Diagnostic Performance of Four Serological Assays for Bovine Brucellosis and Optimised Cutoff Thresholds in an Endemic Region of Iran

ABSTRACT

Brucellosis is a severe zoonotic infection impacting dairy cattle, requiring accurate diagnostic assays for efficient control programs. This cross‐sectional study was conducted in Alborz Province, Iran, to assess the diagnostic efficacy of four serological tests for Brucella detection. One thousand serum samples were obtained from dairy cattle and analysed over 1 year of age. Furthermore, milk specimens from seropositive cows were cultured for bacteriological analysis. Serological testing detected Brucella antibodies in 33% of samples using RBT, 19.4% by SAT, 17.6% by 2‐ME and 33.5% by I‐ELISA. Bacterial culture detected Brucella spp. in 16.6% of seropositive milk samples, with all isolates classified as Brucella abortus biovar 3. Statistical methods were used to evaluate the diagnostic accuracy of each test. Bayesian latent class analysis revealed that I‐ELISA demonstrated the highest diagnostic accuracy for Brucella infection in dairy cattle, with superior sensitivity and specificity. In contrast, SAT and 2‐ME exhibited high specificity but lower sensitivity, while RBT showed moderate sensitivity with low specificity. The receiver operating characteristic (ROC) analysis was performed based on the results obtained from the Bayesian latent class analysis to further evaluate the diagnostic performance of the tests. Furthermore, the ROC curve analysis showed that SAT and 2‐ME displayed strong concordance with RBT. The ideal threshold for SAT and 2‐ME titers was established at 5.00, optimising sensitivity and specificity. Cohen's kappa analysis assessed agreement levels, revealing that RBT demonstrated the highest concordance with I‐ELISA. The results indicate that although RBT offers a simple screening approach, its sensitivity constraints require validation by I‐ELISA. A significant portion of infected animals (20%) might be undetected using RBT. These findings underscore the need for various serological assays to identify brucellosis in endemic areas accurately.

A Venn diagram representing the overlaps and unique sets between four common serological tests used for Brucellosis detection: I‐ELISA, RBT, Wright and 2‐ME. Each circle corresponds to one of the tests, with varying levels of overlap indicating the number of positives shared between these tests. Notably, the 2‐ME and Wright positive tests show high consistency with RBT and I‐ELISA. Therefore, using RBT and I‐ELISA together ensures the detection of all positives identified by 2‐ME and Wright. These two tests can thus serve as efficient initial screening methods to capture the majority of positives detected by the other tests. However, at the recommended cutoff of 1:80, it is important to note that both Wright and 2‐ME failed to detect approximately 43% of the positives identified by I‐ELISA, making them ineffective as screening methods for Brucella infections in endemic areas.

1. Introduction

Brucella bacteria are intracellular Gram‐negative, non‐spore‐forming, non‐motile and facultative bacteria that cause the contagious zoonotic disease known as brucellosis (Dadar, Shahali, Fakhri, et al. 2021). Brucella abortus is the main species influencing cattle (Jamil et al. 2017). However, other species, such as Brucella suis and Brucella melitensis, can also infect cattle in some cases (Dadar, Shahali, and Fakhri 2021; Ewalt et al. 1997). It affects humans and livestock in Iran, resulting in public health and economic concerns. The disease is endemic in most of the country, with city incidence rates ranging from 0 to 41 cases per 100,000 (Mirnejad et al. 2017). Cows are most affected, with 14.7% recorded in specific research (Dadar, Shahali, and Fakhri 2021). Brucella spp. is detected in 22% (95% confidence interval [CI]: 16%–28%) of Iranian dairy products, with raw goat milk being the most highly contaminated (Shahbazpour et al. 2025). In Iran, B. melitensis and B. abortus are the predominant species. B. melitensis is the most common and causes the majority of human and bovine brucellosis cases in the country (Dadar and Alamian 2025; Bahonar et al. 2025).

Particularly in endemic areas, such as Iran, the disease has a significant financial impact on the livestock industry. It causes lower milk production, reproductive failure, infertility in male and female‐infected animals and culling of positive animals (Alamian et al. 2021; Alamian, Bahreinipour, et al. 2023; Dadar, Shahali, and Fakhri 2021). Recovering animals may also continue to shed bacteria, thereby increasing the risk of transmission to other cattle and humans (González‐Espinoza et al. 2021). The disease poses serious public health concerns because humans can contract it through direct contact with infected animals, consuming contaminated animal products or inhaling airborne pathogens (Franc et al. 2018). It has been revealed that in endemic areas such as Algeria, 15% of veterinarians in 30 regions were affected. Most infections were caused by direct contact with contaminated animals or products, unprotected immunisation and the consumption of raw milk. Poor cleanliness, limited vaccine‐handling training and inadequate safety were the most important risk factors. These findings emphasise the importance of occupational exposure education and awareness (Lounes et al. 2022). Brucellosis affects low‐income countries. Therefore, the World Health Organization (WHO) ranks it as one of the top ‘neglected zoonotic diseases’ worldwide (Franc et al. 2018). Effective control and surveillance of bovine brucellosis depend on the early and accurate detection of brucellosis in animals, which is performed using laboratory tests due to the lack of specific clinical signs (Cárdenas et al. 2019). Furthermore, it is important to consider the circumstances of brucellosis in neighbouring countries when evaluating regional control measures. A study in Pakistan, east of Iran, highlighted the ongoing zoonotic risk of brucellosis in cattle. It reported that 7.54% of samples were seropositive, while 1.88% were confirmed positive by RT‐PCR (Ullah et al. 2022).

Identifying brucellosis in control and eradication programs typically relies on bacteriological and serological assessments (Gall and Nielsen 2004). The isolation and identification of Brucella spp. are the gold standards for the confirmatory and precise diagnosis of brucellosis (World Organisation for Animal Health 2019; Nielsen 2002). The chances of obtaining a positive culture from a live‐infected animal are minimal if samples are not collected from an abortion (McGiven 2013). Moreover, culture is impractical for large‐scale application due to the associated expenses and biosafety issues, as B. abortus is classified as a Biosafety Level 3 pathogen (Poester et al. 2010). This method poses a significant risk of infection to laboratory personnel (Pereira et al. 2020; Dadar, Tabibi, et al. 2023). Alternative diagnostic techniques for brucellosis, including polymerase chain reaction (PCR) tests utilising serum/blood, swabs and milk; however, the accuracy of these tests remains inadequately established, and the available data are inconsistent (McGiven 2013; Nielsen 2002; Ducrotoy et al. 2018). Therefore, due to the restrictions of direct tests, serological tests are used as the foundation for diagnosing bovine brucellosis in the control program, as the clinical samples, such as blood and milk, are easily accessible and the methods are comparatively cost‐effective and simple (Nielsen and Yu 2010; Dadar, Wareth, et al. 2021). Although various serological tests are available for diagnosing brucellosis, none achieve absolute accuracy. There is no standalone test concept because of the differences in diagnostic specificity (DSp) and diagnostic sensitivity (DSe), which may be influenced by both the test itself and the epidemiology of the disease (Greiner and Gardner 2000; McKenna and Dohoo 2006). The serological diagnosis of bovine brucellosis is typically conducted using a sequential approach that involves both screening and confirmatory tests (Ducrotoy et al. 2018). Screening tests, including the Rose Bengal test (RBT), buffered plate agglutination test (BPAT) and indirect enzyme‐linked immunosorbent assay (I‐ELISA), are known for their low cost, high sensitivity and quick execution. However, positive results from these tests require confirmation through additional testing (Gall and Nielsen 2004). Conversely, confirmatory tests have high sensitivity and specificity. However, the complement fixation test (CFT) and fluorescence polarisation assay (FPA) typically require skilled interpretation of results and specialised equipment (Poester et al. 2010).

Diagnostic techniques in Iran, ranging in sensitivity and specificity, are bacterial isolation, PCR methods and serological assays like the RBT, serum agglutination test (SAT) and I‐ELISA or direct ELISA (Dadar, Shahali, and Fakhri 2021; Ghorbani et al. 2013). However, the control program of veterinary organisations utilised serological tests such as RBT, SAT, 2‐mercaptoethanol test (2‐ME), and I‐ELISA to screen for industrial and semi‐industrial bovine brucellosis. However, the cost of the ELISA kit often prevents its use (Izadi et al. 2024). This study thoroughly evaluated the accuracy (DSe and DSp) of the most commonly used serological tests for diagnosing bovine brucellosis in Iranian dairy cattle farms to identify the most reliable assays and accurate cutoffs for the SAT and 2‐ME diagnostic tests, which are the most frequently utilised methods in endemic regions like Iran.

2. Materials and Methods

2.1. Study Area

This cross‐sectional study was conducted from 2023 to 2024 on 1000 serum samples collected from industrial dairy farms to evaluate and compare the four Brucella diagnostic tests using various serological methods. The samples were supplied by Alborz Province, located in the north‐central part of Iran, in various regions, which is crucial for dairy production and has a dynamic agricultural economy (Bahonar et al. 2025). The province borders Tehran, Mazandaran, Qazvin and Markazi, connecting Iran's capital to the Caspian region (Figure 1). Approximately 10 mL of blood was drawn from the jugular of cows older than 1 year. The blood samples were preserved without anticoagulant in a refrigerated container. They were quickly sent to the Veterinary Diagnosis and Treatment Department (Karaj, Iran) for serum separation via centrifuge for 5 min at 3000 rpm. The sampled cows appeared to be vaccinated, healthy cattle with non‐mastitic milk and normal lactate levels. The estimated volume of each serum sample after centrifugation was 2 cc.

FIGURE 1.

Map showing the geographical location of Alborz Province in Iran, highlighting the study area.

2.2. Serological Analysis

Serum samples (n = 1000) were tested by the RBT, SAT, 2‐ME and I‐ELISA. The production of Brucella antigens for RBT, SAT and 2‐ME was conducted at the Razi Vaccine and Serum Research Institute in Karaj, Iran, using B. abortus S99 antigen. The WOAH standards indicate that a titer of 1:80 or higher for SAT and 2‐ME is considered positive for specific agglutination antibodies against Brucella (Alton et al. 1988). In addition, an I‐ELISA was conducted to validate the results of other serology tests in cattle. To achieve this objective, an I‐ELISA was conducted using the Ingezim Brucella Bovina 2.0 ELISA Test Kit (Ingenasa, Madrid, Spain), following the manufacturer's guidelines (https://ingenasa.eurofins‐technologies.com). This kit is designed to assess specific antibodies against B. abortus. The optical density (OD) of the analysed samples was measured at 450 nm and positive samples were identified using the subsequent formula:

$$\mathrm{Cutoff}=\mathrm{O}{\mathrm{D}}_{450}\mathrm{nm}\mathrm{positive}\mathrm{control}\times 0.4=40\%\mathrm{positivity}$$

The OD of samples was considered positive for OD value ≥ 1.0 and negative for OD value ≤ 0.2.

2.3. Bacteriological Examination of Milk Samples and Bacterial Biotyping

Milk samples from all seropositive cows were spread onto Brucella selective media (Brucella agar; HiMedia, India) supplemented with vancomycin (10.0 mg), nalidixic acid (2.5 mg), cycloheximide (50.0 mg), nystatin (50,000 IU), polymyxin B (2500 IU), bacitracin (12,500 IU) (Oxoid, UK), and inactivated 5% horse serum and were subsequently incubated at 37°C with 10% CO₂ for 14 days. A validated panel of Brucella biotyping assays was used to identify the bacteria on the isolated colonies. The tests comprised CO₂ dependence, lysis by specific phages, agglutination with acriflavine, Brucella monospecific antisera, H₂S generation, and growth in thionin and fuchsin‐coloured conditions (Alton et al. 1988).

2.4. Statistical Analysis

2.4.1. Latent Class Model

A Bayesian latent class model (LCM) was employed to estimate the Se and Sp of four diagnostic tests, including RBT, SAT, 2‐ME, and I‐ELISA, for detecting Brucella infection in dairy cattle farms. The analysis was performed using the runjags package in R, which interfaces with Just Another Gibbs Sampler (JAGS) for Markov chain Monte Carlo (MCMC)‐based model fitting (Dendukuri and Joseph 2001). Each model was run with three parallel chains, each consisting of 50,000 iterations, following a burn‐in period of 10,000 iterations to allow the chains to stabilize. Convergence of the MCMC simulations was assessed using the Gelman–Rubin diagnostic (R‐hat), with values less than 1.05 indicating satisfactory convergence. This statistical approach allowed for the reliable estimation of diagnostic test parameters within a flexible Bayesian framework (Toft et al. 2007). To compare the posterior accuracy estimates of the tested tests and those in the literature, the step function in OpenBUGS was used to estimate the probability and Monte Carlo error (MCE) of accuracy parameters exceeding a chosen value.

2.4.2. ROC Curve

The diagnostic performance of SAT and 2‐ME with RBT and I‐ELISA was assessed using receiver operating characteristic (ROC) curve analysis. Although I‐ELISA is not a definitive gold standard, it was used as a reference comparator due to its widespread use and high reported diagnostic performance in Iran (99.9% specificity and up to 99% sensitivity, according to the manufacturer). The area under the curve (AUC) was calculated to evaluate the overall discriminatory power of each test. Sensitivity and specificity were determined at various cut‐off points, and the optimal cutoff was identified based on the highest combined sensitivity and specificity. The ROC curve was plotted for each test (SAT titer and 2‐ME titer). The AUC and its 95% CI were computed to assess diagnostic performance. The asymptotic significance (p value) was calculated to test the null hypothesis that the true AUC is 0.5 (indicating no discriminatory power). The optimal cut‐off value for each test was identified as the point that maximised both sensitivity and specificity. It is important to note that the sensitivity and specificity values reported are relative to I‐ELISA and should be interpreted as comparative rather than absolute diagnostic measures. All statistical analyses were performed using IBM SPSS Statistics (version 25) and Microsoft Excel.

2.4.3. Cohen's Kappa Analysis

The agreement between the reference I‐ELISA test and the other tests (RBT, SAT and 2‐ME) was assessed using Cohen's kappa (κ) statistic. The kappa statistic measures the level of agreement between two diagnostic tests beyond what would be expected by chance alone. Kappa values range from −1 to +1, where values ≤ 0 indicate no agreement, 0.01–0.20 indicate slight agreement, 0.21–0.40 indicate fair agreement, 0.41–0.60 indicate moderate agreement, 0.61–0.80 indicate substantial agreement and 0.81–1.00 indicate almost perfect agreement. The strength of agreement between I‐ELISA and each reference test was interpreted based on the kappa value and its corresponding 95% CI. All statistical analyses were performed using SPSS Ver. 25, and a p < 0.05 was considered statistically significant.

3. Results

3.1. Serological Test Results and Milk Culture Outcomes

All cattle sera were examined for Brucella antibodies by RBT, and 330 (33%) sera specimens were serologically positive. All sera tested under the SAT, 2‐ME and I‐ELISA methods revealed 19.4%, 17.6% and 33.5% seropositive cases, respectively (Table 1). The culture results of 330 milk samples of all seropositive cows showed that Brucella spp. were isolated and identified from 16.6% (55/330) of milk specimens collected from seropositive cows. Brucella was visible under the microscope as single or small pairs of Gram‐negative coccobacilli, originating from small pinpoint colonies with translucent and shiny honey‐coloured surfaces. Conventional biotyping tests on isolated bacteria identified B. abortus biovar 3 in all cows (Table 1).

TABLE 1.

Seroprevalence of cattle through RBT, SAT, 2‐ME, I‐ELISA and culture.

Test	Totals samples	Positive sample (%)
RBT	1000	330 (33)
SAT	1000	193 (19.3)
2‐ME	1000	176 (17.6)
I‐ELISA	1000	334 (33.4)
Culture	330 a	55 (16.6)

^a Note: Culture was performed only on a subset of 330 seropositive samples.

3.2. LCM Results

Bayesian latent class analysis (LCA) estimated the diagnostic accuracy of four tests for Brucella infection in dairy cattle. Under the main model with constrained priors, the Se of RBT, SAT, 2‐ME and I‐ELISA was estimated at 80.5% (95% CrI: 76.1–84.4), 61.6% (56.2–66.8), 58.7% (53.3–64.0) and 93.1% (88.9–96.0), respectively. The corresponding specificities (Sp) were 59.0% (55.8–62.1) for RBT and 100% (99.4–100) for SAT, 2‐ME and I‐ELISA. In the unconstrained model using vague priors (β(1,1)), posterior sensitivity estimates remained similar, with I‐ELISA showing the highest Se at 98.9% (96.3–99.9), followed by RBT at 82.3% (77.9–86.1), SAT at 62.1% (56.7–67.3) and 2‐ME at 59.3% (53.9–64.5). Specificity estimates were consistent with the main model, confirming 100% specificity for SAT, 2‐ME and I‐ELISA, and 59.0% for RBT (Table 2). These findings highlight the superior sensitivity and specificity of I‐ELISA, while SAT and 2‐ME showed high specificity but lower sensitivity. RBT demonstrated moderate sensitivity and low specificity.

TABLE 2.

Bayesian latent class analysis (LCA) of brucellosis diagnostic tests.

Parameter	Main model (constrained priors)		Unconstrained model
	Prior density	Posterior estimate (95% CrI)	Prior density	Posterior estimate (95% CrI)
Se_RBT	β(36.1,11.3)	80.5 (76.1–84.4)	β(1,1)	82.3 (77.9–86.1)
Se_SAT	β(47.7,16.6)	61.6 (56.2–66.8)	β(1,1)	62.1 (56.7–67.3)
Se_2‐ME	β(30.0,30.0)	58.7 (53.3–64.0)	β(1,1)	59.3 (53.9–64.5)
Se_I‐ELISA	β(47.7,16.6)	93.1 (88.9–96.0)	β(1,1)	98.9 (96.3–99.9)
Sp_RBT	β(108.7,6.0)	59.0 (55.8–62.1)	β(108.7,6.0)	59.0 (55.8–62.1)
Sp_SAT	β(108.7,6.0)	100 (99.4–100)	β(108.7,6.0)	100 (99.4–100)
Sp_2‐ME	β(108.7,6.0)	100 (99.4–100)	β(108.7,6.0)	100 (99.4–100)
Sp_I‐ELISA	β(108.7,6.0)	100 (99.4–100)	β(108.7,6.0)	100 (9.4–100)

Note: All sensitivity (Se) and specificity (Sp) estimates are percentages. Vague priors for sensitivity used β(1,1) uniform distributions.

Abbreviations: β(a,b), beta distribution priors; CrI, credible interval (Bayesian equivalent of confidence intervals).

Based on Bayesian posterior estimates from a total of 1000 cattle, animals were classified into four diagnostic categories to guide brucellosis management decisions (Table 2). Class 1 (definite active infection) included 214 animals (21.4%; 95% CrI: 19.8%–23.1%) who tested positive on I‐ELISA and at least two of RBT, SAT or 2‐ME. This group had a positive predictive value (PPV) of 99.2% and a negative predictive value (NPV) of 98.7%, warranting immediate culling. Class 2 (probable early infection) comprised 277 animals (27.7%; 95% CrI: 25.8%–29.7%) with positive I‐ELISA and RBT results only, but negative or inconclusive results on SAT and 2‐ME. This group had a PPV of 89.4% and an NPV of 95.1%, leading to a recommended strategy of retesting and culling if positive on follow‐up. Class 3 (isolated I‐ELISA positive) involved 67 animals (6.7%; 95% CrI: 5.8%–7.6%) that were I‐ELISA‐positive but negative on all other tests. With a PPV of 75.3%, this group raises concerns about possible cross‐reactivity or laboratory error, and thus warrants further confirmation via culture or PCR. Finally, Class 4 (uninfected) represented the majority, with 588 animals (58.8%; 95% CrI: 56.7%–60.9%) testing negative on all assays. This group showed 100% NPV and 99.9% PPV, supporting no action. Notably, only one false negative was observed in this group (RBT+/SAT+/2‐ME+ but I‐ELISA−), indicating high diagnostic reliability overall (Table 3).

TABLE 3.

Classification and management of brucellosis in cattle based on serological testing (n = 1000 cases; Bayesian posterior estimates).

Class	Definition	Diagnostic pattern	Cases	Prev. (95% CrI)	PPV	NPV	Action
1	Definite active infection	ELISA+ and ≥ 2 of (RBT+/Wright+/2‐ME+)	214	21.4% (19.8–23.1)	99.2%	98.7%	Cull
2	Probable early infection	ELISA+ and RBT+ only	277	27.7% (25.8–29.7)	89.4%	95.1%	Retest and cull if positive
3	Needs re‐testing	ELISA+ alone	67	6.7% (5.8–7.6)	75.3%*	99.8%	Rule out lab error or cross‐reactivity
4	Uninfected	All tests negative	587	58.8% (56.7–60.9)	99.9%	100%	No action

^*Reflects cross‐reactivity risks.

3.3. ROC Curve Results

The diagnostic performance of the recommended SAT and 2‐ME titers for identifying positive samples was evaluated against I‐ELISA, which demonstrated the highest sensitivity in the LCA. ROC curve analysis was used to assess their accuracy (Figure 2). The case processing summary indicated 1000 valid cases, with 334 positive and 666 negative results. One case was excluded due to missing data. The AUC for the SAT was 0.859 (95% CI: 0.830–0.888, p < 0.001), demonstrating excellent discriminatory power. Similarly, the 2‐ME test showed an AUC of 0.819 (95% CI: 0.786–0.851, p < 0.001), indicating good diagnostic performance. Both AUC values were statistically significant (p < 0.001), confirming that the tests performed better than chance (AUC = 0.5). Using the I‐ELISA as the reference comparator, the SAT titer achieved a sensitivity of 75.4% and a specificity of 91.4% (1 − specificity = 0.086) at the optimal cut‐off point of 5.00. For the 2‐ME titer, the sensitivity and specificity were 66.2% and 94.7% (1 − specificity = 0.053), respectively. A detailed summary of sensitivity and specificity estimates across various cut‐off points, using I‐ELISA as the reference comparator, is presented in Table 4.

FIGURE 2.

ROC analysis of SAT and 2‐ME titers for Brucella diagnosis, using I‐ELISA as the comparative reference test. The blue line represents the SAT (Wright) titer and the red line represents the 2‐ME titer. The green diagonal line indicates the reference line (AUC = 0.5), representing no discriminatory ability. The curves demonstrate that SAT had slightly higher diagnostic accuracy than 2‐ME, with optimal cut‐off points indicating corresponding sensitivity and specificity values.

TABLE 4.

Diagnostic performance of SAT titer and 2‐ME titer against I‐ELISA as the reference comparator with a recommended cutoff (1:10).

Test	AUC	95% CI	p	Sensitivity (Se)	Specificity (Sp)	Optimal cutoff
SAT titer	0.859	0.830–0.888	< 0.001	75.4%	91.4%	5.00
2‐ME titer	0.819	0.786–0.851	< 0.001	66.2%	94.7%	5.00

ROC curve analysis also evaluated the diagnostic performance of the SAT and 2‐ME in comparison to RBT results (Figure 3). The results showed that the AUC for SAT titer was 0.958 (95% CI: 0.940–0.975, p < 0.001), indicating excellent diagnostic accuracy. Furthermore, the AUC for 2‐ME titer was 0.879 (95% CI: 0.850–0.907, p < 0.001), indicating good diagnostic accuracy. Both AUC values were statistically significant (p < 0.001), confirming that both tests performed significantly better than chance (AUC = 0.5). For the SAT, at the optimal cut‐off point of 5.00, the sensitivity was 75.4% and the specificity was 91.4% (1 − specificity = 0.007) (Table 4). For the 2‐ME test, at the optimal cut‐off point of 5.00, the sensitivity was 66.2% and the specificity was 94.7% (1 − specificity = 0.003) (Table 5).

FIGURE 3.

The ROC curve for assessing the diagnostic performance of the SAT and 2‐ME tests in comparison to RBT. The blue curve represents the SAT (Wright) titer, while the red curve represents the 2‐ME titer. The green diagonal line indicates a test with no discriminatory ability (AUC = 0.5). Both curves demonstrate the ability of the tests to distinguish between infected and non‐infected animals, with SAT showing slightly higher diagnostic accuracy based on its curve positioning.

TABLE 5.

Diagnostic performance of SAT and 2‐ME titer against RBT as the screening test.

Test	AUC	95% CI	p	Sensitivity (Se)	Specificity (Sp)	Optimal cutoff
SAT titer	0.958	0.940–0.975	< 0.001	92.1%	99.3%	5.00
2‐ME titer	0.879	0.850–0.907	< 0.001	76.1%	99.3%	5.00

The chart (Figure 4) compares the performance of the 2‐ME and SAT tests across different dilution ratios. The dilution ratio 80 has the highest percentage of ‘positive’ I‐ELISA results for both SAT and 2‐ME methods (more than 99%).

FIGURE 4.

A combined bar chart compares the 2‐ME and SAT dilution ratios with an I‐ELISA positive percentage for each dilution. This bar chart illustrates the percentage of positive findings from 2‐ME and SAT (Wright) serological tests at four dilution ratios (1:10, 1:20, 1:40, and 1:80). The blue bars represent 2‐ME results, while the red bars represent SAT outcomes. Both procedures show a growing positive percentage with greater dilution ratios, with 2‐ME having a higher positivity rate at most dilutions.

The Venn diagram analysis comparing the RBT and I‐ELISA test outcomes for positive samples supports the observation that RBT did not detect 20% of positive samples identified by I‐ELISA. This finding indicates a significant difference in the sensitivity of the diagnostic methods (Figure 5).

FIGURE 5.

The intersection of the two circles in the Venn diagram represents the positive cases detected by both I‐ELISA and RBT tests. This set of cases, confirmed by both methods (n = 266), suggests that these tests are reliable for identifying 80% of brucellosis cases. However, the non‐overlapping areas (the regions where only I‐ELISA or RBT are positive) suggest that 20% of positive samples (n = 68) cannot be detected using RBT alone.

As shown in Figure 6, the shaded region of the Venn diagram represents the 174 common positives shared by all tests. The intersections highlight the overlap between the test results, illustrating where agreement occurs in identifying cases. Notably, the 2‐ME and SAT positive tests were consistent with the RBT and I‐ELISA results. Therefore, using RBT and I‐ELISA together ensures the detection of nearly all positives identified by 2‐ME and SAT. These two tests can thus serve as efficient initial screening methods to capture the majority of positives detected by the other tests. However, at the recommended cutoff of 1:80, it is important to note that both SAT and 2‐ME failed to detect approximately 43% of the positives identified by I‐ELISA, making them ineffective as screening methods for Brucella infections in endemic areas. To improve detection, suspicions of brucellosis should be considered at a much lower titer of 1:10, with the serological test repeated after several days. Even when including cases with titers ranging from 1:10 to 1:80, around 25% of infected samples remain undetected when compared to I‐ELISA.

FIGURE 6.

A Venn diagram representing the overlaps and unique sets between four tests: I‐ELISA, RBT, SAT and 2‐ME. Each circle corresponds to one of the tests, with varying levels of overlap indicating the number of positives shared between these tests.

3.4. Cohen's Kappa Analysis

The agreement among several serological tests and I‐ELISA, as determined by the kappa statistic, indicates concordance beyond chance. All tests demonstrated statistically significant agreement with the I‐ELISA (p < 0.001), supporting the diagnostic importance of these results. The RBT revealed the highest agreement, with a kappa value of 0.702 (95% CI: 0.654–0.752), indicating substantial agreement. Among the SAT and 2‐ME tests, the kappa values were 0.636 and 0.592, respectively, indicating moderate agreement. These results suggest that the SAT and 2‐ME tests are less reliable than the I‐ELISA. The RBT's highest kappa value might be the most reliable test for detecting brucellosis cases (Table 6).

TABLE 6.

Agreement between different serological tests used for brucellosis (n = 1000).

Comparison	Standard error	Kappa value	95% CI of kappa	p *
SAT and I‐ELISA	0.026	0.636	0.583–0.689	< 0.001
2‐ME and I‐ELISA	0.027	0.592	0.538–0.645	< 0.001
RBT and I‐ELISA	0.025	0.702	0.654–0.752	< 0.001

^* p < 0.05 is considered significant.

4. Discussion

The control of brucellosis in animals and humans depends on the reliability of the procedures employed for detecting and identifying the causal agent (Andrade et al. 2024). Diagnosing this bacterial disease accurately is very challenging, and determining the most effective way to control or manage it is also difficult (Helmy et al. 2025).

Studies indicate that test performance may differ between cattle and animals, such as sheep and water buffalo. This underscores the necessity of species‐specific validation of diagnostic assays (Fosgate et al. 2002). Serological assays are crucial for detecting bovine brucellosis, particularly in endemic areas where accurate and rapid diagnosis is essential (Ibarra et al. 2023). Screening tests, such as the RBT and BPAT, are extensively utilised due to their higher sensitivity. However, they can result in false positives (Ibarra et al. 2023). RBT detects antibodies against Brucella spp., specifically targeting the lipopolysaccharide (LPS) on the surface of the bacteria. However, other bacteria and microorganisms may have structurally similar antigens to those of Brucella. Cross‐reacting bacteria mainly include Yersinia enterocolitica, Escherichia coli and Francisella tularensis. These organisms may possess LPS molecules or other surface proteins that resemble the antigens found in Brucella species, triggering an immune response that is mistakenly interpreted as brucellosis. The Venn diagram in Figure 5 shows that RBT detected 64 positive samples that I‐ELISA has not confirmed. This may be due to antibodies from organisms other than Brucella, leading to false positives, justifying the use of confirmatory tests like the SAT or ELISA for improved specificity (Chachra et al. 2009; Salman and El Nasri 2012). This is particularly important in brucellosis diagnostics, as RBT is commonly used as the first screening tool in endemic areas. The SAT provides quantitative data but exhibits lower sensitivity than the RBT and the BPAT. The RBT exhibits greater sensitivity than I‐ELISA, enabling the detection of more positive samples; nevertheless, its poorer specificity may result in false positives due to cross‐reactivity. In a study on cetaceans, RBT identified positives that I‐ELISA failed to confirm; many false positives were associated with cross‐reactivity to other Gram‐negative bacteria, including Photobacterium damselae and Vibrio parahaemolyticus (Di Febo et al. 2025). The cross‐reactivity arises because RBT employs whole‐cell antigens that may interact with antibodies directed against similar bacteria, while I‐ELISA is often more selective by focusing on specified antigens. Furthermore, RBT's antigen preparation (heat‐killed whole cells, acidified and dyed) may modify antigenic characteristics, leading to false positives or inconsistencies with I‐ELISA outcomes (Di Febo et al. 2025). Research in cattle and other species indicates that while the RBT exhibits excellent sensitivity in identifying genuine positives, its specificity is lower than that of the I‐ELISA. This implies that RBT may identify antibodies resulting from infections or exposures unrelated to Brucella, as well as instances of non‐specific agglutination (Barkay et al. 2024; Ipola et al. 2018; Zakaria 2018).

To confirm, it has been advised to conduct more specialised tests such as the CFT and the I‐ELISA (Fosgate et al. 2002). I‐ELISA offers high sensitivity and the ability to differentiate between antibody classes, while CFT is considered the gold standard due to its high specificity (Aggad and Boukraa 2006).

The Venn diagram in Figure 5 suggests that I‐ELISA may be more sensitive for detecting a wider range of brucellosis cases than RBT. A significant portion of infected animals (20%) might go undetected using RBT, which is widely used as the first screening test in endemic areas. These results suggest that I‐ELISA might have a higher sensitivity than RBT, SAT and 2‐ME. I‐ELISA can identify cases in the early stages of infection or those with antibody levels too low to be detected by other methods. This highlights the need to combine the strengths of each test to achieve a more comprehensive detection of brucellosis.

However, serological tests selected for detecting bovine brucellosis in endemic areas are based on several criteria, including vaccination status, rapid results and the balance between specificity and sensitivity (Andrade et al. 2024; Ducrotoy et al. 2018; Pereira et al. 2020). Whereas the RB51 vaccination does not interfere with most serological techniques, vaccination with the S19 strain can produce false positives in some tests for several months (Ibarra et al. 2023). In endemic areas, a combination of tests enhances diagnostic accuracy and reduces false negative results (Alamian, Amiry, et al. 2023).

Detecting and controlling bovine brucellosis in Iran critically depends on the use of serological tests. The RBT, SAT and the 2‐ME test are the three serological tests most commonly used by the Iranian Veterinary Organization (IVO) for brucellosis screening in cattle (Alamian, Bahreinipour, et al. 2023; Izadi et al. 2024). Recent research suggests that this method may have limitations and that a combination testing strategy would be advantageous (Alamian, Amiry, et al. 2023). SAT and 2‐ME are widely used in brucellosis diagnosis due to their low cost, ease of use, and minimal equipment needs, making them suitable for low‐resource settings. These tests remain preferred in many regions because of their historical use, established protocols, and ability to function without advanced infrastructure. The SAT is practical and accessible for basic labs, while 2‐ME adds value by distinguishing active from past infections (Corbel 2006; Díaz et al. 2011). In contrast, I‐ELISA, PCR, and culture methods offer greater accuracy but are limited by high costs and technical demands. Thus, prioritising SAT and 2‐ME ensures broader diagnostic reach and aligns with local capacities in endemic areas.

Our analysis of the SAT and 2‐ME titers for positive samples revealed a higher likelihood of missing positive cases in screening by these tests, as they showed less sensitivity than the recommended new cutoff. Their lesser sensitivity makes them less reliable in identifying all true positive instances, even if they have excellent specificity. SAT is widely utilised in India and Pakistan as a confirmatory test, providing quantitative titers that help differentiate between active infection and past exposure (Sharma 2016; Hassan et al. 2022).

Using the I‐ELISA as the reference comparator, our analysis showed an 85.5% and 84% accuracy for SAT and 2‐ME, respectively, indicating moderate overall performance. The PPV for these tests highlighted that when the test was positive, it was highly likely to be correct. However, their NPV was lower, indicating a higher risk of false negatives.

The LCA results in Table 2 provide detailed insight into the diagnostic accuracy of four serological tests for brucellosis in cattle, with two model approaches, and showed that I‐ELISA shows the highest sensitivity, with posterior estimates of 93.1% (88.9–96.0%) in the constrained model and 98.9% (96.3–99.9%) in the unconstrained model, indicating a strong ability to identify infected animals correctly. Our result also showed that RBT sensitivity is moderately high at about 80%–82%, suggesting it misses some positive cases, but remains effective as a screening tool. Furthermore, according to this analysis, SAT and 2‐ME sensitivities are lower, ∼59%–62%, indicating these tests miss a substantial proportion of true positive animals when used alone. These findings correspond with other global studies where ELISA typically surpasses agglutination‐based tests in sensitivity, likely due to better detection of chronic or low‐antibody titer infections (Rahman et al. 2019, 2013; Bodenham et al. 2021). The lower RBT specificity at 59.0% (55.8%–62.1%) with frequent false positives contrasts with some previous studies that reported higher specificity but highlights RBT's known limitations, especially cross‐reactivity and false positives due to vaccination or related bacteria (Bodenham et al. 2021; Rahman et al. 2013). The very high specificity of SAT, 2‐ME, and I‐ELISA aligns with their role as confirmatory tests, as reflected in international brucellosis diagnostic guidelines (Rahman et al. 2019).

Through the ROC analysis, we identified a new cutoff of 1:10 for positive samples in this study for the SAT and 2‐ME test, which showed higher sensitivity and better detection of positive samples. The new cutoff for SAT and 2‐ME in this study suggests that tests detect positive samples more reliably and with high specificity. Other studies reported a sensitivity of 32.5% (95% CI: 22.8–42.3) and a specificity of 96.4% (95% CI: 89.5–100) for the SAT test, which differs from our study (Mohsenpour et al. 2011). Furthermore, another Iranian study reported sensitivity, specificity and accuracy rates of 65.31%, 98.85% and 94.12% for bovine brucellosis, respectively (Alamian, Amiry, et al. 2023).

The starting cutoff in 2‐ME and Wright, from which positive samples were detected using I‐ELISA, was 1:10. This indicates that considering gradual cutoffs ranging from 1:10 to 1:80 may enhance the detection of positive bovine brucellosis cases in endemic areas using these methods. However, since I‐ELISA and RBT together identify nearly all of the positives from the 2‐ME and SAT, they provide broader coverage for detecting the target condition or disease without missing significant cases. Therefore, a confirmatory test with high specificity, such as I‐ELISA or culture for cattle, is essential to reduce the number of livestock improperly classified as infected (Fosgate et al. 2002). As depicted in Figure 5, many positive samples may be missed when using these two tests separately or even in combination.

However, I‐ELISA is not the gold standard for diagnosing brucellosis, as bacterial culture and PCR remain the definitive methods. However, in endemic regions such as Iran, practical constraints, including limited availability of bacterial culture facilities and molecular diagnostic tools, pose significant challenges. Given these limitations and based on the results of Bayesian LCA, we selected I‐ELISA as the most reliable and feasible reference method available for our study. I‐ELISA offers high sensitivity and specificity, is cost‐effective and can be readily implemented in resource‐limited settings.

According to our analysis, positive results on I‐ELISA and RBT but negative on other confirmatory tests have a slightly lower PPV (89.4%), suggesting some may not truly be infected or are at an early infection stage. International studies recommend prompt retesting, as acute or early infections may yield such partial serological profiles (Bodenham et al. 2021).

The classification used in this study closely mirrors the tiered approach advocated in South Asia (Rahman et al. 2019) and Latin America (Samartino 2002), where a highly sensitive screening test (e.g., RBT or ELISA) is followed by more specific confirmatory tests (SAT, 2‐ME or I‐ELISA).

Another Iranian study, a comparative analysis of serological testing for bovine brucellosis, also revealed that I‐ELISA exhibited superior sensitivity and specificity relative to alternative procedures, including the SAT. I‐ELISA may be more reliable than our assessed SAT and 2‐ME tests. However, due to its costs, it could not be used in the control program for brucellosis on Iranian farms (Saadat et al. 2017).

Screening tests exhibiting high sensitivity yield a reduced incidence of false‐negative classifications, thereby enhancing their efficacy in identifying infected animals for the test‐and‐slaughter program in endemic areas, such as Iran. However, diagnostic test characteristics (sensitivity and specificity) variations among species and locations are crucial when developing disease management programs (Ducrotoy et al. 2018; Fosgate et al. 2002). Research findings from one species and country may not be directly applicable to another without validation in the relevant livestock populations (Fosgate et al. 2002).

Selection bias may have affected our study due to the testing of only a subset of animals from the larger herds. Population list frames were unavailable, and the handling facilities were unable to inspect all animals effectively. This may have affected the population prevalence estimates, as the examined animals may not have accurately represented the entire herd.

We did the culture test to confirm seropositive animals. The results of our study also showed the isolation of field strains of B. abortus biovar 3 from seropositive vaccinated dairy cows according to classical biotyping. These cows had been vaccinated as adults with the reduced RB51 vaccine. The occurrence of B. abortus infections in the milk of seropositive cows was also reported in different studies in dairy cattle farms with a history of abortion (Abnaroodheleh et al. 2023). Several studies also reported the presence of B. abortus biovar 3 in cows in endemic areas (Arellano‐Reynoso et al. 2013; Zoghi et al. 1992; Islam et al. 2018). Another study reported that the B. abortus strains from Iran showed close genetic similarity to strains from neighbouring countries such as Iraq and Egypt, suggesting transboundary livestock movement and regional circulation (Dadar, Brangsch, et al. 2023). Regarding neighbouring countries and the broader region around Iran, brucellosis persists as endemic in the Middle East, in countries such as Saudi Arabia, Syria, Palestine, Jordan and Oman, where elevated incidence rates have been documented (Refai 2002). In the Mediterranean basin, encompassing several neighbouring countries. B. melitensis biovar 3 and B. abortus biovar 1 are predominantly isolated from humans and livestock. While B. suis biovar 2 is generally more prevalent in European Mediterranean countries and less common in Afro‐Asian regions due to cultural influences (Abdeen et al. 2019), recent reports of its isolation from Egypt suggest it may also be present in other countries within the region (Wareth et al. 2023). B. melitensis can cross species barriers and has established reservoirs in cattle and bovines in nations like Albania, Egypt, France, Israel, Iran, Italy and Turkey (Abdeen et al. 2019; Dadar, Brangsch, et al. 2023; Dadar, Shahali, Fakhri, et al. 2021).

However, some diseases, like brucellosis, rely on bacterial isolation for diagnosis, which may not be reliable for all animals. Most bacterial approaches fail to retrieve live organisms from diseased animals. Animals with positive bacterial cultures may be at a different stage of infection than those with negative bacterial cultures. False‐positive culture findings are unlikely, while false‐negative results may affect sensitivity and specificity estimations (Fosgate et al. 2002). In this regard, PCR has demonstrated superior sensitivity compared to culture in diagnosing brucellosis across multiple studies, particularly in detecting cases missed by conventional methods or in chronic or latent cases where the bacterial load may be low. A study showed that PCR effectively detected Brucella in blood samples from seronegative animals, underscoring its utility in diagnosing subclinical or early‐stage infections. Their findings underscore that PCR can work as a complementary tool to address the shortcomings of serology and culture, especially in endemic areas where serological cross‐reactions or sporadic bacterial shedding may complicate diagnosis (El‐Diasty et al. 2021).

In comparison to bacterial culture and PCR, I‐ELISA has significant limitations. Bacterial culture is the primary method for identifying brucellosis, although it requires time, utilises Biosafety Level 3 facilities, and is less sensitive in chronic cases. Furthermore, PCR can detect Brucella DNA quickly and precisely, especially in acute infections; however, its accuracy may be reduced in chronic or low‐bacterial load infections. However, I‐ELISA identifies host antibodies rather than the pathogen, which makes it susceptible to false positives due to cross‐reactivity with other Gram‐negative bacteria and limits its capacity to distinguish between current and prior infections or vaccination status. I‐ELISA's sensitivity and specificity vary by antigen, host species, and infection stage, which can lead to under‐ or overdiagnosis (Bassiony et al. 2011; Mohseni et al. 2017).

In many regions and scenarios, there is no gold standard diagnostic for brucellosis, which complicates efforts to assess test performance accurately. In such cases, LCA provides a valuable statistical alternative for evaluating diagnostic accuracy. LCA enables estimation of the true sensitivity and specificity of tests like I‐ELISA, PCR and culture by considering the observed test results as arising from unobserved (latent) disease states, thereby bypassing the need for a gold standard. Bayesian variants of LCA can incorporate prior information and account for covariance between tests, providing robust estimates that facilitate calibration of test thresholds and improve diagnostic accuracy in both human and animal populations (Elsohaby et al. 2022; Kneipp et al. 2024). Together, these considerations highlight the importance of combining multiple diagnostic approaches and advanced statistical methods like LCA to achieve a reliable brucellosis diagnosis when definitive tests are lacking.

One limitation of our study is the definition of a proper gold standard for evaluating the diagnostic performance of the tested assays. Although bacterial culture is traditionally considered the gold standard for brucellosis diagnosis, its application, particularly in milk samples, is severely constrained by its low sensitivity. In our study, Brucella spp. were isolated from only 16.6% (55/330) of milk samples collected from seropositive cows, underscoring the limited reliability of culture as a reference method in this context. Therefore, we employed the I‐ELISA as a comparator for ROC analysis due to its widely recognised diagnostic accuracy and high manufacturer‐reported sensitivity and specificity. However, we acknowledge that these results should be interpreted as measures of comparative performance rather than definitive diagnostic accuracy.

5. Conclusion

Sensitivity and specificity are essential for effective disease screening and population management, and understanding test accuracy is key for designing successful eradication strategies. Given the substantial influence of screening method selection on diagnostic efficacy, the results of this study strongly suggest that SAT and 2‐ME should not be used as screening tools for brucellosis in endemic areas. Our analysis demonstrated that I‐ELISA and, to a lesser extent, RBT are highly effective in detecting Brucella infections, capturing all positive cases identified by 2‐ME and SAT. Their combined use ensures broader coverage, reduces the reliance on more complex or costly confirmatory tests and facilitates early detection. This approach is crucial for timely intervention, effective disease control and successful management of brucellosis in endemic regions.

Author Contributions

Faranak Abnaroodheleh: conceptualisation, investigation, funding acquisition, writing – review and editing, methodology, validation, resources, supervision, data curation. Fereshteh Ansari: conceptualisation, investigation, funding acquisition, writing – original draft, methodology, validation, visualisation, software, formal analysis, resources, supervision. Youcef Shahali: conceptualisation, investigation, project administration, formal analysis, software, data curation, supervision, resources, visualisation, writing – review and editing, writing – original draft. Maryam Dadar: conceptualisation, investigation, funding acquisition, writing – original draft, writing – review and editing, visualisation, validation, methodology, software, formal analysis, project administration, resources, supervision, data curation.

Ethics Statement

All animals in this study were treated according to the ethical standards for field studies approved by the Iranian Veterinary Organization in Tehran, Iran (Approval Code: IR. IVO.1399.001 at 1 June 2023). The dairy farmers were informed about the purpose of the investigation and gave their informed consent.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://publons.com/publon/10.1002/vms3.70566.

Abnaroodheleh, F. , Ansari F., Shahali Y., and Dadar M.. 2025. “Diagnostic Performance of Four Serological Assays for Bovine Brucellosis and Optimised Cutoff Thresholds in an Endemic Region of Iran.” Veterinary Medicine and Science 11, no. 5: 11, e70566. 10.1002/vms3.70566

Data Availability Statement

The data supporting this study's findings are available on request from the corresponding author.