoring. Dogs were evaluated for scoring daily during hospitalization until discharge, or to a maximum of 7 days. Dogs satisfying criteria of overt DIC (inciting disease and coagulation test abnormalities) or bone marrow failure (bi‐ or pancytopenia) were excluded from the study. Results of the clinical evaluation were used to classify ITP dogs into 2 subgroups using a previously described scheme. ³⁷ In brief, the classification of sITP included dogs with evidence of an infectious, metabolic, inflammatory, or neoplastic disease, or drug exposure history. Diagnostic testing for classification was performed at admission or while hospitalized or both, with inclusion of any pending test results at time of release or death. Dogs with no apparent coincident disease process identified at or within 7 days of enrollment (or by pending tests) were classified as pITP. The clinicians enrolling each case designated the final ITP classification, and results of platelet surface‐associated immunoglobulin G testing were not a criterion for study entry or ITP classification. At study enrollment, 8‐10 mL whole blood was collected by peripheral venipuncture or from indwelling IV catheters to yield 2 mL serum, 2 mL 3.2% or 3.8% citrate plasma, and 2‐3 mL EDTA whole blood. Serum and citrate plasma samples were stored frozen (−20°C) and EDTA blood was held at room temperature before shipping to the central laboratory. Samples were shipped with a refrigerant cold pack within 24 hours of collection for overnight delivery. Diagnostic testing, other than tests outlined below, was at the clinicians' discretion.

2.3. C‐reactive protein, hemostasis tests, and blood typing

Serum samples were stored at −80°C at the central laboratory until batch assay of CRP at approximately 3‐month intervals. Serum CRP concentration was measured using a commercial CRP assay for dogs (Tridelta Phase Canine CRP Assay, Tri‐Delta Diagnostics, Morris Plains, New Jersey) using the manufacturer's standards and controls. ³⁸ Coagulation profiles, consisting of activated partial thromboplastin time (aPTT), prothrombin time (PT), Clauss fibrinogen, antithrombin activity (AT), and D‐dimer concentration were performed upon receipt of the plasma samples, following previously described methods. ³⁶ The fibrinogen and AT assay standard curves were derived from dilutions of pooled canine plasma, prepared in‐house from 20 healthy dogs. The fibrinogen content of the canine standard was determined by a gravimetric method and the AT activity had an assigned value of 100%. The D‐dimer assay utilized a human plasma standard, with results reported in ng/mL. Residual citrate plasma was stored frozen (−20°C) and assayed within 48 hours of receipt to measure von Willebrand factor antigen concentration (VWF:Ag) using a sensitive ELISA. ³⁹ Blood typing for the DEA1 antigen was performed using EDTA blood and an immunochromatographic strip technique (Alvedia Canine DEA 1 Lab Kit, Animal Blood Resources Inc, Stockbridge, Michigan). The coagulation profile, VWF:Ag, and DEA 1 blood type results were provided to clinicians.

2.4. Platelet flow cytometry

Platelet surface‐associated immunoglobulin G (PSAIg) and expression of constitutive and activation‐dependent platelet membrane markers were analyzed in platelet‐rich plasma (PRP) isolated from EDTA blood within 24 hours of receipt at the central laboratory. The PSAIg assay was performed as described previously. ⁴⁰ Briefly, 900 μL of EDTA blood combined with 300 μL of HEPES/EDTA buffer was centrifuged for 30 seconds at 1900 g to harvest supernatant PRP. A 50 μL aliquot of PRP was reserved for membrane marker labeling, and the remaining PRP was washed in phosphate buffered saline (PBS), and then resuspended in 150 μL of 2% fetal bovine serum in PBS. The washed PRP was labeled with a fluorescein isothiocyanate (FITC)‐conjugated rabbit anti‐dog IgG antibody to detect PSAIg. Microspheres conjugated with canine IgG were labeled in parallel with the test samples as controls to establish a consistent fluorescence boundary for defining positive PSAIg. Aliquots of washed PRP also were labeled with an FITC‐conjugated rabbit isotype antibody as nonspecific antibody binding controls. Platelet membrane markers were detected in the reserved, unwashed PRP in 2 separate labeling reactions, each containing 10 μL PRP in a total reaction volume of 100 μL. In a triple‐label reaction, PRP was combined with 10 μL each of the following labels: a pan‐cell membrane marker, bio‐maleimide ⁴¹ (1:5000 dilution, BODIPYFL Maleimide, Thermo Fisher, Millersburg, Pennsylvania), anti‐GPIIIa antibody (1:50 dilution, CD61‐PE, clone SZ21, BD Biosciences, East Rutherford, New Jersey) and anti‐tetraspanin antibody (1:4 dilution, CD9‐APC, clone H19a, Biolegend, San Diego, California). A double‐label reaction was configured with 10 μL of an anti‐P selectin antibody (CD62P‐PE, clone AC1.2, BD Biosciences) and 10 μL of a PS‐binding protein (1:50 dilution, Lactadherin‐FITC, Haematologic Technologies, Essex Jct, Vermont). The reactions were incubated in the dark at room temperature for 15 minutes and then quenched in 400 μL PBS for analysis. Single use aliquots of frozen/thawed PRP (FT‐PRP) from a healthy dog were single‐labeled with each fluorescent probe to use as compensation and thresholding controls for each test day. All samples were analyzed using a BD Accuri C6+ cytometer (BD Biosciences). The manufacturer's instrument controls (Accuri C6+, CS&T RUO Beads) were analyzed daily before test samples. Data acquisition was set to obtain 7500 events in a platelet gate defined by forward vs side scatter (FSC‐A/SSC‐A) properties. Data files were analyzed using dedicated software (FlowJo, Treestar Inc, Ashland, Oregon). For analyses of membrane markers, the FT‐PRP samples first were gated on platelet‐sized events in FSC‐A/SSC‐A density plots (Figure S1A). Thresholds for positive vs negative events then were defined from single parameter histograms (Figure S1B‐F). The threshold boundaries for positive PSAIg were defined by the fluorescence intensity of control IgG‐conjugated microspheres, as previously described. ⁴⁰ Results were reported as percentage of platelets with “positive” values (ie, platelets with fluorescence values above defined thresholds for bound CD marker, lactadherin, or IgG). Sample plots from 2 dogs depicting gating and threshold boundaries to define positive values for reactions to detect platelet membrane markers and PSAIg are shown in Figure S2A‐E.

2.5. Treatment and study outcomes

Attending clinicians recorded treatment for up to 7 days of hospitalization and outcome using standard data collection forms (Data S3). Treatment information included immunosuppressive and immunomodulating drugs and dosages, gastroprotectants and other supportive treatments, and transfusion of blood component products, including platelet concentrates, or derivatives. Study outcome measures were receipt of any transfusion, days of hospitalization, survival to discharge, survival with or without relapse at 3 months after diagnosis, and survival with or without relapse at 6 months after diagnosis. Survival after discharge data was only compiled for pITP cases.

2.6. Statistical analyses

Patient demographics, test results, treatment, and DOGiBAT scores were transferred from study data collection forms into a secure, web‐based software platform (Research Electronic Data Capture [REDCAP]). ⁴² Clinicians were contacted by email and telephone for pending diagnostic test results at the time of hospital discharge, and for patient relapse status and survival at 3 months and 6 months after study enrollment. Statistical analyses were performed using SAS 9.4 (SAS Institute, Cary, North Carolina). The SAS procedures FREQ and UNIVARIATE were used to obtain descriptive statistics for categorical and continuous variables, respectively and PROC CORR to obtain correlations between variables. The Wilcoxon 2‐sample test in PROC NPAR1WAY was used to compare continuous variables between pITP and sITP dogs. The SAS procedure LOGISTIC was used to obtain models for factor effects on 3 outcomes: diagnosis of pITP, receipt of blood product transfusion, and survival. Factors were considered significant at P < .05. For each outcome, models with only 1 predictor were fit, and then model selection procedures were used to find combinations of predictors. Finally, models containing significant and near‐significant (P < .10) predictors identified above were fit to obtain the best predictive model. Odds ratios and predicted probabilities were calculated from these models. Model fit was assessed using R ² and by comparing Akaike's information criterion for the models with and without covariates. Receiver operating curves also were generated for each model.

3. RESULTS

3.1. Animals

A total of 103 dogs were enrolled between December, 2018 and August, 2021, with 6‐month follow‐up for the last dogs enrolled completed in February, 2022. Five enrolled dogs with nonimmune thrombocytopenia subsequently were excluded from analyses (Figure 1). The remaining 98 dogs included 58 diagnosed with pITP and 40 with sITP. The number of dogs enrolled from each site ranged from 4 to 23, with 55 dogs (56%) enrolled from sites in the Northeast. Within the 98 ITP dogs were 50 different breeds and mixed breed dogs. Most frequent were Labrador Retriever (n = 11), followed by Cockapoo, Golden Retriever, mixed breed, and German Shepherd Dog (n = 5 each).

FIGURE 1.

Summary of enrollment, diagnosis, and outcome for 103 dogs with thrombocytopenia (platelet count < 50 000/μL).

3.2. Clinical and diagnostic evaluations

Clinicopathologic data collected for the study are summarized in Table 1 and Table S1. The observed abnormalities common in pITP and sITP dogs were high median CRP and fibrinogen concentrations, low expression of CD9 and CD61, and high PSAIg percentages (Table 1). Significant differences were observed between dogs with pITP and sITP for platelet count, platelet membrane marker expression, and D dimer concentration (Figure 2). The median platelet count for pITP dogs was lower than for sITP dogs (5.5 vs 16.0 × 10³/μL; P < .0001). However, median DogiBAT scores for both groups were 5 (Table 1). Moreover, platelet count did not significantly correlate with admission bleeding score for all dogs (r _s = 0.058; P = .57) or for either group alone: pITP (r _s = −0.128; P = .35), sITP (r _s = 0.274; P = .09).

TABLE 1.

At‐admission patient characteristics, diagnostic test results and bleeding scores for dogs with pITP compared to dogs with sITP.

Characteristic	pITP (n = 58) a	sITP (n = 40) a	P‐value
Age (year)	9 (0.6‐14.4)	8 (0.5‐15)	.46
Sex (M/F)	27/31	13/27	.16
Season enrolled (Spring/Summer/Fall/Winter)	15/14/9/20	12/11/6/11	.89
Vaccinated within 3 months (Y/N)	11/43	10/29	.52
DogiBAT Day 1 total score	5 (0‐11)	5 (0‐14)	.75
DogiBAT Day 1 GI score	0 (0‐2)	0 (0‐2)	.31
Platelet count <10 vs 10‐50 × 103/μL	44/14	16/24	.0003
Platelet count (×103/μL)	5.5 (0‐26)	16 (0‐70) b	<.0001
MPV (fL)	16.6 (0‐34.7)	16 (8.6‐31.3)	.31
Hematocrit (%)	36.7 (12‐57.9)	34.4 (17.3‐62.5)	.17
WBC (×103/μL)	14.8 (2.0‐46.9)	11.6 (2.3‐40.7)	.09
aPTT (s)	12.6 (8‐31.5)	12.9 (8.4‐180)	.51
PT (s)	12.8 (10.1‐14.6)	12.8 (11.3‐180)	.31
AT (%)	91.5 (54‐134)	90.5 (43‐132)	.63
D‐dimer (ng/mL)	337 (45‐2869)	566.5 (0‐5797)	.007
Fibrinogen (mg/dL)	590.5 (229‐1498)	567.5 (15‐1391)	.37
VWF:Ag (%)	126 (24‐401)	124.5 (10‐401)	.30
BUN (mg/dL)	16 (4‐93)	17 (9‐109)	.76
CRP (ng/mL)	82.3 (0‐342.2)	98.4 (0‐500)	.30
CD9 (% of platelets that express antigen)	0.75 (0‐41)	6.4 (0‐90)	.002
CD61 (% of platelets that express antigen)	4.2 (0.2‐50.5)	12 (0.5‐91)	.02
CD62P (% of platelets that express antigen)	1.6 (0.3‐12.2)	2 (0.3‐18)	.04
PSAIg (% of platelets with membrane‐bound IgG)	12.8 (0.9‐55.7)	18 (2.5‐83)	.06

Note: Values are medians (and ranges), except for sex, season enrolled, vaccinated within 3 months, and platelet count <10 vs 10‐50 × 10³/μL, for which they are numbers of dogs. Bolded number indicates significance levels.

Abbreviations: aPTT, activated partial thromboplastin time; AT, antithrombin activity; CRP, C‐reactive protein; MPV, mean platelet volume; PSAIg, platelet surface associated IgG; PT, prothrombin time; VWF:Ag, von Willebrand factor antigen.

^a For some comparisons, the number of dogs in each group varied due to missing data: vaccinated: 54 (pITP), 39 (sITP); total DogiBAT score: 56 (pITP), 39 (sITP); GI DogiBAT score: 56 (pITP), 39 (sITP); MPV: 31 (pITP), 25 (sITP); Hematocrit: 55 (pITP); WBC: 55 (pITP); BUN: 53 (pITP), 38 (sITP); CD9: 42 (pITP), 31 (sITP); CD61: 42 (pITP), 31 (sITP); CD62P: 42 (pITP), 31 (sITP); PSAIg: 43 (pITP), 31 (sITP).

^b One dog with sITP had an estimated platelet count at admission of 50 × 10³/μL and was enrolled. This dog had a platelet count of 70 × 10³/μL from the complete blood count sample submitted after hospital admission.

FIGURE 2.

At‐admission laboratory tests with significant differences between primary and secondary immune thrombocytopenia (ITP) cases.

Qualitative platelet abnormalities were a common feature of ITP. Whereas >90% of healthy dog platelets express the constitutive membrane markers CD9 and CD61, ⁴³ , ⁴⁴ , ⁴⁵ the median CD9 and CD61 expression for both groups of ITP dogs were <15%. Expression of CD61 and CD9 were strongly correlated with each other (r _s = 0.866; P < .0001), but PSAIg was not significantly related to either CD61 (r _s = −0.012; P = .92) or CD9 (r _s = 0.055; P = .64). The strength of relationship for these variables with platelet count varied somewhat for the pITP and sITP groups (Table 2), and the pITP group had significantly lower expression of both platelet membrane markers than the sITP group (Figure 2B). Although CD62P expression significantly differed between pITP and sITP groups, this difference is unlikely to be clinically relevant because the median values were similar (1.6% vs 2.0%; Table 1). Increased CRP concentration (>25 ng/mL) was common in both groups, with 70% of pITP and 72% of sITP dogs having high concentrations. Among the hemostatic and inflammatory proteins, the only significant difference between groups was increased D dimer concentration in sITP dogs (Figure 2C).

TABLE 2.

Spearman correlations (r _s) for admission platelet count and expression of platelet membrane CD61 and CD9 and platelet surface associated IgG (PSAIg) for dogs with primary and secondary ITP.

Platelet parameter	Primary ITP (n = 42)		Secondary ITP (n = 31)
	Admission platelet count		Admission platelet count
	r s	P‐value	r s	P‐value
CD61	0.304	.05	0.413	.02
CD9	0.322	.04	0.479	.006
PSAIg	−0.263	.09	−0.028	.88

Note: Bolded number indicates significance levels.

Most study dogs underwent imaging and infectious disease screening. Abdominal ultrasound was the most common imaging modality, evaluated in 76% of pITP and 75% of sITP cases. A point‐of‐care test (SNAP 4‐DX) was the most common infectious disease screen, evaluated in 86% of pITP cases and 85% of sITP cases (Table S1). The most common coincident diseases identified in the 40 dogs with sITP included vector‐borne infectious diseases (37.5% of cases) and malignant and nonmalignant hematologic disorders, accounting for 27.5% of cases (Table 3). Other potentially associated or coincident diseases included recent antimicrobial drug treatment, solid tumors, and immune‐mediated and metabolic diseases (Table 3). Of the 58 dogs classified as pITP, 3 had recorded positive results for pathogen screening (n = 2 SNAP 4‐DX Lyme, n = 1 Brucella canis serology). These results were not assessed by the enrolling clinicians as evidence of coincident infection based on follow‐up testing (ie, Lyme C6 antibodies, Brucella canis PCR) and clinical judgment. Positive serologic screening test results for dogs classified as sITP were pursued at clinicians' discretion. Moreover, not all coincident diseases present in the sITP dogs have been reported previously to be associated or causally related to ITP in dogs, nor did our study assess whether any of the comorbidities were coincidental or causative.

TABLE 3.

Disease diagnoses at discharge for dogs with secondary ITP.

Category	n	n
Arthropod Vector‐Borne Disease a	15
Anaplasma spp.		7
Erhlichia spp.		4
Borreliosis		3
Hemotropic mycoplasma		1
Hematopoietic disorders b	11
Lymphoma/leukemia		7
Myelodysplasia/bone marrow aplasia		2
Splenic tumor		2
Antimicrobial drug therapy c	6
Penicillins/Aminopenicillins		5
Potentiated sulfonamides		1
Solid tumors b	4
Hepatic		1
Pulmonary/pleural		1
Mast cell disease		1
Oral tumor		1
Evans syndrome d	2
Cutaneous lupus erythematosus b	1
Diabetes mellitus e	1

^a 4DX positive n = 13; Other serology positive n = 8; PCR positive n = 6.

^b Cytologic or histologic diagnoses.

^c Clinical history.

^d Coombs positive.

^e Chemistry panel abnormality.

3.3. Survival

Survival data were compiled for all ITP dogs (Figure 1). The proportions of dogs surviving to hospital discharge and the duration of hospitalization were not different between dogs with pITP vs sITP, and the median duration of hospitalization was 3 days in both groups (Table 4). Additional 3‐month and 6‐month follow‐up information was compiled for the pITP cases (Figure 1). At 3 months after admission, 37/45 dogs with follow‐up information were alive and relapse‐free (82% relapse‐free survival). At 6 months after admission, 29/34 dogs with follow‐up information remained relapse‐free (85% relapse‐free survival). For the entire study population of 58 pITP cases, 33 dogs were alive at 6 months, 16 dogs had died, and 9 dogs were lost to follow‐up.

TABLE 4.

Outcome at seven days from enrollment for dogs with pITP compared with sITP.

Outcome	pITP (n = 58)	sITP (n = 40)	P‐value
Survived to discharge [n (%)]	49 (85%)	32 (80%)	.56
Status by category at 7 days [n (%)]			.79
Alive, discharged	45 (78%)	30 (75%)
Alive, hospitalized	4 (7%)	2 (5%)
Died, hospitalized	1 (2%)	0 (0%)
Euthanized	8 (14%)	8 (20%)
Hospitalization duration in days [median (range)]	3 (0 to >7)	3 (0 to >7)	.43

Note: Values are number (percent) of dogs, or median (range) for duration of hospitalization.

3.4. Treatment: Primary ITP cases

Blood product transfusion and immunosuppressive or immune‐modulating drug administration were compiled for the pITP dogs (Figure 3). At least 1 transfusion was administered to 20/55 (36.4%) of the pITP dogs with treatment information. Most transfused dogs (52%) received only packed red cells, with the remaining dogs receiving either fresh whole blood, lyophilized platelets, or multiple blood products (Figure 3). Data on drug treatment were obtained for 54 pITP cases (Figure 3). Similar percentages of dogs received glucocorticoids only (31.5%), glucocorticoids and additional immunosuppressive agents (30%), or glucocorticoids with vincristine and additional immunosuppressive agents (28%). Intravenous immunoglobulin was administered to 3 dogs, as the only immunosuppressive agent combined with glucocorticoids (n = 1) or as a second (n = 1) or third agent (n = 1) combined with glucocorticoids.

FIGURE 3.

Summary of immunosuppressive and transfusion therapy for 58 dogs with primary immune thrombocytopenia.

3.5. Logistic regression model for diagnosis of primary ITP

The best‐fit model for pITP diagnosis included the following 6 factors: patient sex, platelet count, aPTT, PT, D dimer, and PSAIg (Table 5):

$$\text{Probability(pITP)}=\begin{array}{l}\exp (3.0274\left[\text{if male}+1.5514\right]-0.145\\ \times \text{platelet count}+0.4161\times \text{aPTT}-0.4048\times \mathrm{PT}\\ -0.0023\times \mathrm{D}\text{dimer}-0.0233\times \text{PSAIg})/1\\ +\exp (3.0274+1.5514-0.145\\ \times \text{platelet count}+0.4161\times \text{aPTT}-0.4048\times \mathrm{PT}\\ -0.0023\times \mathrm{D}\text{dimer}-0.0233\times \text{PSAIg}).\end{array}$$

TABLE 5.

Effects of clinicopathologic characteristics on differentiation of primary immune thrombocytopenia (pITP) from secondary ITP (sITP).

Parameter	DF	Estimate	SE	Wald Chi2	OR	95% CI	P‐value
Intercept	1	3.0274	1.0055	9.0652			.002
Sex (Male vs female)	1	1.5514	0.7182	4.666	4.718	1.245, 21.976	.03
Platelet count	1	−0.145	0.0455	10.1618	0.865	0.78, 0.935	.001
aPTT	1	0.4161	0.188	4.8977	1.516	1.092, 2.288	.03
Prothrombin time	1	−0.4048	0.1889	4.5915	0.667	0.395, 0.927	.03
D dimer	1	−0.0023	0.00089	6.3836	0.998	0.996, 0.999	.01
PSAIg	1	−0.0233	0.0238	0.964	0.977	0.93, 1.022	.32

Note: Probability modeled is diagnosis of pITP (vs sITP). Although PSAIg is non‐significant, it is included in the model because removing it caused several of the other variables to become non‐significant. Regarding model fit, Akaike's Information Criterion (AIC) for the model with intercept‐only was 102.631, adding the 6 covariates decreased it to 72.566. The model R ² was 0.4336. Bolded number indicates significance levels.

Abbreviations: aPTT, activated partial thromboplastin time; CI, confidence interval; DF, degrees of freedom; OR, odds ratio; PSAIg, Platelet‐surface associated immunoglobulin G; SE, standard error.

The odds of a pITP diagnosis were 4 to 5 times higher for male vs female dogs. The odds of a pITP diagnosis increased with increases in aPTT and decreased with increases in PT and D dimer concentration. A receiver operated characteristic (ROC) curve (area under the curve [AUC], 0.881; 95% confidence interval [CI], 0.796‐0.965; P < .0001) constructed from the model is presented in Figure 4A. The diagonal line denotes an AUC of 0.5 (ie, chance alone).

FIGURE 4.

Receiver operating characteristic (ROC) curves for models of primary immune thrombocytopenia (ITP) diagnosis, receipt of transfusion, and survival. (A) Diagnostic model. ROC curve for diagnosis of primary immune thrombocytopenia (pITP) vs secondary ITP (sITP) (predictor variables: sex, platelet count, activated partial thromboplastin time, prothrombin time, D dimer, and platelet‐surface associated immunoglobulin G). (B) Transfusion model. ROC curve for whether a pITP dog received a blood product transfusion (predictor variables: hematocrit, number of days hospitalized). (C) Survival model. ROC curve for whether a pITP dog survived or died (predictor variables: hematocrit, blood urea nitrogen).

3.6. Indicators of pITP severity‐correlations and probability modeling

No significant correlations among disease severity indicators were observed including between platelet count or DOGiBAT score with blood product administration, days of hospitalization, or survival to discharge (Table 6). Probability models for predicting blood transfusion receipt and survival to discharge were constructed. At‐admission hematocrit influenced transfusion wherein for each unit decrease in hematocrit, the likelihood of receiving a transfusion increased by approximately 1.3 times. For each additional day of hospitalization, the likelihood of transfusion increased by approximately 3.5 times (Table 7). Admission hematocrit and BUN concentration were associated with survival to discharge, wherein likelihood of non‐survival increased by approximately 1.1 times for every unit decrease in hematocrit, or for every unit increase in BUN concentration (Table 8). The ROC curves constructed from hematocrit and BUN models had AUC of 0.960 (95% CI, 0.915‐0.1.000; P < .0001) and AUC of 0.904 (95% CI, 0.809‐0.998; P = .01) respectively and are presented in Figure 4B,C.

TABLE 6.

Spearman correlations (r _s) for admission platelet count and DOGiBAT score for dogs with primary ITP and select disease severity indicators of transfusion (yes/no), days hospitalized, and survival to discharge (survived/died).

Outcome event	Platelet count (n = 56) a		DOGiBAT Score (n = 55) a
	r s	P‐value	r s	P‐value
Blood product administration	−0.1847	.17	0.1524	.27
Hospitalization duration	−0.0638	.64	0.0978	.48
Survival to discharge	−0.096	.47	0.0728	.59

^a Excluded from analyses are 2 pITP cases with recorded platelet count range but no platelet count value, and 3 pITP cases missing an admission DOGiBAT score.

TABLE 7.

Effects of factors influencing the probability of receiving any transfusion for dogs with primary ITP.

Parameter	DF	Estimate	SE	Wald Chi2	OR	95% CI	P‐value
Intercept	1	2.9412	1.985	2.1955			.14
Hematocrit	1	−0.2358	0.08	8.6943	0.79	0.649, 0.896	.003
Hospitalized (days)	1	1.2904	0.437	8.7184	3.634	1.851, 10.923	.003

Note: Regarding model fit, Akaike's Information Criterion (AIC) for the model with intercept‐only was 73.1888, adding the 2 covariates decreased it to 31.243. The model R ² was 0.5729. Bolded number indicates significance levels.

Abbreviations: CI, confidence interval; DF, degrees of freedom; OR, odds ratio; SE, standard error.

TABLE 8.

Effects of factors on probability of non‐survival to discharge from diagnosis of primary immune thrombocytopenia.

Parameter	DF	Estimate	SE	Wald Chi2	OR	95% CI	P‐value
Intercept	1	−0.5003	1.5936	0.0986			.75
Hematocrit	1	−0.109	0.0512	4.5249	0.897	0.795, 0.98	.03
Blood urea nitrogen	1	0.0794	0.0376	4.4619	1.083	1.016, 1.177	.03

Note: Probability modeled includes death or euthanasia vs survival. Regarding model fit, Akaike's Information Criterion (AIC) for the model with intercept‐only was 43.373, adding the 2 covariates decreased it to 31.939. The model R ² was 0.2526. Bolded number indicates significance levels.

Abbreviations: CI, confidence interval; DF, degrees of freedom; OR, odds ratio; SE, standard error.

4. DISCUSSION

Our study aimed to compare combined clinical and clinicopathologic features of dogs with pITP and sITP to identify criteria that might be diagnostically discriminating, provide pathophysiologic insights, and enhance the delivery of care to individual animals. Several differences between the clinicopathologic variables of pITP and sITP, defined in our study as dogs having any coincident disease, were observed. Dogs with pITP had lower platelet counts, decreased expression of platelet membrane proteins, and lower concentrations of D dimer than dogs with sITP. However, no single test or variable was fully discriminating because of substantial overlap in measured values between diagnostic categories. A diagnostic model, constructed from platelet count and D dimer results, combined with additional variables of patient sex, coagulation profile (aPTT, PT), and PSAIg enhanced differentiation of pITP from sITP dogs and may prove useful for early patient management. Specifically, a male dog with a very low platelet count, a long aPTT, but normal PT and D dimer results was significantly more likely to have pITP than sITP. Previous studies of thrombocytopenic dogs have found that ITP generally is characterized by lower platelet counts than consumptive or production defects, although no definitive threshold platelet count exists. ¹³ , ³⁷ , ⁴⁶ Our study supports the diagnostic relevance of platelet count, suggesting that it can aid in differentiating pITP from sITP for dogs with platelet counts <50 000/μL. Thrombocytopenia severity also aids diagnosis of ITP in humans, with enhanced accuracy of clinical predictive models that include the lowest platelet count. ⁴⁷

The residual circulating platelets of dogs with pITP and sITP had decreased expression of platelet membrane antigens CD61 and CD9. Platelet CD61 (GPIIIa; integrin β3) is a subunit of the platelet fibrinogen receptor. Combined with CD41 (GpIIb; integrin αIIb) this heterodimeric complex is the most abundant platelet membrane adhesion receptor, and is densely expressed on resting and activated platelets. ⁴⁸ Tetraspanin‐29, or CD9, is also a high abundance, transmembrane platelet protein. ⁴⁹ Although decreased platelet CD61 expression has been reported previously in dogs with pITP, ¹² the underlying cause for this decrease is unknown. It is possible that abnormal CD61, and the decrease in CD9 labeling we observed, was caused by interference by platelet‐bound anti‐fibrinogen receptor antibodies or polyclonal autoantibodies targeting both the fibrinogen receptor and CD9. Unlike the fibrinogen receptor complex, anti‐tetraspanin antibodies have not been identified in human patients with ITP. Tetraspanin is structurally distinct from integrins, but CD9 colocalizes with the fibrinogen receptor on the membrane of activated platelets and platelet‐derived microparticles. Our finding that PSAIg was unrelated to CD61 or CD9 expression is similar to the results of an ITP study in children that described no association between PSAIg and decreases in CD61 expression. ²² Rather than direct autoantibody interference, an alternate explanation is that immune‐mediated platelet membrane injury non‐specifically alters or disrupts multiple platelet membrane antigens such that the monoclonal antibodies no longer recognize epitope targets. Our cytometry protocol used bio‐maleimide labeling as a cell‐type nonspecific membrane marker to eliminate machine noise, cell debris, and protein precipitates from the platelet analysis gate. However, we cannot exclude the potential presence of microparticles or membrane fragments originating from non‐platelet cell‐types as an alternate explanation for decreased CD61 and CD9 labeling.

A model constructed to identify predictors of pITP disease severity indicated that low hematocrit and high BUN concentration at admission were associated with non‐survival to hospital discharge. Only hematocrit at admission was associated with transfusion requirement. Importantly, in dogs with pITP, platelet count at admission was not correlated with clinical bleeding score and did not predict transfusion or survival. These findings are in agreement with previous case reports and case series of pITP in dogs and people that describe an association between high BUN concentration and non‐survival, ¹⁰ , ²⁰ , ⁵⁰ and that low hematocrit is related to receipt of transfusion. ⁵¹ , ⁵² In our study, DOGiBAT score at admission did not correlate with duration of hospitalization or transfusion, in contrast to results from a previous study by our research group. ³⁷ Transfusion was strongly correlated with days hospitalized in our current study, and the duration of hospitalization was shorter (3 days vs 4.5 days), suggesting that differences in patient population, case management, or both across a broader group of collaborating institutions may underlie discrepancies between studies.

Some limitations of our study are inherently related to the need for a gold standard for pITP diagnosis. The classification of sITP was assigned to dogs having any at‐admission disease process or recent drug treatment that the enrolling clinician considered relevant, thus incorporating clinician bias and uncertainty to include conditions that were not an immune trigger. Additionally, some pITP dogs with early or occult infection or neoplasia, or relatively incomplete diagnostic evaluations, may have been misclassified. A lack of a standard protocol for pathogen testing resulted in non‐uniform screening and confirmatory testing for all enrolled dogs. Similar proportions of pITP and sITP dogs had point‐of‐care screening tests performed. However, the sITP cases had more intensive follow‐up testing based on the clinicians' evaluation of abnormal or positive results of these tests. Of the 3 pITP cases with positive initial pathogen screening, clinicians pursued additional testing in 2 cases to combine with their clinical assessment of pITP. Further study bias was introduced from selection of enrollment centers. Cases were recruited from 8 different institutions, most were in the Northeast, and all were referral centers. The populations in our study thus are subject to some geographic bias and may not be representative of cases seen by primary care veterinarians. Case accrual and follow‐up were hindered by the COVID‐19 pandemic, leading to decreased numbers of recruitment sites and fewer collaborating clinicians. Finally, our study was not designed as a treatment trial of pITP or sITP, thus restricting the focus of data collection and follow‐up. Classification of pITP vs sITP was not based on response to treatment. Our study had no restrictions on case management, thus introducing biases based on clinician preference and precluding direct comparisons in treatment‐related outcomes for the pITP cases.

In conclusion, we identified clinicopathologic features of dogs with pITP that differed at the time of admission from dogs with sITP. We found qualitative changes in platelet membrane antigen expression in both pITP and sITP dogs that may provide pathophysiologic clues or have applications as biomarkers of immune injury. The associations and models developed from our study population will require further evaluation for their clinical utility in follow‐up studies, but the DOGiBAT scoring tool and standardized data collection forms are directly applicable to these future studies. In concert with previous studies, our results support the importance of high BUN concentration and low hematocrit as indicators of severe pITP. Together these findings will aid in designing treatment trials to guide individualized patient management.

CONFLICT OF INTEREST DECLARATION

Authors declare no conflict of interest.

OFF‐LABEL ANTIMICROBIAL DECLARATION

Authors declare no off‐label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION

Approved by Cornell University IACUC, #2017‐0117.

HUMAN ETHICS APPROVAL DECLARATION

Authors declare human ethics approval was not needed for this study.

Supporting information

Data S1: Supplementary File 1.

Data S2: Supplementary File 2.

Data S3: Supplementary File 3.

Data S4: Supplementary ModelTemplate.

Table S1: Summary of at‐admission testing performed for dogs classified with pITP, sITP, or non‐ITP.

Figure S1: Flow cytometric gating strategy using canine frozen thawed platelet rich plasma (FT‐PRP) as platelet membrane marker control. (A) Density plot [forward scatter (FSC‐A) vs side scatter (SSC‐A)] displaying the platelet gate applied to single‐labeled aliquots of FT‐PRP. (B–F) Histogram plots of labeled FT‐PRP within the platelet gate to display event counts (Y axis) vs fluorescence intensity (X‐axis). In each plot, a bisecting gate differentiates positive from negative events for a series of fluorescent markers identified as follows: (B) CD61PE = CD61 phycoerythrin; CD61 label binds platelet integrin β₃ and is used as a constitutive platelet membrane marker. (C) CD9APC = CD9 allophycocyanin; CD9 label binds tetraspanin and is used as a non‐integrin constitutive platelet membrane marker. (D) Bio‐maleimide is a fluorescent marker that binds cell‐membrane thiol groups regardless of cell origin and is used as a marker of cell membranes, including microparticles. (E) Lactadherin FITC = fluorescein isothiocyanate conjugated lactadherin protein; lactadherin binds to outer membrane phosphatidylserine (PS) and is a marker of apoptotic or procoagulant platelets and microparticles. (F) CD62P‐PE = CD62P phycoerythrin; CD62P label binds to the platelet alpha granule protein P selectin and is used to identify degranulated platelets. Note that FT‐PRP contains no events positive for CD62P.

Figure S2: Sample cytometry plots of two dogs with immune thrombocytopenia (ITP) demonstrating gating strategy to characterize platelet membrane markers and platelet‐bound antibody. Top row: ITP case 1. Bottom row: ITP case 2. (A & D) Ancestry plots from triple‐labeled platelet rich plasma samples displaying a forward scatter vs side scatter platelet gate (top inset) and histogram to define a biomaleimide positive gate (bottom inset) and density plot displaying proportions of gated events expressing CD9 (APC‐A) in Q1 & Q2, and CD61 (CD61‐PE) in Q2 and Q3. (B & E) Density plots from double‐labeled platelet rich plasma samples (within a forward scatter vs side scatter platelet gate) displaying proportions of events expressing phosphatidyl‐serine (Lact‐FITC) in Q1 and Q2, and P selectin (CD62P‐PE) in Q2 and Q3. (C & F) Overlay histograms from washed platelet rich plasma suspensions (within a forward scatter vs side scatter platelet gate) labeled with a rabbit anti‐dog IgG reagent (thick black line) and in separate reactions labeled with a rabbit isotype antibody (gray fill). The positive vs negative events are indicated with a bisected gate (IgG FITC+ and IgG FITC−).

Brooks MB, Goggs R, Frye AH, et al. A prospective cohort study to identify clinical diagnostic and prognostic markers of primary immune thrombocytopenia in dogs. J Vet Intern Med. 2024;38(2):1022‐1034. doi: 10.1111/jvim.16985