Tropical Animal Health and Production

, Volume 41, Issue 5, pp 723–729

Effectiveness of Rose Bengal test and fluorescence polarization assay in the diagnosis of Brucella spp. infections in free range cattle reared in endemic areas in Zambia

Authors

    • Department of Disease Control, School of Veterinary MedicineUniversity of Zambia
  • A. Lund
    • National Veterinary Institute
  • K. Nielsen
    • Animal Disease Research, InstituteCanada Food Inspection Agency
  • G. Matope
    • Department of Paraclinical Veterinary Studies, Faculty of Veterinary MedicineUniversity of Zimbabwe
  • M. Munyeme
    • Department of Disease Control, School of Veterinary MedicineUniversity of Zambia
  • K. Mwacalimba
    • Department of Disease Control, School of Veterinary MedicineUniversity of Zambia
  • E. Skjerve
    • Department of Food Safety and Infection BiologyNorwegian School of Veterinary Science
Original Paper

DOI: 10.1007/s11250-008-9244-0

Cite this article as:
Muma, J.B., Lund, A., Nielsen, K. et al. Trop Anim Health Prod (2009) 41: 723. doi:10.1007/s11250-008-9244-0

Abstract

The effectiveness of Rose Bengal test (RBT) and fluorescence polarization assay (FPA) in diagnosing cattle brucellosis in endemic areas was assessed and RBT and FPA test agreement was compared (n = 319). The sensitivity of RBT and FPA in detecting low Brucella titres were evaluated in paired sera (n = 34). A logistic regression model was constructed to predict cattle test result in FPA using RBT as the main predictor and incorporating bio-data and animal history. There was 79.3% agreement between the RBT and FPA (Kappa = 0.59; Std error = 0.05; p = 0.000) and a high correspondence between high RBT scores and positive FPA results suggesting that sera with high RBT score may not require confirmation with tests such as competitive-ELISA or CFT. High FPA cut-off points were more likely to miss animals with low antibody titres. The RBT had a reduced ability in detecting low antibody titres compared to the FPA. FPA test interpretation was improved if a priori information, such as sex and age was used. Under the challenging disease surveillance conditions prevailing in rural Africa, field-testing methods that are sensitive and specific; allow single animal contact, low technical skills in data interpretation are suitable.

Keywords

CattleBrucellosisEndemic areaFPARBTZambia

Abbreviations

CFT

Complement fixation test

ELISA

Enzyme linked immunossorbent assay

FPA

Fluorescence polarization assay

mP

Mili-polarisation units

RBT

Rose Bengal test

Se

Sensitivity

Sp

Specificity

Introduction

Brucellosis, caused by Brucella abortus, is still a big problem in livestock production in sub-Saharan Africa (McDermott and Arimi 2002). This disease causes both social and economic problems in affected communities. Control and prevention of brucellosis involve use of diagnostic tests of varying effectiveness and diagnostic sensitivity. Various diagnostic techniques, their effectiveness and limitations have been described for field and laboratory diagnosis of brucellosis (Alton et al. 1975; Nielsen et al. 1996b; Muma et al. 2007).

Diagnosis of brucellosis in pastoral cattle offers a number of challenges that include: farmer compliance, lack of diagnostic facilities, and limited resources of accessing animals and running the tests. Most conventional brucellosis diagnostic tests have certain limitations that make them ineffective diagnostic tools under the prevailing African traditional farming practices. Procedures that are laboratory based such as the Complement Fixation Test (CFT) are costly, time consuming and require a second visit when taking the results back to the farmer (Nielsen 2002). In addition, tests such as the Brucellin skin test that involve double contact with the animals are also difficult to apply because of the reduced chances of accessing the animals on a second visit. In traditional livestock farming, animal grazing areas or water sources are often located at long distances from homes (Perry et al. 1984; Muma et al. 2006). Therefore, animal testing may be perceived by farmers to be an inconvenience, especially during farming season. In addition, lack of animal holding facilities make animal restraining a tiresome process since animals have to be physically restrained and a double visit makes the exercise daunting. Other tests such as the Enzyme Linked Immunossorbent Assay (ELISA) that require elaborate equipment and skilled interpretation may not be very applicable (Nielsen 2002). All these factors may inevitably affect farmer compliance and ultimate success of a testing programme. A suitable test under such conditions should be affordable, diagnostically sensitive and specific rapid test that is convenient to use with a single contact with the animals. The RBT and FPA seem to have such features. RBT has been used to diagnose brucellosis in animals despite its limitations (OIE 2004). The RBT is reported to have high sensitivity (Se), but a low specificity (Sp) (OIE 2004). The FPA, which is a more recent test, is considered suitable for detection of serum antibodies to Brucella spp in multiple species including, humans, livestock and wildlife. In ruminants, the test has a high diagnostic specificity and sensitivity compared to other tests (Nielsen et al. 2000; Duran-Ferrer et al. 2004). FPA has been observed to be superior in the diagnosis of brucellosis in buffaloes compared to RBT (Montagnaro et al. 2008) and in sheep, the test has been observed to be efficient and accurate in diagnosing B. melitensis (Minas et al. 2007).

In our earlier study, we estimated the Se and Sp of RBT, FPA and c-ELISA in Zambian traditional cattle (Muma et al. 2007) and observed that RBT and FPA had better test performance indices (Table 1). This study therefore was aimed at testing the effectiveness of FPA and RBT in diagnosing Brucella spp. infection under the endemic situation in traditional cattle rearing system in Zambia.
Table 1

Highest Sensitivity and Specificity performance estimates with 95% confidence limits for brucellosis serological tests in Zambian traditional cattle in endemic areas

Serological tests

Sensitivity (Se) %

Specificity (Sp) %

Test index (Se + Sp)

RBT

93 (84–99)

82 (60–99)

175

FPA

89 (74–99)

93 (85–99)

182

c-ELISA

97 (93–100)

60 (33–96)

157

Source: Muma et al. 2007.

Materials and methods

Cattle sera

The study was based on randomly collected sera from traditionally reared cattle from Blue Lagoon, Lochinvar and Kazungula. Sera were collected between August 2003 and November 2005. In this study, we examined 391 of the 1245 sera earlier collected (Muma et al. 2006, 2007). We further examined paired sera (n = 34) that were collected approximately a year apart. For paired sera, the first set was collected between August 2003 and September 2004 and was tested on RBT and c-ELISA while the second set was collected between September 2004 and November 2005 and tested on RBT and FPA. Details of the sampling techniques are described in details in our earlier reports (Muma et al. 2006, 2007).

Rose Bengal test

Antibodies to Brucella spp. were detected by testing of serum samples using RBT as described by Alton et al. (1988). Brucella abortus antigen (VLA, UK) was used to screen sera for the presence of antibodies to Brucella spp. The degree of agglutination was graded on an ordinal scale from 0 (no agglutination) to 3 (coarse clumping), with corresponding RBT scores of 0, 1, 2 and 3. All doubtful reactions were recorded as negative or zero scores.

ELISA test

Svanovir™ Brucella-Ab c-ELISA kits (Svanova Biotech AB Uppsala, Sweden) was used to determine anti-Brucella spp. antibody titres and conducted according to the manufacturer’s instructions. Sera and controls were run in duplicates. The optical densities (OD) were measured at 450 nm in a micro-plates photometer (Humareader, Model 18500/1, Awareness Technology, Inc. Germany). The threshold for determining sero-positivity was based upon the manufacturer’s recommendations (≥30%), with antibody titres recorded as percentage inhibition (PI) defined by the ELISA kit supplier as:
$$PI = 100 - \frac{{\left( {Mean..OD_{Samples/Ctrl} . \times ..100} \right)}}{{Mean..OD_{Conjugate..control..Cc} }}$$

PI=percentage inhibition, OD=optical density for sample, control (Ctrl) and conjugate control (Cc), respectively.

Fluorescence polarisation assay

The FPA assay was done as earlier described (Nielsen et al. 1996a, 1998, 1999, 2002) using the Sentry Fluorescence Polarisation Analyser (Diachemix Sentry TM 100, single tube reader, Diachemix LLC, Wisc. USA). Positive and negative reference cattle sera were included in each lot of samples tested. Data was expressed as milli-polarisation units (mP). The threshold for determining seropositivity was set at >90 mP (Nielsen et al. 1996a, 2002; Nielsen and Gall 2001).

Statistical analysis

Data was stored and managed in excel before transferring to Stata statistical package (version 9 for Windows, Stata Corporation, college station, TX). For practical purposes, an individual animal was assumed to be the unit of interest. Further, having antibody titres against Brucella spp. and thus testing positive both on RBT and FPA was the outcome of interest. The association between RBT and FPA was assessed using the Kappa agreement test while the logistic regression model was used to predict a positive FPA test outcome. The model was constructed as prescribed by (Dooho et al. 2003). We built the model using the forward selection procedure by applying the iterative maximum likelihood estimation procedure and statistical significance contribution of individual predictors (or group of predictors) to the models tested using the Walds test and the likelihood ratio test (Dooho et al. 2003). Using the paired sera, the test performance of RBT and FPA in detecting declining antibody titres was assessed. Paired sera comprised a sub-set of animals with history of clinical signs that are compatible with Brucella spp. infections such as abortion and fertility problems. In order to assess titre increase/decrease in paired samples, some assumptions were made. Presence of a clinical picture compatible with brucellosis and demonstration of anti-Brucella specific antibodies was set as criteria to assume Brucella spp. infection (Kiel and Khan 1987). Further, we assumed that all animals that tested positive (T+) in both RBT and c-ELISA in 2003/4 were infected (I+) and non-infected (I) if positive in only a single test (T) or negative in both. For further FPA analyses, we estimated an intrinsic cut-off point based on the distribution of antibodies in the negative and positive sub-populations. Under this assumption, the number of animals infected (I+) and non-infected (I) in 2004/5 were determined and titre increase or decrease was also assessed.

Results

There was 79.3% agreement between the RBT and FPA (Kappa = 0.59, Std error = 0.05; p = 0.000). The effect of RBT screening prior to testing with FPA was assessed using the basic and full logistic regression models with RBT scores as a main predictor and dichotomised FPA results as the outcome of interest (Table 2). The basic model was found to be significant (Number of observations = 391; Hosmer-Lemeshow chi2 (2) = 0.00; Prob > chi2 = 1.0) and made correct predictions 82.6% of the times. We observed a high correspondence between high RBT score and high FPA readings and a poor correspondence between low RBT score and FPA positivity (Table 2). The usefulness of other epidemiological information in assisting accurate FPA predictions was assessed in the full model that included sex and age. Equally, the full model was significant (Hosmer-Lemeshow chi2 (20) = 20.02; Prob > chi2 = 0.46) and made correct classification 86.5% of the times showing a better prediction when other background information was considered (Fig. 1). The final model showed that sensitivity and specificity were optimised when the probability of cut-off was 55% (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs11250-008-9244-0/MediaObjects/11250_2008_9244_Fig1_HTML.gif
Fig. 1

ROC-curve, a plot of sensitivity as a function of 1-specificity for the full model for the FPA

https://static-content.springer.com/image/art%3A10.1007%2Fs11250-008-9244-0/MediaObjects/11250_2008_9244_Fig2_HTML.gif
Fig. 2

Sensitivity (Se) and specificity (Sp) at different cut-off values for the full model for the FPA test

Table 2

Basic and full logistic regression models for prediction of a positive Brucella spp. infection in the FPA assay applied on traditional Zambian cattle (n = 391)

Variable

Levels

Basic model

Full model

aOdds Ratio (95% CI)

p-Value

aOdds Ratio (95% CI)

p-Value

RBT agglutination score

Zero (0)

1

-

1

-

One (1)

13 (7–23)

0.000

9 (5–7)

0.000

Two (2)

20 (10–41)

0.000

18 (8–17)

0.000

Three (3)

48 (17–137)

0.000

51 (17–156)

0.000

Sex

Male

-

-

1

-

Female

-

-

2 (1–5)

0.060

Age category

2–4 years

-

-

1

-

4.5–5 years

-

-

5 (2–10)

0.000

5.5–7 years

-

-

5 (2–11)

0.001

>7 years

-

-

7 (3–18)

0.000

aOR showing increased chance of an FPA test positive result with increasing RBT agglutination score (and increasing age).

Out of 34 animals analysed in paired samples, 20 and 14 animals were classified as ‘infected’ and ‘non-infected’, based on clinical history and serology according to 2003/4 results. Figure 3 shows the distribution of antibodies as estimated by FPA in milli-polarization units (mP) from paired sera plotted according to their RBT results approximately one year later. Two distinct data sub-populations with median and inter-quartile ranges of 74.3 mP (65.8 to 78.5) and 199 mP (80.9 to 268.3) were observed for the RBT negative and positive populations, respectively. For further analysis of FPA data in this data set, the intrinsic cut-off point was set to 80 mP based on the analysis of the inter-quartile ranges in the negative and positive sub-populations.
https://static-content.springer.com/image/art%3A10.1007%2Fs11250-008-9244-0/MediaObjects/11250_2008_9244_Fig3_HTML.gif
Fig. 3

Comparison of the distribution of anti-Brucella antibody titres (mP) of FPA in RBT positive (RBT = 1) and negative (RBT = 0) groups (n = 34)

Table 3 shows the classification of those animals (n = 34) were paired sera were obtained according to RBT2004/5 and FPA (80 and 90 mP) test results.
Table 3

Cross-tabulation of Brucella spp. antibody FPA and RBT results and brucellosis infections reference tests (clinical signs and seropositivity) (n = 34) in brucellosis endemic areas of Zambia

Infection status (Clinical signs + seropositivity)

  

I+

I

FPA80 mP

T+

19

7

 

T

1

7

  

20

14

  

I+

I

FPA90 mP

T+

13

7

 

T

7

7

  

20

14

  

I+

I

RBT2004/5

T+

12

5

 

T

8

9

  

20

14

Notations T+, T for positive and negative test result; I+, and I for positive and negative infection status by reference method (Clinical sign + RBT2003/4 + c-ELISA); (n = 34: 20 ‘infected’ and 14 ‘non-infected’).

Results indicate that there was a distinct difference in the classification of animals between RBT and FPA90 mP and also between the two FPA cut-off points (90 mP and 80 mP). The FPA positive and negative predictive values increased from 65% to 73% and 50% to 87.5%, respectively when the FPA cut-off point was lowered from 90 mP to 80 mP. Further, a significant difference was observed in animal classification when FPA80 mP was used alone or when RBT and FPA80 mP were interpreted in parallel (p = 0.000).

Discussion

The aim of this study was to assess the effectiveness of using RBT and FPA in testing for Brucella spp. infections in traditional cattle in Zambia, taking into account several challenges encountered in testing animals under the prevailing rural conditions. Considering that brucellosis is endemic in most parts of Zambia (Muma et al. 2006) and that the majority of Zambian cattle (80%) are raised under the traditional livestock production system (Anon 2000), it was important to test the effectiveness of usage of these test assays under the prevailing conditions. Since success of any testing program is a function of many factors including effectiveness of usage of a diagnostic test, this study showed that both the RBT and FPA could be effectively used in brucellosis diagnosis in traditional livestock system because both allow single animal contact, cow-side field testing of animals and on-spot delivery of test results. In our earlier report we indicated that both RBT and FPA have reasonable Se and Sp under the Zambian condition although they still require optimisation (Muma et al. 2007). Although the RBT is recommended under this set-up, it has some problems and the most obvious and most dangerous is prozoning, where sera with high levels of antibody results in non-visible reactions with the RBT antigen (Alton et al. 1988). Thus a strong Brucella positive serum may be classified as negative in contrast to results on other serological assays. In contrast, primary binding assays such as the FPA do not prozone to the point of diagnostic confusion making the FPA more preferable.

We observed a high correspondence between high RBT score results which corroborates with what has been earlier observed and consolidated the observation that sera with high RBT score may need not be re-tested with confirmatory tests such as c-ELISA or CFT (Omer et al. 2001). Further, the logistic regression model showed that FPA test interpretation was more reliable if a priori information, such as sex, age, and history of abortion was used in selecting appropriate cut-off points for the FPA. Omer et al. (2001) observed that information about a population or individuals may assist in the interpretation of a test result. Information such as sex, age, and previous disease history may increase the likelihood of correctly diagnosing Brucella infection in the FPA.

Analysis of paired sera collected one year apart from same animals showed that the FPA maintained a better classification of animals into ‘infected’ and ‘non-infected’ compared to RBT between the one year interval. Our initial classification of these animals into the ‘infected’ and ‘non-infected’ groups was based on serological response to the RBT and c-ELISA and presence of Brucella-related clinical signs. Since c-ELISA positive animals that were negative on RBT were classified ‘non-infected’, this could introduce misclassification of truly infected animals due to the prozoning effect encountered with RBT assay. However, only one animal was classified ‘non-infected’ on this basis and thus misclassification bias was less likely to influence the final interpretation of results.

It would then be suggested that in situations of active Brucella infections, lower FPA cut-off point from the conversional 90 mP would be suitable to use in order to increase the chance of detecting low titres. In our study, we observed the cut-off 80 mP gave a better interpretation in the face of active Brucella infections in cattle herds. This however requires further study since these tests need to be optimised under the prevailing local conditions. The need to adjust the FPA cut-off point in different epidemiological situations has been recommended in order to improve FPA test performance (Ramirez-Pfeiffer et al. 2007; Montagnaro et al. 2008).

Despite the limited sample size, the study, has demonstrated that RBT and FPA could effectively and efficiently be used to diagnose Brucella infections in traditionally reared cattle in endemic areas taking into account the raised concerns. Since both the RBT and FPA can be performed in the field with no or little energy demand, they possibly could be suitable for use under the existing conditions in traditional cattle farming areas. However, both tests may need to be optimised under the same conditions they are to be used in order to achieve maximum performance. The study has further shown the need to take into account other epidemiological information to improve the detection of Brucella infected animals. Further, for re-testing of previously brucellosis-diagnosed animals between intervals, the FPA would be recommended because of its relative good ability to detect low antibody titres.

Acknowledgement

The Norwegian Council for Higher Education’s Programme for Development Research and Education (NUFU) funded this study and we are greatly indebted to this organization. We also acknowledge the cooperation we received from the farmers and help from field staff under the Ministry of Agriculture. We are further grateful to the staff at the University of Zambia, School of Veterinary Medicine who helped with the work.

Copyright information

© Springer Science+Business Media B.V. 2008