Meta-analysis of the utility of culture, biopsy, and direct KOH examination for the diagnosis of onychomycosis

Onychomycosis is a highly prevalent infection worldwide with a range between 2% and 30%, corresponding to 50% of nail diseases and 30% of superficial mycoses [1, 2]. Onychomycosis is a cosmopolitan disease, and its incidence increases according to age, climate, physical activity, occupation, and underlying diseases [3]. A higher prevalence has been reported in men, individuals over 60 years of age, patients with immunosuppressive diseases, such as human immunodeficiency virus (HIV) infection or immunological defects, diabetics, and patients with peripheral vascular disease [1]. Its incidence is also higher in humid and tropical climates, under poverty and overcrowding conditions, and in athletes or sportspersons in whom a higher incidence of tinea paedis has been reported [3, 4].

This disease occurs via fungal invasion of the nail. Over the development course of the infection, there is initial colonisation with subsequent invasion of the nail bed and plate that cause changes in the nail colour, texture, and shape. There are different clinical presentations, including distal subungual, proximal subungual, white superficial, and total onychodystrophy. The distal subungual form is the most common [3, 5].

In addition to the change in nail shape, onychomycosis is related to low self-esteem because those with the condition often experience shame at being associated with poor hygiene and as a source of transmission to other individuals in their surroundings. Onychomycosis is also related to economic and work-related problems, social rejection, and a decreased quality of life [5‐8].

Causative agents are divided into the following three main groups: dermatophyte moulds, non-dermatophyte moulds, and yeasts. Dermatophyte moulds are moulds directly associated with the infection and clinical signs. The diagnostic value of the isolation of non-dermatophyte moulds or yeasts is controversial because it depends on the amount of inocula with positive growth in culture; moulds are microorganisms that can be present as skin colonisers (although they are transient colonisers in nails) and thus are often considered contaminants. In these cases, it is advised to repeat the procedure before considering them the infectious agents [6‐8].

The tests used to establish the diagnosis of onychomycosis include direct potassium hydroxide (KOH) examination, culture, histopathology, confocal laser microscopy, phase-contrast microscopy, and state of the art techniques, such as Vitek, matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry, and polymerase chain reaction (PCR) [9]. Because some of these tests are expensive and require the use of specialised equipment and materials that are not routinely used, the most commonly used methods are direct KOH examination, culture, and to a lesser extent nail biopsy, which are simpler and less costly than confocal microscopy and MALDI-TOF.

Despite the technical advantages of the traditionally used tests for the detection of the causative agents, none can be considered as a standard test alone from the viewpoint of their diagnostic utility. Therefore, several criteria are typically used for diagnostic validity studies because their simultaneous use can increase the sensitivity and specificity [10]. However, there is currently no consensus on the most appropriate combination of tests because mixed results have been reported for performance and validity. In this regard, studies have shown a high variability of results in the application of individual tests or their combinations. Sensitivity values have been reported between 23% [11] and 84.6% [12] for culture, between 44% [13] and 100% [12] for KOH, and between 81% [14] and 91.6% [15] for biopsy. The sensitivity values reported for test combinations are 57% for biopsy and KOH [11] and 98.3% for biopsy and culture [15].

In addition to the high variability in individual results, another limitation of the available studies is the inclusion of small sample sizes. Studies have performed diagnostic evaluations with samples of 40 [12], 50 [13], 63 [11], or 96 [14] individuals who are usually not selected probabilistically. Similarly, most studies present an incomplete diagnostic assessment to the extent that only data on sensitivity, specificity and predictive values are reported and relevant parameters, such as likelihood ratios, the Youden index (J), and the receiver operating characteristic (ROC) curve, among others, are ignored. These limitations can be overcome by a meta-analysis of diagnostic tests. This type of study allows the calculation of these indices via comparison among studies, which shows the parameters related to the diagnostic utility in a greater number and different types of patients and identifies potential sources of heterogeneity of the results, among other advantages.

The objective of this study was to evaluate the diagnostic validity, performance, and accuracy of culture, nail clipping with PAS staining (biopsy), and direct KOH examination in the investigation of onychomycosis.

Based on the search terms, 167,825 studies were found, of which only 2,073 included the terms in the title, abstract, or keywords. Of these, only 12 fulfilled the research protocol, and the others were excluded based on the protocol, as shown in Fig. 1. In addition to the 12 studies that fulfilled the search and selection protocol, a study conducted by the researchers was added, which corresponded to unpublished research.

Only 5 studies reported the place of study as follows: one in Iran [20], one in Germany [21], one in Taiwan [22], one in the United States [13], and one in Colombia [23]. All of the studies included in this meta-analysis defined the study population as subjects with clinical suspicion or clinical signs of onychomycosis, and the diagnostic positivity or reference criterion was considered to be any clinical suspicion of onychomycosis and at least one positive test for the condition.

The clinical criteria applied to patients in the included studies were not undergoing antifungal therapy 1 or 3 months prior to the study [13, 14, 24], not presenting dermatological diseases such as psoriasis, lichen planus, or other nail dystrophies [11, 14, 22, 24], and presenting a more serious clinical form than proximal subungual onychomycosis.

Regarding the clinical presentations, all studies reported some degree of dystrophy as an inclusion criterion but only 3 specified the number of patients broken down by type of lesion. For instance, in the study of Haghani et al. [20], the distal subungual form was more frequent (n = 88), followed by the white superficial (n = 5), proximal subungual (n = 5), proximal distal (n = 1), and total onychodystrophy (n = 1) forms. Karimzadegan-Nia et al. [14] reported more cases of total onychodystrophy (n = 77), followed by the distal subungual (n = 14), and white superficial (n = 5) forms. Finally, Lawry et al. [11] also included a greater number of patients with total onychodystrophy (n = 59), followed by the white superficial (n = 3) and distal subungual (n = 1) forms.

Table 1 describes some characteristics of the included studies, which were published between 2000 and 2016. A total of 2,858 subjects were studied. In the studies that reported the distribution of patients by gender, a greater proportion of men (57.1%) was found, the age range was wide (1–98 years), and the most common aetiological agents were dermatophytes (53.3% in all studies that reported the frequency of the fungi identified), followed by non-dermatophyte moulds (28.2%), and yeasts (18.5%).

Regarding the assessment of the methodological quality, five studies met the 14 criteria of the QUADAS tool [20, 21, 23‐25], four met 13 criteria [13‐15, 26], and the other studies met 10 [27], 9 [11], and 8 criteria [22, 28]. The least applied criteria that were most related to potential bias were those related to the correct classification of the standard and to the independence between the application and interpretation of the evaluated test and the standard without affecting the applicability of the results to the clinical scenarios in which they were traditionally used (Fig. 2).

In the meta-analysis of the direct KOH examination, heterogeneity was found in all of the evaluated parameters (p Chi-square, and Cochran-Q <0.05). In the combined measure under the random effects model, the sensitivity was 61% (95% CI = 59–64; I² = 94.7%), specificity was 95% (95% CI = 94–97; I² = 93.8%), positive likelihood ratio was 9.6 (95% CI = 1–95; I² = 98.5%), negative likelihood ratio was 0.4 (95% CI = 0.3–0.5; I² = 76.2%), diagnostic OR was 27 (95% CI = 10–74; I² = 69.6%), area under the curve was 0.87 (Fig. 3), positive predictive value was 96.3% (95% CI = 96.2–96.4), negative predictive value was 55.9% (95% CI = 55.8–56.0), and accuracy was 72.8 (95% CI = 72.7–72.9) (Table 2).

For culture, the overall sensitivity was 56% (95% CI = 54–59; I² = 90.4%), specificity was 99% (95% CI = 98–100; I² = 64.1%), positive likelihood ratio was 17.3 (95% CI = 6–49; I² = 61.8%), negative likelihood ratio was 0.4 (95% CI = 0.3–0.5; I² = 85.8%), diagnostic OR was 45 (95% CI = 18–111; I² = 40.2%), area under the curve was 0.86 (Fig. 4), positive predictive value was 99.4% (95% CI = 99.3–99.5), negative predictive value was 52.0% (95% CI = 51.9–52.1), and accuracy was 70.3 (95% CI = 70.2–70.4) (Table 2).

For biopsy, the overall sensitivity was 84% (95% CI = 82–86; I² = 85.6%), specificity was 89% (95% CI = 87–91; I² = 96.3%), positive likelihood ratio was 7.2 (95% CI = 3–19; I² = 96.4%), negative likelihood ratio was 0.2 (95% CI = 0.1–0.3; I² = 86.8%), diagnostic OR was 45 (95% CI = 13–153; I² = 90.1%), area under the curve was 0.92 (Fig. 5), positive predictive value was 93.4% (95% CI = 93.3–93.5), negative predictive value was 75.4% (95% CI = 75.3–75.5), and accuracy was 85.75 (95% CI = 85.7–85.8) (Table 2).

Table 2 shows the overall diagnostic evaluation parameters, the parameters broken down by the methodological quality of the studies, and the number of individuals analysed in the different subgroups. Generally, the quality of the articles was not related to the conclusion on the validity and performance of the three diagnostic tests, with the exception of the sensitivity, negative predictive value, and positive predictive value for biopsy. In the moderate quality studies, the following differences were found compared to the total group: higher sensitivity for direct KOH examination and biopsy and lower sensitivity for culture, lower positive predictive value for KOH and culture, and higher positive and negative predictive values for biopsy. Finally, the efficiency or accuracy of the three tests was influenced by the methodological quality of the studies, with a lower value in the moderate quality studies for KOH and culture and a higher value for biopsy.

This meta-analysis presents a comprehensive evaluation of the validity, performance, and accuracy of diagnostics tests based on the evaluation of 2,858 individuals with clinical suspicion of onychomycosis, of which 2,125 were analysed with direct KOH examination, 2,525 with culture, and 2,757 with biopsy. The evaluation was performed on individuals over a wide age range, with a similar gender distribution, different causative agents, and from different countries in Europe, Asia, and the Americas. Many of the advantages of meta-analyses were available, such as analysing a broader reference population, increasing the statistical power of the analyses, improving the external validity or extrapolation of the results, and producing scientific evidence of greater validity, reproducibility, and cost-effectiveness compared to individual studies [29].

Meta-analyses of diagnostic tests help determine their validity, performance, and accuracy, limit or evaluate bias, minimise random findings, and generate conclusions of greater scientific quality [30]. For onychomycosis, reviews are available on risk factors and comorbidities [31], treatments [32], and disease overview [33]. However, no similar studies have evaluated the three traditional diagnostic methods (culture, biopsy, and direct KOH examination), which demonstrates the importance of this study.

Nail clipping with PAS staining (biopsy) was the most sensitive test, with an overall sensitivity of 84%; the highest sensitivity was obtained by Shenoy et al. [28] with 98% and the lowest by Alkhayat et al. with 61% [25]. Overall, all studies reported a sensitivity > 60% for nail clipping with PAS staining. This finding is contrary to the result observed for culture, which had an overall sensitivity of 56%, with the lowest sensitivity of 29% reported by Jung et al. [27] and the highest of 82% reported by Weinberg et al. [26]. Nail clipping with PAS staining is useful to confirm the presence of fungi in the nail plate and its invasion by visualisation of fungal structures [23]. Although culture was found to be the least sensitive test in this study, this test should not be ignored in the diagnosis of onychomycosis because it is the only one of the three tests that can detect the aetiological agent with greater accuracy and provides highly useful information to decide on the most appropriate antifungal therapy.

In contrast to the sensitivity reported in the studies, in which culture was the least valid test, the present study found that culture was the most specific test, with an overall result of 99% and several reports of 100% [13, 14, 20‐22, 27]. These high specificity values may be biased because some studies reported 100% for this criterion as well as for the positive predictive value. These studies included the test evaluated in the standard, indicating that false positive results would never be obtained for the evaluated test.

Regarding sensitivity and specificity, it should be taken into account that these aspects are a priori intrinsic properties or probabilities of the tests that are not fully useful when applied in a clinical setting (i.e., they do not provide the certainty with which a clinician could determine whether a positive or negative result comes from an infected or healthy person) [34]. This information can be obtained using predictive values, which are the most useful clinical parameters [35]. Specifically in this study, all three tests showed excellent performance in infected individuals, with positive predictive values > 90%; however, the same finding was not observed for the negative predictive values, because only biopsy had an acceptable negative predictive value. This finding could be explained by the influence of the prevalence of the screened event on these parameters, which was very high in the case of onychomycosis [2].

Sensitivity and specificity are intrinsic test characteristics whose individual interpretations may differ when one result is high and the other is not, and predictive values are influenced by the prevalence of the screened infection or disease. Although other parameters are available for the diagnostic evaluation to overcome these limitations, they are rarely used in the evaluation of tests for onychomycosis. These parameters are the likelihood ratio, test accuracy, diagnostic OR, and Youden index (J). Likelihood ratios combine sensitivity and specificity to indicate the degree of certainty with which an infected patient can be detected in the presence of infection or a healthy individual in its absence [36]. Accuracy reflects the likelihood of valid results in all healthy and sick individuals tested. Diagnostic OR correlates test results with the presence of the disease. The Youden index (J) combines sensitivity and specificity and provides a global measure of validity and test agreement or disagreement [37].

Overall, all of the tests showed low utility when analysed separately. Nail clipping with PAS staining showed the highest likelihood ratios. With a positive likelihood ratio (PLR) of 7.20 and negative likelihood ratio (NLR) of 0.21; the best PLR was reported by Haghani et al. [20] with 845.03 and the best NLR by Jung et al. [27] with 0.07. Similar results were shown for the OR with 44.6, second only to the culture OR with 44.7, allowing the conclusion that these tests are the most useful to discriminate between healthy and infected individuals with no significant differences between them. Similarly, the area under the curve (AUC) for this test was 0.92, suggesting that it was the most valid test; similar results were obtained for the Youden index (J) and efficiency, confirming that nail clipping with PAS staining is the test with the greatest utility of the three tests evaluated.

In the meta-regression analysis, few evaluation parameters were influenced by the methodological quality of the studies. However, in this review, it was not possible to perform subgroup analyses based on other relevant variables, such as the degree or type of dystrophy, the causative agent, and the progression time of the lesion, among other microbiological, epidemiological, and demographic characteristics; this issue is one of the main limitations of this study.

The diagnostic tests independently evaluated in this meta-analysis show acceptable validity, performance, and efficiency, with nail clipping with PAS staining outperforming the other two tests. However, the results indicate the need to combine the three tests to establish a diagnosis given their complementarity in terms of being able to identify the affected patients, causative agent, and degree of invasiveness.