Negative correlation between serial changes in the reticulocyte number and C-reactive protein level in dogs with delayed reticulocytosis

Abstract

Inflammation suppresses erythropoiesis, leading to non-regenerative anemia in dogs. Currently, clinical evidence of this inhibitory effect remains limited. This study provides important clinical insight by demonstrating a negative correlation between C-reactive protein (CRP) concentration and reticulocyte count over time in dogs with inflammatory anemia. This study aimed to leverage inflammatory anemia as a model to examine the longitudinal relationship between CRP levels and reticulocyte counts in dogs. Medical records of 23 dogs diagnosed with various inflammatory conditions and persistently non-regenerative erythropoiesis were retrospectively reviewed. Serial results of hematological and serological parameters were analyzed using linear mixed-effects models to account for intra- and inter-individual variation. Reticulocyte counts in anemic dogs with resolving underlying inflammatory conditions were significantly negatively correlated with CRP levels (standardized β=−0.53, 95% confidence interval (CI) −0.68 to −0.37; P<0.001). Moreover, cross-sectional analysis revealed that higher CRP levels were significantly associated with lower reticulocyte counts, suggesting that the severity of inflammation may be linked to reduced reticulocyte counts (Pearson’s r=−0.5; P=0.01). Similarly, erythropoietic responses to darbepoetin alfa appeared to be attenuated in dogs with elevated CRP concentrations. In these dogs, whose mean CRP concentration was 6.15 mg/dL, the onset of reticulocytosis was delayed approximately twofold, and the increase in peak reticulocyte count was approximately 38% lower than that in dogs with normal CRP concentrations. This study provides clinical evidence supporting an association between systemic inflammation and reduced or delayed reticulocytosis in dogs, highlighting the clinical utility of concurrently monitoring CRP and reticulocyte levels to assess disease progression and inflammatory recovery.

INTRODUCTION

Reticulocytes are immature erythrocytes produced in the bone marrow and serve as a reliable biomarker for erythropoietic activity in anemic dogs [10]. Reticulocytosis typically develops in dogs within 2 to 4 days of anemia onset [49]. Meanwhile, various factors, such as inflammation, chronic kidney disease (CKD), bone marrow stem cell injury, or endocrinopathies, can result in non-regenerative anemia [19]. Inflammatory anemia, also known as anemia of inflammatory disease (AID), is the most common cause of non-regenerative anemia in veterinary medicine [5, 41]. AID arises secondary to a wide spectrum of inflammatory conditions, including infections, pancreatitis, neoplasia, and immune-mediated disorders [5]. The pathogenesis of AID involves multiple complex mechanisms, including dysregulated iron metabolism, cytokine-mediated blunting of the reticulocyte response, and desensitization of erythropoiesis [16].

A study showed that inflammation impairs reticulocyte maturation in anemic dogs by reducing the immature reticulocyte fraction (IRF) and hemoglobin synthesis [33]. However, a subsequent study found no significant influence of inflammation on canine IRF [26]. Both studies found that inflammation did not affect total reticulocyte counts, suggesting that the relationship between inflammation and reticulocyte biology warrants further study. In humans, inflammatory dynamics, including both magnitude and fluctuations, markedly influence erythropoiesis and reticulocyte biology, including proliferation, differentiation, and maturation [13, 21]. These dynamics have been primarily investigated using longitudinal approaches rather than single-timepoint assessments, thereby providing a more comprehensive evaluation of the impact of inflammation on erythropoietic activity [4, 21]. Conversely, studies in dogs examining correlations among relevant parameters have predominantly employed cross-sectional designs [7, 26, 33, 35].

C-reactive protein (CRP) is a sensitive and reliable biomarker of both local and systemic inflammation in dogs and is routinely measured using point-of-care test (POCT) analyzers [12, 30]. In dogs, CRP levels have been used as a barometer of inflammatory dynamics, but do not reflect underlying disease mechanisms or indicate specific disease types [26, 33]. In this context, POCT platforms for key mediators involved in inflammatory pathways, such as cytokines and iron-regulating hormones (e.g., hepcidin), would be valuable but remain unavailable for routine veterinary use [51].

This study aimed to evaluate the currently descriptive concept that inflammation inhibits reticulocyte maturation in dogs by using linear mixed models (LMMs) to assess the longitudinal correlation between CRP concentrations and hematological parameters, including reticulocyte counts, in 23 dogs with inflammatory anemia and delayed reticulocytosis.

MATERIALS AND METHODS

Data collection

A retrospective review was conducted at a veterinary teaching hospital using a commercial electronic medical record system (IntoVET EMR) from January 2022 to August 2025. Cases were identified by searching the IntoVET EMR database using the keyword “anemia.” Written informed consent for data use was obtained from the dog owner before study enrollment.

Diagnostic criteria for hemolytic anemia

Cases were identified by searching the IntoVET EMR database using the keywords “hemolytic anemia,” “babesiosis,” and “immune-mediated hemolytic anemia.” Hemolytic anemia was diagnosed based on hemolytic findings, including spherocytosis, autoagglutination, or hemoglobinemia. Immune-mediated hemolytic anemia (IMHA) was diagnosed by confirming hemoglobinemia in plasma, identifying spherocytes and schistocytes on blood smears, detecting hemoglobinuria on urinalysis, obtaining a positive Coombs’ test result, and observing a positive response to prednisolone therapy after exclusion of infectious causes.

Diagnostic approach for AID

AID or inflammatory anemia was diagnosed according to previously described laboratory criteria, defined as persistent non-regenerative anemia accompanied by elevated serum CRP concentrations and a chronic inflammatory leukogram [49]. Dogs with end-stage CKD, chronic blood loss, iron deficiency, or endocrine disorders causing non-regenerative anemia were excluded.

Clinicopathologic analyses

A complete blood count (CBC) was performed on a ProCyte Dx hematology analyzer (IDEXX Laboratories, Westbrook, MA, USA) to evaluate the erythrogram, including morphological classification of anemia and absolute reticulocyte count, as well as the leukogram. Resident veterinarians examined peripheral blood smears to assess the morphology of circulating blood cells. Manual reticulocyte staining with new methylene blue (NMB) (Sigma-Aldrich, St. Louis, MO, USA) was performed to confirm the automated reticulocyte counts. CRP concentrations were measured in fresh serum or plasma using a clinical chemistry analyzer (IDEXX Laboratories). When multiple measurements were obtained on the same day, the mean CRP value was used. Complete urinalysis, including dipstick analysis, and evaluation of serum or plasma coloration were performed to assess intravascular hemolysis.

Inclusion and exclusion criteria

The study scheme is shown in Fig. 1. In this study, regeneration of anemia was defined as an absolute reticulocyte count ≥110,000/µL, corresponding to the upper reference limit of the automated hematology analyzer, as mild reticulocytosis can also occur in clinically healthy dogs. Inclusion criteria were as follows: (i) non-regenerative status persisting more than 4 days from the onset of anemia; (ii) availability of more than four consecutive follow-up blood tests, enabling within-patient trend analysis. Exclusion criteria included dogs that had received chemotherapy, owing to the potential confounding effects of drug-induced bone marrow suppression [6], and dogs in which the impact of inflammation on erythropoiesis could not be accurately evaluated. Dogs that underwent blood transfusion or received darbepoetin treatment during the study period were retained for downstream analyses.

Fig. 1.

Flowchart of case selection. Flow diagram of the inclusion and exclusion criteria for dogs enrolled in this study. Cases are included if dogs present with anemia and delayed reticulocytosis (>4 days), have sufficient blood test results, and are not undergoing chemotherapy. Dogs that do not meet these criteria are excluded, e.g., those with anemia but without an increase in CRP concentrations. CRP, C-reactive protein.

Statistical analysis

Associations between reticulocyte counts and clinicopathologic parameters were explored using Pearson correlations and univariate LMMs with patient-specific random intercepts to account for repeated measurements [9]. Pairwise Pearson correlation coefficients were first calculated to examine associations between reticulocyte counts and CBC parameters. Two-tailed P-values (α=0.05) were used without multiple comparison correction, as these analyses were descriptive and exploratory. Continuous predictors were transformed using the Yeo–Johnson method to approximate normality, then standardized to z-scores to enable comparison of regression coefficients.

Associations between clinicopathologic parameters and reticulocyte counts were assessed using univariate LMMs with a random intercept for each patient. Recurrent episodes within the same dog were analyzed at the episode level, with each dog included as a random effect to account for clustering of episodes within individuals. Competing residual correlation structures (discrete-time AR(1) and continuous-time CAR(1)) were compared under identical fixed effects, and the structure with the lowest Akaike information criterion (AIC) was selected. To address multiple-comparison issues in univariate LMMs, we applied the Benjamini–Hochberg false discovery rate correction and reported associations with q≤0.10. To account for irregular visit intervals, we conducted a sensitivity analysis to assess whether applying time-interval-based weights influenced the results. The primary analysis used the unweighted model, and the weighted model was reported only for robustness (sensitivity) checks. Fixed-effect significance was assessed using Wald t-tests. Model performance was summarized using marginal and conditional R², AIC, the Bayesian information criterion, and the intraclass correlation coefficient.

We also fitted a multivariate LMM with CRP and canine pancreatic lipase immunoreactivity (cPLI) as fixed effects to assess any independent associations with reticulocyte counts, using the same random-effects structure as the univariate LMMs. Model fit was evaluated using AIC and marginal/conditional R².

For visualization, we generated forest plots of standardized effect estimates with 95% confidence intervals (CIs), scatterplots with ordinary least-squares regression lines (for descriptive purposes), and representative time-series plots of reticulocyte counts and CRP concentrations for typical cases of inflammatory and regenerative anemia, selected based on within-patient correlation and temporal change. All analyses were performed in R (v4.5.1; R Foundation for Statistical Computing, Vienna, Austria). LMMs were fitted using the nlme package (v3.1-168). Variable transformations were conducted using bestNormalize, and figures were created with ggplot2 (v3.5.2) and assembled using patchwork (v1.3.1).

Comprehensive clinical evaluation and diagnostic workup of underlying diseases in dogs

All dogs underwent thorough clinical evaluation, including history taking, assessment of clinical signs, physical examination, and routine laboratory testing. Serum biochemistry was analyzed using the Catalyst One analyzer (IDEXX Laboratories). Additional diagnostic assays were applied as appropriate: canine pancreatic lipase immunoassays (SNAP^® cPLI^TM, IDEXX Laboratories; Vcheck cPL, Bionote Inc., Hwaseong-si, Republic of Korea), endocrine function tests (serum total and free thyroxine, low-dose dexamethasone suppression, and ACTH stimulation tests), and infectious disease panels (rapid antibody kits, PCR assays, and multiplex screening kits). To screen for Babesia infection, both a point-of-care antibody rapid test (Canine Babesia Antibody Rapid Kit, Bionote Inc.) and real-time PCR (CareDx™ Canine Babesia Real-Time PCR Kit, Carevet Co., Ltd., Siheung-si, Republic of Korea) were used. Diagnostic imaging included radiography (VXR-E40, DRGEM, Gwangmyeong, Republic of Korea), ultrasonography (V8, Samsung Medison, Seoul, Republic of Korea), computed tomography (SOMATOM Emotion, Siemens Healthineers, Erlangen, Germany), and magnetic resonance imaging (MAGNETOM Sempra, Siemens Healthineers). Neoplastic lesions were confirmed cytologically or histologically when feasible. Immune-mediated diseases were diagnosed after excluding infectious, metabolic, and neoplastic etiologies, with confirmation based on clinical response to immunosuppressive therapy. All dogs received pharmacologic treatment appropriate for their underlying conditions as determined by the resident veterinarian in consultation with a faculty member in internal medicine.

Comparison of CBC parameters between anemia of inflammatory disease and hemolytic anemia, and among hemolytic anemia subgroups

A subset of anemic dogs with hemolytic disorders, such as IMHA or babesiosis, can present with a non-regenerative response, suggesting that inflammation may contribute to bone marrow suppression of erythropoiesis, a pattern that was also observed in the AID cases from our study [29, 46, 48]. To investigate whether such non-regenerative or delayed regenerative responses were associated with inflammatory indicators, we compared CBC parameters between dogs with AID and those with hemolytic anemia, as well as between the regenerative and non-/delayed-regenerative subgroups within the hemolytic anemia group. For AID cases, the first available test during the study period was used, whereas for hemolytic anemia cases, data obtained at the onset of anemia were analyzed. Statistical analyses were performed using the built-in statistical functions of GraphPad Prism (version 8.0.1; GraphPad Software, San Diego, CA, USA). Mean values were compared between groups using unpaired two-sample t-tests. All tests were two-tailed, and a P-value of <0.05 was considered statistically significant. At the date of anemia onset in hemolytic cases, any unavailable CBC parameter was coded as missing; CRP values reported as “under” were also coded as missing, whereas “over” results were top-coded at 10.

Analysis of cPLI and CRP to evaluate the impact of pancreatitis on reticulocytosis

Pancreatitis is a common inflammatory condition in dogs and is associated with hematologic alterations, including anemia [45]. Therefore, cPLI was included in the multivariate LMM to account for potential confounding and to capture sources of inflammation that may not be fully reflected by CRP alone. For this analysis, only dogs with at least three consecutive serum cPLI measurements during the study period were included, and these cases were analyzed in the same manner as described above to assess the longitudinal relationship between cPLI concentrations and reticulocyte counts.

RESULTS

Demographic information and data collection

The demographic information is summarized in Supplementary Table 1. The underlying diseases of all patients were defined based on the associated status at study inclusion. Among canine patients who visited the referral hospital during the retrospective study period, 235 were initially screened for inclusion as anemic. First, 116 dogs were excluded due to insufficient follow-up samples. Second, 89 dogs were excluded for failing to meet the inclusion criterion of remaining non-regenerative for at least 5 days after the onset of anemia. Third, seven patients were excluded for receiving chemotherapeutic agents during the delayed reticulocytosis period. As a result, 23 dogs met the inclusion criteria and were included in the final analysis. Of the 23 dogs included in the final analysis, eight received darbepoetin. Among those diagnosed with pancreatitis, only six had adequate serial data for inclusion. The breeds included: Pomeranian (n=5), Cocker Spaniel (n=1), Labrador Retriever (n=1), Border Collie (n=1), Welsh Corgi (n=1), Poodle (n=4), Maltese (n=6), Dachshund (n=1), Mixed breed (n=1), Jindo (n=1), and Papillon (n=1). The age range was 2–18 years, with a median age of 10 years. The sex distribution was castrated males (n=11), intact males (n=2), spayed females (n=8), and intact females (n=2). All dogs had at least one underlying inflammatory condition. The most common inflammatory diagnosis was pancreatitis, present in 56.5% of cases. Hepatobiliary diseases, including cholangitis and acute hepatitis, were identified in 17.4% of cases. Urinary tract inflammation (bacterial cystitis, pyelonephritis, polypoid cystitis) was observed in 13.0% of patients. Dermatologic inflammation (chronic atopic dermatitis, otitis externa) accounted for 8.7%, and pyometra for 8.7%. Meanwhile, 45 dogs were strongly suspected of having AID, of which six (13%) had anemia and normal CRP at the first visit.

Morphological classification and anemia characteristics

Among the 23 dogs included in this study, 52% exhibited normocytic, normochromic anemia, whereas 48% presented with microcytic, normochromic anemia. The non-regenerative status persisted for 5–267 days, with a median duration of 10 days. At the first examination, the mean hemoglobin (HGB) was 11.8 g/dL (SD=2.59), the mean hematocrit was 32.9% (SD=7.64), the red blood cell (RBC) count averaged 5.32 × 10⁶/µL (SD=1.25), the red cell distribution width–coefficient of variation (RDW–CV) averaged 17.63% (SD=3.29), and the reticulocyte count averaged 37.21 × 10³/µL (SD=24.46); the mean corpuscular volume (MCV) and mean corpuscular hemoglobin concentration (MCHC) averaged 62 fL (SD=4.36) and 35.9 g/dL (SD=1.39), respectively (Fig. 2; Supplementary Table 2). Blood smear examination of all anemic dogs confirmed the absence of hypochromic RBCs, poikilocytosis, RBC subpopulations (e.g., Heinz bodies), or other abnormalities (e.g., platelet clumping). Meanwhile, NMB staining revealed punctate and aggregated reticulocytes, confirming that no false-positive results were detected in the automated reticulocyte counts.

Fig. 2.

Distribution of red blood cell indices for dogs enrolled in this study. Dot plots of the distribution of RDW–CV, RETIC#, MCV, and MCHC at the first examination included in the study. Each dot represents an individual case, and red bars indicate the mean ± SEM. RDW–CV, red cell distribution width–coefficient of variation; RETIC#, reticulocyte count; MCV, mean corpuscular volume; MCHC, mean corpuscular hemoglobin concentration; SEM, standard error of the mean.

Univariate LMM results

The regression coefficients and P-values for the independent variables in relation to the dependent variable, reticulocyte count, are presented in Table 1 and Fig. 3. Several parameters showed statistically significant, moderate associations with reticulocyte count, including CRP (standardized β=−0.53; P<0.001), lymphocyte count (β=0.32; P=0.001), and platelet count (β=0.37; P<0.001) (Fig. 4). However, darbepoetin administration, monocyte count, reticulocyte–hemoglobin content, MPV, MCH, neutrophil-to-monocyte ratio (NMR), hematocrit (HCT), RBC count, HGB, neutrophil count, and white blood cell (WBC) count were not significantly associated with reticulocyte count. Meanwhile, in anemic dogs with pancreatitis, cPLI was not correlated with reticulocyte count (β=−0.29; P=0.09; Table 1).

Table 1. Univariate linear mixed model results for associations between standardized reticulocyte count and complete blood count parameters, C-reactive protein concentration, and darbepoetin administration in 25 cases.

Variable	Standardized β	Standard error	DF	t-value	P-value	95% CI lower	95% CI upper
Darbepoetin administration	−0.39	0.3	23	−1.31	0.2	−1	0.23
RDW-CV	1.16	0.12	144	9.51	<0.001***	0.92	1.4
CRP	−0.53	0.08	146	−6.63	<0.001***	−0.68	−0.37
Plateletcrit	0.46	0.08	146	5.51	<0.001***	0.29	0.62
MCHC	−0.5	0.09	146	−5.8	<0.001***	−0.67	−0.33
Platelet count	0.37	0.09	146	4.07	<0.001***	0.19	0.55
MLR	−0.22	0.08	146	−2.63	0.01*	−0.39	−0.06
Eosinophil count	0.18	0.084	146	2.1	0.04*	0.01	0.34
Lymphocyte count	0.32	0.1	146	3.3	0.001**	0.13	0.5
MCH	−0.15	0.13	146	−1.2	0.23	−0.4	0.1
NLR	−0.22	0.09	146	−2.61	0.01*	−0.39	−0.05
MPV	0.22	0.11	146	1.92	0.06	−0.007	0.44
MCV	0.38	0.14	146	2.69	0.01*	0.1	0.65
Monocyte count	0.01	0.09	146	0.09	0.93	−0.17	0.19
Retic-HGB	−0.09	0.09	146	−0.97	0.33	−0.27	0.09
NMR	−0.02	0.08	146	−0.24	0.81	−0.18	0.14
HCT	0.19	0.101	146	1.86	0.06	−0.01	0.39
RBC count	0.13	0.1	146	1.32	0.19	−0.07	0.33
HGB	0.11	0.1	146	1.05	0.29	−0.09	0.3
Neutrophil count	0	0.09	146	0.004	1	−0.18	0.18
WBC count	0.05	0.1	146	0.54	0.59	−0.14	0.24
cPLI concentration†	−0.29	0.16	29	−1.75	0.09	−0.62	0.05

*P<0.05, **P<0.01, and ***P<0.001. ^†Statistical analysis of cPLI concentration was conducted in six cases only. RDW-CV, red cell distribution width–coefficient of variation; NLR, neutrophil-to-lymphocyte ratio; MCV, mean corpuscular volume; MLR, monocyte-to-lymphocyte ratio; MCHC, mean corpuscular hemoglobin concentration; CRP, C-reactive protein; MPV, mean platelet volume; HCT, hematocrit; RBC, red blood cell count; HGB, hemoglobin concentration; WBC, white blood cell count; NMR, neutrophil-to-monocyte ratio; Retic-HGB, reticulocyte hemoglobin content; MCH, mean corpuscular hemoglobin; cPLI, canine pancreatic lipase immunoreactivity; CI, confidence interval; DF, degrees of freedom.

Fig. 3.

Associations between reticulocyte count and various hematological and inflammatory variables. Forest plot of the standardized regression coefficients (β) with 95% confidence intervals for the associations between reticulocyte count and selected hematological and inflammatory parameters. Variables such as RDW–CV, plateletcrit, platelet count, eosinophil count, lymphocyte count, NLR, MCV, MLR, MCHC, and CRP were significantly correlated with reticulocyte count. Statistical significance was adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate procedure, with adjusted P-values reported as q-values. Black circles indicate variables that remained statistically significant after correction, whereas white circles indicate those that were not significant after correction. RDW–CV, red cell distribution width–coefficient of variation; NLR, neutrophil-to-lymphocyte ratio; MCV, mean corpuscular volume; MLR, monocyte-to-lymphocyte ratio; MCHC, mean corpuscular hemoglobin concentration; CRP, C-reactive protein; MPV, mean platelet volume; HCT, hematocrit; RBC, red blood cell count; HGB, hemoglobin concentration; WBC, white blood cell count; NMR, neutrophil-to-monocyte ratio; Retic–HGB, reticulocyte–hemoglobin content; MCH, mean corpuscular hemoglobin.

Fig. 4.

The associations between standardized reticulocyte counts and C-reactive protein, lymphocyte counts, and platelet counts. Results of the univariate LMM analyses for standardized CRP, standardized lymphocyte count, and standardized platelet count are presented as scatter plots. (A) Reticulocyte counts are significantly negatively correlated with CRP concentrations. (B) A weak but positive correlation is observed between standardized reticulocyte counts and lymphocyte counts. (C) Standardized platelet counts exhibit a modest positive correlation with reticulocyte counts. Shaded areas indicate 95% confidence intervals. CRP, C-reactive protein; LMM, linear mixed model.

Multivariate LMM results

Of the 23 dogs, six were included in the longitudinal relationship analysis between cPLI concentrations and reticulocyte counts. In the multivariate LMM, with reticulocyte count as the dependent variable, and CRP and cPLI as independent variables, results for cPLI (β=−0.3; P=0.1) and CRP (β=−0.6; P<0.001) indicated that the association between cPLI and reticulocyte count remained statistically nonsignificant (Table 2).

Table 2. Multivariate linear mixed model results for associations between standardized reticulocyte count and C-reactive protein and canine pancreatic lipase immunoreactivity concentrations.

Variable	Standardized β	Standard error	DF	t-value	P-value	95% CI lower	95% CI upper
cPLI	−0.3	0.18	28	−1.6	0.12	−0.67	0.08
CRP	−0.6	0.14	28	−4.28	<0.001***	−0.88	−0.31

***P<0.001. cPLI, canine pancreatic lipase immunoreactivity; CRP, C-reactive protein; CI, confidence interval; DF, degrees of freedom.

Cross-sectional association between reticulocyte count and CRP level at the onset of anemia

Previous studies found no association between inflammation, measured by CRP levels, and reticulocyte counts [26, 33], but did find a significant association with reticulocyte maturation, indicated by the IRF [33]. However, in our study, a cross-sectional analysis of CRP concentration and reticulocyte count at the initial examination demonstrated a moderate negative correlation (Pearson’s r=−0.5; P=0.01). Conversely, no association was observed between reticulocyte count and reticulocyte–hemoglobin content (Pearson’s r=0.15; P=0.5). Furthermore, when dogs were classified into mild and severe inflammation groups based on a CRP threshold of 5 mg/dL [22], both reticulocyte count and reticulocyte–hemoglobin content were significantly lower in the severe group. The severe group had a significantly lower reticulocyte count than the mild group (mean ± SD: 27.44 ± 18.8 versus 51.9 ± 27.6 × 10³/μL; P=0.04); reticulocyte–hemoglobin content was also significantly lower in the severe group (mean ± SD: 22.5 ± 4.8 vs. 26.2 ± 2.8 pg; P=0.02).

Impact of CRP levels on reticulocyte response to darbepoetin

Eight dogs received darbepoetin during the analysis period; two developed reticulocytosis within three days of administration, three responded within seven days, and the remaining three did not respond to darbepoetin. To assess the influence of inflammation further, we compared the timing and magnitude of reticulocyte responses between these eight dogs with elevated CRP levels and three dogs with non-inflammatory disease and normal CRP levels. The clinical characteristics of three control dogs are provided in Supplementary Table 3. In dogs with elevated CRP concentrations (an average of 6.15 mg/dL), the onset of reticulocytosis was delayed by approximately twofold, and the peak reticulocyte count increase was reduced by ~38% compared to those with normal CRP levels (Table 3).

Table 3. Comparison of reticulocyte response between dogs with high and normal C-reactive protein concentrations.

Group	Time to reticulocytosis (days, mean ± SD)	Magnitude of reticulocytosis (mean ± SD)
High CRP (n=8)	12 ± 12.7	164 ± 42.3
Normal CRP (n=3)	6 ± 2.6	266 ± 25.8

CRP, C-reactive protein.

Analysis of CRP elevation and regeneration in hemolytic anemia

We analyzed the proportion of patients with regenerative hemolytic anemia associated with IMHA or babesiosis who had elevated CRP and examined the related regeneration patterns. Among 47 dogs diagnosed with hemolytic anemia, 39 (83.0%) had elevated CRP levels, including 18 with IMHA and 29 with babesiosis. Of these 47 dogs, 42 (89.4%) had regenerative anemia, 3 (6.4%) had non-regenerative anemia, and 2 (4.3%) exhibited delayed reticulocytosis. In our dataset, most (42/47, approximately 89%) showed a typical regenerative response, whereas the remaining non- or delayed-reticulocyte cases were associated with babesiosis. One-way ANOVA revealed statistically significant differences in RBC, HCT, HGB, MCV, reticulocyte count, lymphocyte count, and platelet count among groups, while monocyte count showed a tendency toward significance (P=0.05) (Supplementary Table 4). Compared with AID, hemolytic anemia cases were characterized by more severe anemia, higher reticulocyte counts (approximately 200,000/µL versus 37,211/µL at admission), and more pronounced macrocytosis and anisocytosis. Within the hemolytic group, the regenerative subset had significantly higher lymphocyte and monocyte counts than non-regenerative hemolytic cases.

DISCUSSION

Although inflammation has long been implicated in suppressing erythropoiesis and causing non-regenerative anemia in dogs, clinical evidence supporting this association remains limited. This study addressed this gap by demonstrating a link between inflammation and impaired erythropoiesis through three complementary lines of evidence: (i) serial CRP and reticulocyte counts were negatively correlated, (ii) higher CRP levels were significantly associated with lower reticulocyte counts, and (iii) darbepoetin responsiveness was reduced in anemic dogs with inflammation.

Our longitudinal analysis demonstrated a negative association between CRP concentration and reticulocyte counts, supporting the hypothesis that inflammation is associated with impaired bone marrow erythropoiesis in anemic dogs. Although impaired erythroid proliferation and reticulocyte maturation have been reported in human and canine AID studies, clinical evidence for these mechanisms remains limited [5, 37]. Two prior canine studies that attempted to address this issue reported conflicting results: one reported impaired reticulocyte maturation without a reduction in reticulocyte numbers [33], whereas the other found no difference in IRF between inflammatory and non-inflammatory states [26]. However, neither demonstrated a decrease in absolute reticulocyte counts associated with inflammation. This discrepancy may largely reflect differences in study populations: one investigation included relatively few dogs with inflammatory diseases [26], whereas most dogs in the other were non-anemic [33], which may have limited the ability to detect a negative association. Additionally, methodological differences may explain these divergent findings. For example, most prior veterinary studies have largely relied on cross-sectional analyses, which capture only a single time point and do not account for within-individual temporal dynamics [47]. Such reliance on isolated measurements may lead to misinterpretation, e.g., by underestimating temporal changes and thereby failing to achieve statistical significance [4, 52]. Conversely, longitudinal analysis evaluates within-subject changes over time [4], providing a more accurate evaluation of time-dependent associations, particularly for inflammatory markers [11, 17]. Supporting our approach, longitudinal studies in chronically anemic patients have evaluated relationships between inflammatory and hematological parameters, such as CRP, interleukin (IL)-6, and tumor necrosis factor-α (TNF-α), during the recovery period [2, 24]. Interestingly, our longitudinal analysis was in line with our cross-sectional findings, consistently demonstrating a negative association between CRP concentration and reticulocyte count. This concordance strengthens the robustness of our observations, overcomes the limitations of cross-sectional interpretation, and helps reconcile discrepancies reported in previous studies. In summary, our study uniquely focused on anemic dogs with concurrent inflammatory diseases, stratified cases by anemia type, and applied longitudinal models to characterize the suppressive effects of inflammation on erythropoiesis, thereby highlighting both novelty and clinical significance.

In clinical practice, our findings highlight two key implications. First, simultaneous monitoring of serial CRP levels and reticulocyte counts can provide a practical tool for tracking hematopoietic recovery in anemic dogs with inflammatory conditions. Second, control of inflammation may be important when interpreting erythropoietic recovery and responses to erythropoietin (EPO) therapy. While these clinical implications should be interpreted cautiously given the small sample size, concurrent monitoring of CRP concentrations and reticulocyte counts may still represent a reasonable approach in dogs with AID. Our findings suggest that inflammation may be an important factor contributing to a blunted response to darbepoetin, a notion supported by a recent finding in cats [8].

Inflammation may suppress erythropoiesis through multiple mechanisms. Proinflammatory cytokines inhibit erythroid proliferation and maturation in humans and animals [5, 18, 44]. For example, TNF-α promotes enhanced erythrophagocytosis by activated macrophages [36] and, together with IL-1 and IL-6, contributes to increased hepcidin production, leading to functional iron deficiency [38]. Interferon (IFN)-γ and TNF-α also reduce EPO synthesis and inhibit erythroid precursor proliferation and differentiation [1, 14]. IFN-γ in particular has been shown to induce EPO hypo-responsiveness in bone marrow progenitors by inhibiting erythroid proliferation and differentiation [1]. Consistently, in vitro studies using feline and murine bone marrow cells have shown that IFN-γ is a major suppressor of erythropoiesis [1, 28, 50]. Currently, experimental clarification on which specific inflammatory cytokines predominantly suppress hematopoiesis in dogs remains unclear. Future work should focus on identifying the key cytokines that drive the suppressive phenotypes observed in our study, thereby improving understanding of AID pathophysiology. Alternatively, identifying these inflammatory cytokines may serve as useful ancillary markers for detecting AID in dogs, particularly in cases with low-grade inflammation, in which CRP levels remain within the reference range. Indeed, IL-6, TNF-α, and IFN-γ may be more sensitive than CRP in chronic, low-grade inflammatory conditions [42]. In our study, approximately 13% of dogs with suspected AID and anemia had normal CRP levels at first presentation, underscoring the need for additional inflammatory markers.

In anemic dogs, elevated CRP likely reflects inflammation-mediated bone marrow suppression [5], which in this study was associated with non-regenerative or delayed regenerative responses. Interestingly, this pattern was supported by only a limited number of cases among dogs with hemolytic anemia. Several pathophysiological mechanisms may underlie the relatively low frequency of delayed reticulocytosis observed in cases of hemolytic anemia. First, hemolysis rapidly reduces erythrocyte mass, causing tissue hypoxia, and robust stimulation of renal EPO production and secretion [20, 34]. Consistently, anemic dogs with hemolysis or reduced RBC counts show markedly elevated serum EPO levels [37, 39, 40]. Indeed, in our study, dogs with hemolytic anemia had significantly lower RBC mass and higher reticulocyte counts than dogs with AID, suggesting a stronger EPO drive. Second, increased EPO levels could stimulate erythroferrone production, which might inhibit hepatic hepcidin production and enhance systemic iron absorption in dogs; thus, iron availability increases, and a more robust regenerative response can be mounted, even in the presence of inflammation [25]. Third, hemoglobin and heme released during erythrocyte destruction provide an iron source that can be rapidly and readily recycled and utilized for new erythropoiesis [10]. Finally, erythroid precursors likely remain primed rather than inflammation-induced desensitized, enabling rapid differentiation and proliferation upon EPO stimulation [53]. Overall, these mechanisms may counteract the inhibitory effect of inflammation on reticulocytosis in dogs with hemolytic anemia. Conversely, approximately 17% of dogs with babesia-induced hemolytic anemia were non-regenerative or showed delayed reticulocytosis. Although previous studies have proposed direct bone marrow suppression by the parasite and parasite-derived inhibitory factors, we were unable to confirm these mechanisms [48]. One clue that could differentiate these two subsets was lymphocyte count: compared with regenerative cases, non-regenerative or delayed cases had significantly lower lymphocyte counts, which might reflect cortisol-mediated lymphopenia associated with active disease progression, thereby delaying bone marrow erythropoiesis [23]. Conversely, marginal lymphocytosis observed in regenerative cases might indicate catecholamine-driven lymphocyte mobilization and a transition to a more chronic, less severe disease state [3]. Further research is needed to determine whether hepcidin, inflammatory cytokines, or other molecular pathways, depending on disease severity and progression, contribute to the impaired erythropoiesis observed in these subsets of anemic dogs with babesiosis.

Meanwhile, we identified a significant positive correlation between serial changes in reticulocyte and lymphocyte counts, which may, at least in part, reflect the waning effect of endogenous glucocorticoids as inflammation resolves in dogs under stress. A similar but weaker trend was observed between reticulocyte and platelet counts, potentially indicating the restoration of multilineage hematopoiesis following inflammatory resolution [43]. In dogs with pancreatitis, a mild negative correlation between cPLI concentration and reticulocyte count was noted but did not reach significance, likely due to the limited sample size. Moreover, as cPLI reflects pancreas-specific acinar injury, whereas CRP indicates systemic inflammation, their trajectories may diverge under concurrent conditions [27]. Future studies with larger cohorts are warranted to clarify these associations.

This study has several limitations. First, only 23 of 235 anemic dogs were included in the final analysis. Most exclusions were due to insufficient serial data required for longitudinal evaluation, indicating that CRP and reticulocyte counts have not been routinely assessed in a coordinated manner. Consequently, the study population represents a selected subset of cases in which both parameters were measured over time, which may have introduced selection bias; therefore, the results should be interpreted with caution. Future prospective studies incorporating standardized and synchronized measurements of CRP and reticulocyte counts across a broader and more representative population of anemic dogs might be necessary to improve the generalizability of these findings. Second, we acknowledge that the lack of iron metabolism data in the study population may limit the interpretation of AID. However, iron evaluations are not always a reliable gold standard for diagnosing AID or inflammatory anemia, because markers are significantly influenced by inflammation [32]. In human medicine, CRP and other inflammatory biomarkers are recommended to differentiate functional from absolute iron deficiency in chronic inflammatory conditions [15]. Furthermore, hepcidin has shown limited reliability for distinguishing AID, while markers such as CRP, IL-6, and IL-10 better reflect inflammatory status [21]. Similarly, dogs with AID often show only mild decreases in serum iron and transferrin saturation due to sequestration, and normal iron levels do not exclude inflammation [31]; thus, CRP may be a more sensitive indicator, consistent with previous studies [21, 33]. Additionally, diagnostic tools for assessing iron or hepcidin levels were generally unavailable in small animal practice [32]. Overall, inflammatory markers, including CRP, may be acceptable surrogates for evaluating inflammatory anemia in dogs, in whom iron assessment is often limited. Third, we acknowledge that our cohort included dogs with diverse inflammatory conditions and varying degrees of inflammation, which may have introduced confounding effects on erythropoietic suppression. Currently, a clear association between the severity of inflammatory disease and reticulocyte counts has not been established. However, a previous study has demonstrated that various inflammatory conditions suppress erythropoietic activity [33]. Accordingly, future studies are warranted to evaluate reticulocyte responses according to the severity of inflammation within more homogeneous populations defined by specific inflammatory diseases or conditions. Fourth, irregular sampling intervals across patients may have introduced temporal bias. Given the retrospective nature of the study, uniform control of visit intervals was not feasible. To mitigate this limitation, data points with intervals exceeding 4 days immediately after the onset of anemia were excluded, as the timing of reticulocytosis in such cases would be difficult to interpret.

Finally, CRP, as an acute-phase protein, may not fully capture the chronic nature of AID, which represents a downstream consequence of persistent inflammatory activity. Accordingly, CRP–reticulocyte correlations may be weakened or absent during the reticulocyte-lag phase, when CRP levels have already declined; however, reticulocyte counts remain low due to ongoing inflammation-mediated suppression of bone marrow erythropoiesis. Future studies should include this phase as a critical baseline period for molecular investigations to identify key cytokine dynamics (e.g., IL-6, TNF-α, IFN-γ) involved in inflammation-induced marrow inhibition and subsequent erythropoietic recovery.

In conclusion, our study demonstrates a negative temporal correlation between CRP concentration and reticulocyte count in dogs with AID, supporting the clinical utility of concurrent serial monitoring of these parameters to assess inflammatory resolution and erythropoietic recovery. Delayed reticulocytosis following treatment may reflect inflammation-associated suppression of erythropoiesis, highlighting the importance of considering inflammatory status when interpreting treatment responses.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Funding Statement

This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (Ministry of Science and ICT) (RS-2026-25469201).