Repeatability and Reproducibility of Microscopic Examination of Adhesive Tape Strip Cytology Slides for the Diagnosis of Malassezia Overgrowth and/or Malassezia Dermatitis in Dogs

1. Introduction

Yeasts of the genus Malassezia, particularly M. pachydermatis, are part of the normal cutaneous microbiota of dogs [1]. Under conditions that disrupt cutaneous homeostasis, such as integumentary or systemic disease, these organisms may proliferate excessively (Malassezia overgrowth), leading to pruritus and inflammation (Malassezia dermatitis) [1,2]. In dogs with atopic dermatitis (AD), an additional mechanism has been described whereby humoral and/or cellular hypersensitivity to Malassezia spp. allergens exacerbates pruritus and skin lesions, even in the absence of yeast overgrowth [1,2,3,4,5,6,7]. Consequently, current guidelines recommend that detection of any number of Malassezia organisms on the skin of dogs with AD should prompt a therapeutic trial to evaluate their potential role as flare factors [8].

In clinical practice, the presumptive diagnosis of Malassezia overgrowth and/or dermatitis relies primarily on cytological examination, with adhesive tape strip cytology being the most commonly used sampling technique [1,2,9,10]. We previously demonstrated that diagnostic accuracy improves when slides are examined at 1,000× magnification and that the repeatability and reproducibility of yeast counts are poor, even after evaluation of 50 oil immersion fields (OIFs) [11]. However, the optimal number of OIFs required for reliable presumptive diagnosis of Malassezia overgrowth and/or dermatitis has not yet been established.

The aim of the present study was to determine the minimum number of OIFs that should be examined on adhesive tape strip cytology slides to achieve optimal intraobserver repeatability and interobserver reproducibility for the presumptive diagnosis of Malassezia overgrowth and/or dermatitis.

2. Materials and Methods

A total of 12 examiners participated in the study. Examiner 1 was a Diplomate of the European College of Veterinary Dermatology. Examiners 2–5 included a dermatology intern, clinic staff members, and PhD students with extensive experience in cutaneous cytology. Examiners 6–12 were less experienced and included three interns and four fourth- or fifth-year veterinary students. All examiners had participated in a previous study [11]; however, data from examiners 6–12 were not published because their participation had been intended solely for training purposes.

One week before the study, all examiners attended a 60-min training session conducted by examiner 1. A PowerPoint presentation was used to review the study objectives and design, the criteria for selecting OIFs, and the morphological features used for the identification of Malassezia spp. Examiners were instructed to evaluate OIFs located in the central area of the slide and containing moderate to high numbers of keratinocytes arranged in monolayers. The session concluded with a live demonstration of representative cytological preparations to standardize field selection and yeast identification using a microscope (BX-40 Olympus, Japan) equipped with an Altra 20 screen-projecting color camera (Olympus, Japan).

Twenty archived adhesive tape strip cytology slides obtained during routine clinical work-up were used. Approximately 5-cm strips of transparent adhesive tape (Scotch Crystal Clear Tape, 3Μ) had been pressed 5–10 times onto lesional skin from dogs suspected of having Malassezia overgrowth and/or dermatitis. Tape strips were stained with Diff-Quik (Merck, Germany) using the fixative and both staining solutions, air-dried, and mounted adhesive side down onto clean glass slides. Before the study, examiner 1 confirmed acceptable staining quality for all preparations. Slide identity was then masked, and slides were randomly numbered from 1 to 20 using an online random number generator (https://www.gigacalculator.com/calculators/random-number-generator.php).

Twenty high-quality clinical photographs showing lesions compatible with Malassezia dermatitis [1] were incorporated into a PowerPoint presentation. Each slide included one photograph and the text “Pruritus:” followed by either “yes” (12/20; 60%) or “no” (8/20; 40%). The primary diseases included AD (n = 5); demodicosis and flea allergic dermatitis (n = 2 each); and acral lick dermatitis, alopecia X, cyclic flank alopecia, dermatophytosis, digital trauma, follicular dysplasia, hyperadrenocorticism, hypothyroidism, leishmaniosis, muzzle folliculitis-furunculosis, and pemphigus foliaceus (n = 1 each). Photographs were cropped to minimize recognition of the primary disease.

The order of the photographs was randomized, and each image was assigned a number from 1 to 20 corresponding to a cytology slide bearing the same number. Therefore, cytology slides and photographs did not originate from the same dogs, preventing examiners from inferring the likely presence or absence of Malassezia organisms based on the clinical image.

Each examiner first viewed the clinical photograph and then examined the corresponding cytology slide at 1,000× magnification. The numbers of Malassezia yeasts observed within the first 10, 20, 30, 40, and 50 OIFs were recorded. Examiners then determined whether a presumptive diagnosis of Malassezia overgrowth and/or dermatitis was justified after examination of each number of OIFs according to the following criteria: (a) detection of at least one yeast organism in skin pruritic lesions, or (b) detection of an average of at least one yeast organism per OIF in non-pruritic skin lesions [12,13].

A second evaluation round was performed at least two weeks after completion of the first round. All cytological examinations were performed using the same microscope (Carl Zeiss ICS-KF2 Binocular Research Microscope, Jena, Germany).

Statistical analyses were performed separately for slides paired with photographs of pruritic and non-pruritic dogs. Cohen’s kappa coefficient (κ) was calculated for the binary outcome variable (presumptive diagnosis of Malassezia overgrowth and/or dermatitis or not): (a) between the two examination rounds of the same examiner for the same slide and number of OIFs (intraobserver repeatability), and (b) between all examiner pairs evaluating the same slide, examination round, and number of OIFs (interobserver reproducibility). Overall interobserver reproducibility was assessed using Fleiss’ κ. Agreement was interpreted as poor (κ ≤ 0.20), fair (κ: 0.21–0.40), moderate (κ: 0.41–0.60), substantial (κ: 0.61–0.80), or good (κ > 0.80) [14]. Also, κ values obtained after examination of 10, 20, 30, 40, and 50 OIFs were compared using Friedman repeated-measures ANOVA. When significant differences were detected, post-hoc Wilcoxon signed-rank tests with Bonferroni correction were performed. Statistical analyses were conducted using SPSS 29 for Windows, and significance was set at p < 0.05.

Because only archived adhesive tape strip cytology slides obtained non-invasively during routine diagnostic procedures were used and no additional animal handling or sampling was involved, ethical approval was not required.

3. Results

Among the 12 cytology slides paired with photographs of pruritic dogs, the overall prevalence of a presumptive diagnosis of Malassezia dermatitis ranged from 49.7% after examination of 10 OIFs to 68.8% after examination of 50 OIFs (Table 1). Among the eight slides paired with photographs of non-pruritic dogs, the prevalence of a presumptive diagnosis of Malassezia overgrowth and dermatitis ranged from 9.9% to 12% depending on the number of OIFs examined (Table 2).

3.1. Intraobserver Repeatability

For slides paired with photographs of pruritic dogs, κ values ranged from −0.143 to 1 (Table 1). Agreement was classified as poor in 8/60 (13.3%), fair in 16/60 (26.7%), moderate in 5/60 (8.3%), substantial in 16/60 (26.7%), and good in 15/60 (25%) examiner/number of OIF combinations (Table 3). Good repeatability was not achieved by 5/12 (41.7%) examiners at 10, 20, 30, 40, and 50 OIFs. Friedman repeated-measures ANOVA showed no significant differences in κ values among examinations of 10, 20, 30, 40, and 50 OIFs (p = 0.848).

For slides paired with photographs of non-pruritic dogs, κ values ranged from −0.20 to 1 (Table 2). Agreement was classified as poor in 31/60 (51.7%), fair in 1/60 (1.7%), moderate in 3/60 (5%), substantial in none, and good in 25/60 (41.7%) examiner/number of OIF combinations (Table 3). Good repeatability was not achieved by 4/12 (33.3%) examiners at 10, 20, 30, 40, and 50 OIFs. Friedman repeated-measures ANOVA again demonstrated no significant differences in κ values among examinations of 10, 20, 30, 40, and 50 OIFs (p = 0.108).

3.2. Interobserver Reproducibility

For slides paired with photographs of pruritic dogs, κ values ranged from −0.9 to 1. Agreement was classified as poor in 192/660 (29.1%), fair in 168/660 (25.5%), moderate in 145/660 (22%), substantial in 114/660 (17.3%), and good in 41/660 (6.2%) examiner pair/evaluation round/number of OIF combinations (Table 4). Overall interobserver reproducibility was fair (Fleiss’ κ = 0.379).

Friedman repeated-measures ANOVA demonstrated a significant effect of the number of examined OIFs on κ values (p < 0.001). Post-hoc Wilcoxon signed-rank tests showed significantly higher κ values after examination of 10 OIFs compared with 30 (p = 0.026) and 50 OIFs (p = 0.022), and after examination of 20 OIFs compared with 30 (p = 0.022) and 50 OIFs (p = 0.020; p values have been adjusted by the Bonferroni correction).

For slides paired with photographs of non-pruritic dogs, κ values ranged from −0.20 to 1. Agreement was classified as poor in 300/660 (45.5%), fair in 4/660 (0.6%), moderate in 73/660 (11.1%), substantial in none, and good in 283/660 (42.9%) examiner pair/evaluation round/number of OIF combinations (Table 4). Overall interobserver reproducibility was moderate (Fleiss’ κ = 0.509).

Friedman repeated-measures ANOVA again identified a significant effect of the number of examined OIFs on κ values (p < 0.001). Post-hoc Wilcoxon signed-rank tests demonstrated significantly lower κ values after examination of 10 OIFs compared with 40 (p = 0.018) and 50 OIFs (p = 0.007), and after examination of 20 OIFs compared with 30 (p = 0.027), 40 (p = 0.001), and 50 OIFs (p < 0.001; p values have been adjusted by the Bonferroni correction).

4. Discussion

For a cytological method to be clinically useful, it should provide results that are not only clinically meaningful, but also repeatable when assessed by the same observer and reproducible when assessed by different observers [14,15]. Despite the previously reported poor repeatability and reproducibility of adhesive tape strip cytology for quantitative yeast counts [11], we hypothesized that clinically acceptable agreement might still be achieved when cytology was used to diagnose Malassezia overgrowth and/or dermatitis. However, the present study demonstrated poor overall intraobserver repeatability (Table 1, Table 2 and Table 3) and interobserver reproducibility (Table 4) for both diagnostic criteria evaluated: detection of at least one yeast organism in slides paired with photographs of pruritic dogs, and detection of an average of at least one yeast organism per OIF in slides paired with photographs of non-pruritic dogs.

A small number of examiners achieved consistently high repeatability regardless of the number of OIFs examined, particularly examiner #5 for pruritic cases (Table 1) and examiners #1, #2, and #5 for non-pruritic cases (Table 2). Although factors such as experience, visual acuity, and motivation may influence observer performance, the most likely explanation for the unsatisfactory overall agreement is the heterogeneous distribution of yeasts within the adhesive tape preparations, combined with the fact that different microscopic fields are examined during each evaluation round [16]. Increasing the number of examined OIFs could theoretically improve agreement; however, examination of 50 OIFs already requires approximately 4 min per slide [11], rendering further increases impractical for routine clinical use.

Interestingly, interobserver reproducibility differed according to the diagnostic threshold used. In slides paired with photographs of pruritic dogs, agreement was higher when fewer OIFs (10 or 20) were examined, whereas in slides paired with photographs of non-pruritic dogs, agreement improved when larger numbers of OIFs (30–50) were evaluated. A similar, although non-significant, trend was observed for intraobserver repeatability (Table 3). This discrepancy likely reflects the different diagnostic cut-offs applied to the two groups. For slides associated with pruritic lesions, a positive result was defined as the detection of at least one yeast organism. Consequently, the probability of identifying a positive slide increased with the number of OIFs examined, causing some initially negative slides to become positive after evaluation of additional fields. This likely increased the frequency of false-negative results when only 10 or 20 OIFs were assessed, artificially inflating interobserver reproducibility. In contrast, for slides paired with photographs of non-pruritic dogs, positivity required an average of at least one yeast organism per OIF. In these cases, borderline yeast densities may have produced variable classifications after examination of relatively few fields, whereas evaluation of larger numbers of OIFs reduced variability around the diagnostic threshold.

These findings highlight the major influence of the selected diagnostic cut-off on both repeatability and reproducibility in the cytological diagnosis of canine Malassezia overgrowth and/or dermatitis. Differentiating between normal colonization and pathological overgrowth remains challenging, and a wide range of cut-offs has been proposed, including >1 to >5 yeasts per 400× high-power field and >0.7 to >4 yeasts per OIF [2,7,9,17,18,19]. These thresholds are largely derived from studies of healthy dogs; however, this approach has important limitations because normal yeast populations vary according to body site and breed [9,20,21]. Furthermore, detection of even small numbers of yeasts on clinically affected skin may be diagnostically relevant [22] particularly in dogs with AD that develop hypersensitivity to Malassezia spp. [1,8]. It is therefore possible that no universally applicable cytological cut-off exists for all body sites and underlying diseases, and that definitive diagnosis of Malassezia dermatitis ultimately depends on clinical response to antifungal therapy [2,23].

Previous studies have shown that approximately 50% of dogs with AD and a presumptive diagnosis of Malassezia dermatitis (diagnosed by the detection of at least one yeast organism on adhesive tape-strip cytology) respond to antifungal therapy, although, in that study, treatment was combined with antibacterial therapy in dogs with concurrent bacterial overgrowth or bacterial dermatitis [24]. Further studies are warranted to determine whether antifungal treatment can reduce AD severity in dogs in which no yeast organisms are detected after examination of 50 OIFs.

The present study has several limitations. First, the absence of a gold standard prevented assessment of the sensitivity and specificity of cytological examination. Second, evaluation was limited to 50 OIFs; although examination of additional fields might have improved agreement, this would substantially reduce the practicality of the technique in routine clinical settings. Third, the findings of this study apply specifically to adhesive tape strip cytology. While different sampling methods, such as impression smears, skin scrapings, or swab smears, may show different levels of agreement, the objective of the present study was not to compare different sampling techniques but rather to assess the reliability of adhesive tape strip cytology, which is the most commonly used method for the diagnosis of Malassezia overgrowth and/or dermatitis in clinical practice

5. Conclusions

Adhesive tape strip cytology showed poor intraobserver repeatability and interobserver reproducibility for both the detection of a single yeast organism and the assessment of whether yeast density exceeded one organism per OIF, even after examination of 50 OIFs. Consequently, negative cytology results do not exclude Malassezia overgrowth and/or dermatitis in dogs with compatible clinical signs.