Epidemiology of Canine Cutaneous Histiocytoma: Insights From Age and Topographic Clustering

ABSTRACT

Background

Canine cutaneous histiocytoma (CCH) is a benign Langerhans‐cell tumour that often regresses spontaneously via inflammation.

Objectives

To estimate incidence and characterise age and breed–anatomical‐site patterns using two complementary clustering analyses.

Materials and Methods

A total of 359 histopathologically‐confirmed CCHs from a private veterinary laboratory in Portugal (January 2020–January 2022) were analysed with dog‐population denominators from the national registry (SIAC). Incidence rates (IR) were calculated for Lisbon–Setúbal; breed‐ and sex‐specific relative risks (RR) were estimated. Age was modelled with BIC‐selected Gaussian mixtures; breed–site patterns were summarised by hierarchical clustering of site frequencies (heatmaps).

Results

Cases concentrated in young dogs (≤ 2 years, 48.8%) and males (58.5%). Age clustering defined three bands (≈ ≤ 1.5, 1.5–3.5, > 3.5 years) with differing breed and site distributions. Breed–site clustering revealed recurring anatomical predilections by breed and sex. The annual IR in Lisbon–Setúbal was 4.0 per 10,000 dogs (3.5 in females; 4.4 in males). Risk was highest in French bulldogs (RR 11.5; IR 34.1) and Boxers (RR 8.3; IR 24.5) versus mixed‐breed dogs (p < 0.001).

Conclusions and Clinical Relevance

Canine cutaneous histiocytoma (CCH) shows clear age bands and breed‐linked anatomical predilections; these patterns help prioritise ages, sites and breeds for studies of inflammation‐mediated regression.

Background: Canine cutaneous histiocytoma (CCH) is a benign Langerhans‐cell tumour that often regresses spontaneously via inflammation. Objectives: To estimate incidence and characterise age and breed–anatomical‐site patterns using two complementary clustering analyses. Conclusions and Clinical Relevance: CCH shows clear age bands and breed‐linked anatomical predilections; these patterns help prioritise ages, sites and breeds for studies of inflammation‐mediated regression.

1. Introduction

Canine cutaneous histiocytoma (CCH) is a common benign cutaneous tumour arising from Langerhans cells (LCs) [1]. LCs are antigen‐presenting cells that modulate cutaneous immune responses [2]. Histiocytes derive from CD34+ precursors and differentiate into macrophages, interstitial dendritic cells (DCs) and intraepithelial DCs such as LC [3] which reside in the skin and mucosal epithelia [4]. Given its inflammation‐mediated regression, CCH offers a natural setting in which population‐level patterns can generate mechanistic hypotheses; the present work leverages denominator‐based incidence and complementary clustering analyses to translate biology into testable epidemiological structure.

Canine cutaneous histiocytoma (CCH) can occur at any age, yet its prevalence decreases markedly after 3 years [3]. Nonetheless, histiocytomas in senior dogs are not uncommon [3], and there is no convincing evidence of major sex‐related differences [2]. Macroscopically, CCH typically presents as a solitary, rapidly growing, well‐demarcated nodule measuring ~0.2–3.0 cm [1], most often on the head, ears, neck and limbs, although any cutaneous site may be involved [1].

Although often regarded as non‐breed‐specific [3] several studies suggest possible breed predispositions. Shar‐pei have been associated with persistent or recurrent CCH [3]. Increased odds have been reported in: Labradors, French and English bulldogs, Scottish terriers, and Boxers; [5] in Staffordshire bull terriers and cocker spaniels; [6] and, more recently, in French bulldogs, Boxers, pugs, Jack Russell terriers and beagles [7].

CCHs typically undergo spontaneous regression within ~2–3 months [1]. Proposed mechanisms include LC migration to draining lymph nodes, activation of CD4+ T lymphocytes and recruitment of cytotoxic CD8+ T cells [2, 8, 9, 10]. A similar lesion, LC histiocytoma, occurs in infants and regresses spontaneously [11, 12, 13], suggesting shared biological processes. A notable interspecies distinction is the presence of Birbeck granules (BGs) in humans, mice, cats and other species that express functional langerin [4, 14, 15, 16], and their absence in dogs [3], In humans, BGs aid LC identification and pathogen capture via langerin [17, 18, 19, 20, 21]. Despite this ultrastructural difference, tumour behaviour is broadly comparable across species, supporting comparative hypotheses on immune–tumour interactions and regression.

Key epidemiological gaps remain: population‐based incidence estimates and a coherent description of how age and anatomical site vary across breeds. Using registry denominators for a defined catchment (Lisbon–Setúbal) enables sex‐ and breed‐specific rate comparisons.

Thus, this study aimed to expand epidemiological knowledge of histiocytomas by characterising cases diagnosed at DNAtech Veterinary Laboratory (Lisbon, Portugal), quantifying incidence using population denominators, and describing structured patterns in age‐ and breed‐linked anatomical distribution to provide evidence that can guide subsequent mechanistic research.

2. Materials and Methods

2.1. Ethics Statement

The study was approved by the Animal Welfare Agency of the University of Évora (ÓRBEA‐XX GD 27574/2024/P1).

2.2. Data Collection

Cases of CCH diagnosed between January 2020 and January 2022 were retrieved from the DNAtech Veterinary Histopathology Laboratory (Lisbon). Available variables included age, breed, sex, tumour site, tumour diameter and the geographical location of the submitting veterinary surgeon. Anatomical sites were coded a priori into nine categories: four head sites (head, NOS [not otherwise specified], external ear, eyelid, perilabial region), limbs (excluding digits), digits, perineal, perianal, trunk and unknown site. The separation of ‘limbs (excluding digits)’ and ‘digits’ was defined hierarchically, assigning lesions of toes/phalanges to ‘digits.’ ‘Unknown site’ was retained as its own category for descriptive summaries and excluded from clustering analyses.

2.3. Canine Population Data

Dog population denominators for Lisbon and Setúbal were obtained from the Portuguese companion‐animal registry (SIAC) [22]. These districts constituted the majority of submissions and therefore defined the catchment for rate estimation. The average annual incidence rate (IR) was computed as:

$$\mathit{IR}=\frac{\mathit{\text{number of cases over the study period}}}{\mathit{dog}\mathit{\text{population}}\times \mathit{\text{years observed}}}\times \mathrm{10,000}$$

District‐specific IRs were calculated overall and by sex. Relative risks (RR) were estimated by comparing breed‐specific IRs with those of mixed‐breed dogs, and 95% confidence intervals (95% CI) were calculated. In the absence of official 2020–2022 district counts, denominators were derived from the official 2023 SIAC registry [23], under the assumption of limited annual variation in the dog population.

2.4. Statistical Analysis

Data were validated in excel 2013 and analysed in R v4.1.2. Continuous variables are presented as mean ± standard deviation (SD) and compared with Student's t‐test (two groups) or ANOVA with Tukey's post hoc test (three or more groups), as appropriate. Categorical variables are reported as counts and percentages. Sex distribution was evaluated against a 50% null using an exact binomial test (two‐sided). For multiple pairwise comparisons, Holm adjustment was applied.

Patterns linking breeds and anatomical sites were explored using hierarchical clustering heatmaps ( pheatmap package) of site frequencies by breed, and visualised as clustered heatmaps with dendrograms. Rows (breeds) were mean‐centred and row‐scaled (z‐scores) to emphasise relative predilections; the distance metric was 1 − Pearson correlation, and Ward. D2 linkage was used to form clusters. Breeds with > 10 cases were included in clustering.

Age structure was examined with Gaussian finite mixture models ( mclust package). The number of components was selected by the Bayesian information criterion (BIC); component means and SDs were estimated, and cluster membership was assigned by maximum a posteriori probability. The resulting age clusters (approximately ≤ 1.5, 1.5–3.5 and > 3.5 years) were used for descriptive cross‐tabulations and visualised with kernel density plots (overlaid by cluster). Associations between age clusters and breed, sex, or anatomical site were assessed with chi‐square tests. All tests were two‐sided with α = 0.05.

3. Results

The dataset comprised 359 CCH cases (age 2 months–13 years; mean 3.7, SD 3.0) (Table 1) of which 48.8% occurred in dogs ≤ 2 years (Figure 1). Frequency distribution revealed three phases of increased frequency, confirmed by cluster analysis which identified three distinct groups: ≤ 1.5 years, 1.5–3.5 years and > 3.5 years (Figure S1). Although the chi‐square test was not statistically significant (data not shown), some variation between clusters was evident.

TABLE 1.

Distribution of canine cutaneous histiocytoma (CCH) cases by sex, age, breed and anatomical location. Percentages for anatomical location are calculated among cases with known site (n = 241/359).

	Total		Female		Male
	n (%) in column	Mean age (SD)	n (%)in line	Mean age (SD)	n (%)in line	Mean age (SD)
Total	359 (100)	3.7 (3.0)	149 (41.5)	3.5 (2.8)	210 (58.5) A	3.8 (3.2)
Breed
Mixed‐breed	93 (25.9)	3.9 (3.0)b	44 (47.3)	4.2 (2.9)	49 (52.7)	3.7 (3.2)
French bulldog	86 (24.0)	3.1 (2.4)a	34 (39.5)	2.5 (2.4)	52 (60.5) d	3.4 (2.4)
Labrador retriever	25 (6.9)	4.7 (3.6)c	11 (44.0)	4.4 (3.6)	14 (56.0)	4.9 (3.7)
Boxer	20 (5.6)	4.1 (3.6)	9 (45.0)	3.2 (3.2)	11 (55.0)	5.0 (3.9)
Yorkshire terrier	14 (3.9)	4.9 (3.3)c	8 (57.1)	2.9 (2.2)	6 (42.9)	7.2 (3.0)
Jack Russell terrier	13 (3.6)	3.8 (2.8)	4 (30.8)	5.0 (3.6)	9 (69.2)	3.3 (2.6)
Beagle	10 (2.8)	2.1 (2.3)a	2 (20.0)	1.5 (0.7)	8 (80.0) d	2.3 (2.7)
Cocker spaniel	10 (2.8)	4.6 (4.0) c	3 (30.0)	2.6 (2.2)	7 (70.0)	5.4 (4.5)
Others	88 (24.5)	3.4 (3.1)	34 (38.6)	3.4 (2.7)	54 (61.4) d	3.4 (3.3)
Anatomical location
Head, NOS	72 (20.1)	3.7 (2.9)	28 (38.9)	3.0 (2.4)	44 (61.1)	4.2 (3.2)
External ear	32 (8.9)	3.3 (3.2)	14 (43.8)	2.7 (3.1)	18 (56.2)	3.8 (3.3)
Eyelid	8 (2.2)	3.4 (2.0)	5 (62.5)	2.8 (1.7)	3 (37.5)	4.3 (2.3)
Perilabial	16 (4.4)	5.1 (3.9)	8 (50.0)	3.3 (2.9)	8 (50.0)	6.6 (4.2)
Limbs, NOS	89 (24.8)	2.9 (2.6)	32 (36.0)	2.7 (2.3)	57 (64.0) d	3.0 (2.7)
Digits	10 (2.8)	4.2 (3.6)	3 (30.0)	3.0 (2.7)	7 (70.0)	5.0 (4.1)
Perianal	4 (1.1)	4.0 (3.4)	2 (50.0)	5.5 (5.0)	2 (50.0)	2.5 (0.7)
Trunk	10 (2.8)	3.2 (2.6)	6 (60.0)	2.3 (2.2)	4 (40.0)	4.4 (2.8)
No information	118 (32.9)	4.1 (3.2)	51 (43.2)	4.6 (3.1)	67 (56.8)	3.8 (3.3)

Note: Superscript letters (a–c) denote significant pairwise differences in mean age between groups (two‐sample Student's t‐tests with Holm adjustment); values sharing a letter do not differ (p ≥ 0.05). Full statistical details are available upon request.

Abbreviation: NOS, not otherwise specified.

^A Sex proportion tested against 50% using an exact binomial test; p < 0.05.

FIGURE 1.

Combined chart showing the relative frequency (bars) and cumulative frequency (dotted line) of CCH cases by age.

3.1. Sex Distribution

Canine cutaneous histiocytoma was more frequent in males (n = 210, 58.5%) than females (n = 149, 41.5%) (p < 0.05, exact binomial test), sex ratio 1.5:1. Mean age did not differ materially by sex (males 3.8 years, SD 3.2; females 3.5 years, SD 2.8). Male predominance also was evident within French bulldogs (p = 0.047) and beagles (p = 0.018) (Table 1). The distribution across age clusters was broadly similar by sex, with a modest over‐representation of females in the intermediate cluster; the chi‐square test for sex versus age clusters was not significant (p > 0.05) (Figure 2a) (data not shown).

FIGURE 2.

Kernel density ridgeline plot stratified by age clusters (≈ ≤ 1.5, 1.5–3.5, > 3.5 years). Plots depict densities for (a) sex, (b) breed and (c) anatomical location.

3.2. Breed Distribution

Forty one breeds were represented. Mixed‐breed dogs were most frequent (n = 93, 25.9%), followed by French bulldogs (n = 86, 24.0%) and Labradors (n = 25, 7.0%) (Table 1). Yorkshire terriers showed the highest mean age (4.9 years, SD 3.3), while beagles had the lowest (2.1 years, SD 2.3) (Table 1; Figure 3). Pairwise Student's t‐tests indicated beagles had a lower mean age than Yorkshire terriers (p = 0.01) and Labradors (p = 0.02); French bulldogs differed from mixed‐breed (p = 0.012), Cocker spaniels (p = 0.006) and Labradors (p = 0.008); mixed‐breed had a lower mean age than Yorkshire terriers (p = 0.001). These findings suggest earlier onset in beagles and French bulldogs, with mixed‐breed onset intermediate, and later onset in Labradors, Yorkshire terriers and cocker spaniels.

FIGURE 3.

Violin plot of age distribution by main breeds (n ≥ 10 per breed).

Clustering by age within breeds (Figure 2b) showed mixed‐breed and French bulldogs distributed across clusters; Labradors, Jack Russell terriers, cocker spaniels and beagles occurred mainly in the youngest cluster, with few cases in the intermediate cluster; Yorkshire terriers lacked cases in the youngest cluster. All breeds had cases in the oldest cluster except beagles.

3.3. Anatomical Distribution

Of the 359 cases, 241 (67.1%) had a recorded anatomical site; 118 (32.9%) lacked location data (Table 1). The limbs were the most common location (n = 89), followed by head (NOS) (n = 72). CCH in the perilabial region presented at a later mean age (5.1 years, SD 3.9), although this difference did not reach statistical significance. Lesions on the limbs were more frequent in males (p = 0.006; Table 1). By age clusters, eyelid cases were observed only in the intermediate cluster, while digit cases occurred only in the oldest cluster; perilabial cases were absent from the youngest cluster; and trunk cases appeared in the youngest and oldest clusters (Figure 2c).

3.4. Frequency of Anatomical Locations Across Breeds

Breed‐specific site patterns (Figure 4; Table S1) formed three qualitative groups: (1) Jack Russell terriers with a predominance of limb tumours; (2) cocker spaniels, Yorkshire terriers and beagles with a concentration in head (NOS) and/or external ear; and (3) Boxers, mixed‐breed, French bulldogs, and Labradors with more generalised distributions.

FIGURE 4.

Heatmap of the relative frequency of tumour locations across dog breeds. Darker shades indicate higher frequency. Breeds are clustered by similarity in site profiles (rows z‐scored; distance = 1—Pearson correlation; linkage = Ward. D2). Display includes breeds with n ≥ 10.

3.5. Lesion Diameter

The overall mean diameter was 1.6 cm (SD 1.6) (Table S2). Females had smaller lesions than males (1.3 cm, SD 1.2 versus 1.8 cm, SD 1.7; p = 0.002). Among breeds, Cocker spaniels had the largest mean diameter (2.4 cm, SD 1.8) and Yorkshire terriers the smallest (0.6 cm, SD 0.4); between‐breed differences were not significant. ANOVA indicated variation by site (p = 0.006): perianal lesions were largest (5.0 cm, SD 0), while eyelid lesions were smallest (0.4 cm, SD 0.4). In females, mean diameters were 5.0 cm (perianal), 2.7 cm (digits) and 0.7 cm (eyelid).

3.6. Geographical Distribution and Incidence

Lisbon accounted for 41.0% of cases (n = 147), followed by Setúbal (n = 71, 19.8%), Faro (n = 32, 8.9%), Santarém (n = 27, 7.5%) and Leiria (n = 26, 7.2%). In Lisbon–Setúbal, the IR was 4.0 per 10,000 dogs (3.5 in females; 4.4 in males) (Table S3). French bulldogs showed the highest risk (IR 34.1; RR 11.5, versus mixed breed; p < 0.001) (Figure 5), followed by Boxers (IR 24.5; RR 8.33; p < 0.001). For both breeds, effects were stronger in males (Figure 5).

FIGURE 5.

Forest plots of relative risk (RR) of CCH by breed in the Lisbon–Setúbal catchment. Left: All dogs; centre: Females; right: Males. Points show RR versus mixed‐breed (reference); horizontal lines indicate 95% confidence intervals. Only breeds with n ≥ 10 are displayed.

4. Discussion

This study shows that close to half of CCH cases are diagnosed by 2 years of age, and that three age clusters are apparent, with suggestive differences in breed representation and anatomical distribution across clusters. The limbs were the most frequent site among cases with known location, and breed‐specific site patterns were observed. French bulldogs and Boxers exhibited the highest incidence, with male predominance.

4.1. Age Distribution

The mean age of 3.7 years (range 2 months–13 years) is consistent with Aupperle‐Lellbach et al. [7], who reported a central tendency around 4.0 years (1–17 years), and with earlier reports that frequency declines after ~3 years [2, 3]. Kim et al. [6] described a peak at 6–9 months, reinforcing early‐life susceptibility. The three data‐driven age clusters (≈ ≤ 1.5, 1.5–3.5, > 3.5 years) provide a practical summary of heterogeneity and suggest (rather than prove) age‐linked differences in breed and site; taken together, these patterns motivate genetic and mechanistic studies of CCH pathogenesis and regression in both species.

4.2. Sex Distribution

In contrast to Gross et al. [2] who noted no sex differences, these data show male predominance (ratio 1.50), broadly in line with Kim et al. [6] (ratio 1.72) and Aupperle‐Lellbach et al. [7] The small difference in mean age between sexes (3.5 vs. 3.8 years) does not indicate a material shift in onset. Kim et al. [6] also reported lower odds in sterilised females than in castrated males, suggesting a potential role for reproductive status, which could not be assessed here as a consequence of missing data.

4.3. Breed Distribution

French bulldogs showed the highest prevalence in the sample, followed by Labradors. Although breed predisposition remains debated [2], prior studies reported increased odds in Labradors, bulldogs, Scottish terriers and Boxers [5], and in Staffordshire bull terriers and Cocker spaniels; [6] the study by Kim et al. [6] was restricted to 0–12 months and did not include French bulldogs, limiting comparability. Aupperle‐Lellbach et al. [7] reported similar patterns yet focused on odds ratios. By using IRs with population denominators, the present estimates reduce composition bias and strengthen the epidemiological signal; nevertheless, genetic predisposition should be considered a hypothesis, pending multicentre and genomic validation.

French bulldogs are a popular breed with well‐documented health issues [24]. The process of selective breeding has shaped genetic variation to produce recognisable breed‐specific traits; [25] breed‐specific incidence helps prioritise targets for genetic inquiry.

The Yorkshire terrier showed the highest mean age at diagnosis (4.9 years); this may reflect lifespan differences (longer average longevity reported for this breed [26, 27, 28]), although other explanations (e.g., care‐seeking, clinical recognition) are possible. Overall, broader breed coverage remains important to improve risk estimation across canine demographics.

4.4. Anatomical Location Distribution Across Breeds

Consistent with Gross et al. [2], head, ears, neck, and limbs were common sites; in this dataset, limbs predominated, followed by head (NOS) among cases with known site. Three descriptive breed‐site groups emerged: Jack Russell terriers with limb tumours; Cocker spaniels, Yorkshire terriers, and Beagles with head/ear lesions; and Boxers, mixed‐breed, French bulldogs, and Labradors with more generalised distributions.

Age‐linked patterns also were observed: digits occurred only among older dogs, eyelid lesions only in the intermediate cluster, perilabial lesions were absent in the youngest cluster, and trunk lesions appeared in the youngest and oldest clusters. Given limited counts and nonsignificant chi‐square tests, these should be interpreted as hypothesis‐generating rather than inferential claims. Whether CCH comprises subtypes with distinct regression profiles warrants targeted follow‐up.

4.5. Lesion Diameter Distribution

The mean diameter (1.6 cm) aligns with previous reports [1]. Males had, on average, larger lesions than females (~0.5 cm difference); sexual size dimorphism may contribute [29], yet specific evidence for this tumour is lacking. Cocker spaniels displayed the largest mean diameter (2.4 cm) and Yorkshire terriers the smallest (0.6 cm); the latter may relate to body size [30], while larger values in Cocker spaniels (a medium‐small breed) [31] remain unexplained despite falling within reported ranges [1]. Site differences were plausible: eyelid lesions were smallest, possibly as a consequence of early detection, while perianal lesions were largest, perhaps reflecting lower visibility or delayed presentation; numbers in some sites were small.

4.6. Lisbon and Setúbal Incidence Rates

Incidence was estimated for the first time in these districts, enabling sex‐ and breed‐specific comparisons using registry denominators. The annual rate was approximately four per 10,000 dogs, higher in males, and elevated in French bulldogs and Boxers. These patterns are compatible with a genetic component and justify focused clinical vigilance; however, causal attributions are premature without confirmatory studies that integrate genomics, environment, and care‐seeking behaviour.

4.7. Limitations

Some limitations merit emphasis. First, numerators comprised histopathologically‐confirmed cases from a single laboratory, which may under‐ascertain CCH that regress without biopsy confirmation and may reflect referral patterns. Secondly, 32.9% of cases lacked topography; site‐specific summaries were therefore restricted to cases with known location. Third, denominators used the official 2023 SIAC counts as an approximation of mid‐period population, which may introduce minor imprecision; rates should be interpreted accordingly. Finally, the absence of data on reproductive status and concomitant diseases, and detailed clinical histories may have influenced some comparisons, particularly because previous studies have reported an increased risk of CCH associated with long‐term use of Janus kinase inhibitors and ciclosporin [32, 33, 34]. Additionally, potential shifts in breed popularity over the study period could have affected incidence estimates. These constraints do not negate the findings yet do frame them as population‐based signals requiring validation in multicentre, prospective designs.

5. Conclusion

Distinct age‐cluster and breed‐related patterns in CCH occurrence were identified, including earlier onset in specific breeds and anatomical‐site predilections. These epidemiological signals provide a hypothesis‐generating framework for genetic, molecular and immunological studies of spontaneous regression. Focusing on breeds and age groups with the greatest incidence/risk should enable more precise dissection of the mechanisms underlying CCH regression.

Author Contributions

C. De Moura: conceptualisation; writing – original draft (lead); writing – review and editing. K. Pinello: conceptualisation; methodology; writing – original draft; data curation; formal analysis; writing – review and editing. J. Catarino: writing – review and editing. J. Niza‐Ribeiro: conceptualisation. J. Santos: supervision. P. Faísca: Conceptualisation, project administration, supervision.

Funding

This research was supported by the Faculty of Veterinary Medicine, Universidade Lusófona (Lisbon), and by the Vet‐OncoNet project (ICBAS, University of Porto).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Distribution of CCH cases per anatomical location and breeds.

Table S2: Distribution of canine cutaneous histiocytoma (CCH) cases by sex, breed, anatomical location and lesion diameter (cm). Lesion diameter is reported as mean ± standard deviation (SD).

Table S3: Breed‐specific IRs (cases per 10,000 dogs per year) and RRs of CCH in the Lisbon–Setúbal catchment. Mixed‐breed dogs serve as the reference (RR = 1). Estimates are reported with 95% CIs; results are shown overall (and by sex, where available). Denominators derive from the official 2023 SIAC registry.

Figure S1: Age clustering results. Left: BIC for univariate Gaussian mixture models with 1–9 components; “E” and “V” denote equal‐ and variable‐variance parameterisations. The BIC peaks at three components. Right: Age by cluster (1–3); boxplots show median, IQR and outliers, delineating three age clusters (~ ≤ 1.5, 1.5–3.5, > 3.5 years).

Data Availability Statement

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.