Evaluation of microscopic agglutination test as a diagnostic tool during acute stage of leptospirosis in high & low endemic areas
Vijayachari, P
Background & objectives: Making a diagnosis on the results of a single microscopic agglutination test (MAT) is difficult because of the uncertainties about the cut-off titre. The present study was conducted to determine the significant titre for a single MAT in areas of high and low endemicity for leptospirosis.
Methods: A total of 1944 serum samples were collected from healthy individuals and confirmed patients residing in areas of high and low endemicity. All the sera were screened by MAT using 10 live leptospiral strains as antigens. From the distribution of titres among healthy individuals and in patients, the sensitivity and specificity at different cut-off titres were calculated. Likelihood ratio positive (LR+), likelihood ratio negative (LR-), and LR+/LR- were calculated. Receiver operating characteristics (ROC) curves were plotted for the early and late stages of the disease in both the areas.
Results: The ROC plot was totally below the no benefit line during the first week of illness in high endemic area. During the second to fourth weeks it showed better characteristics and the best cutoff titre was 1:200, where the sensitivity was 93.4 per cent and specificity 74.7 per cent LR+ LR- ratio was 41.82 indicating reasonable separation between the positive and negative test results. In the other states the ROC plot was above the no benefit line even during the first week, the best cut-off being 1:50 where the sensitivity was 56.7 per cent and specificity was 90.6 per cent. During the second to fourth weeks the test showed the best characteristics in the low endemicity regions with an ROC curve having the ideal shape. Best cut-off was at 1:100 where the sensitivity was 96.6 per cent and specificity 94.8 per cent LR+ LR- ratio was 523.25 indicating a wide separation between the positive and negative test results.
Interpretation & conclusion: MAT does not have any diagnostic value during the first week, particularly in high endemic areas. The best cut-off to be used will be 1:50 in low endemicity areas during the first week, 1:100 during the second to fourth week and 1:200 in high endemicity regions during the second to fourth weeks.
Key words Leptospirosis – microscopic agglutination test (MAT)
Leptospirosis, an emerging public health problem in developing countries, is difficult to diagnose due to protean clinical presentations and complexities in the diagnostic techniques’. Isolation of the organism from clinical material, which is definite proof of the diagnosis, is a time consuming procedure requiring expensive laboratory facilities and the success rate is very low2. Due to these problems, the diagnosis of leptospirosis usually depends upon serological techniques.
The microscopic agglutination test (MAT) remains the corner stone of sero-diagnosis of leptospirosis. MAT is also helpful in understanding the epidemiology of the disease. Despite the fact that MAT is the oldest among the serological tests for diagnosis of leptospirosis, difference of opinion exists regarding its technique, and the interpretation and significance of the results.
MAT is ideally performed on paired serum samples. The criterion for a definite diagnosis of current leptospiral infection is a four fold rise in titre or seroconversion. However, in actual practice, obtaining paired samples from patients is very difficult. Very often the diagnosis is made on a single sample received during the initial stage of the disease. This has led to a debate on the cut-off titre to be used as a diagnostic titre indicating current leptospiral infection.
The cut-off titre used to demarcate a positive test result from a negative one is a factor strongly influencing the accuracy of the results. The sensitivity of the test is the probability of it giving a titre equal to or above the cut-off in samples from true patients. Similarly the specificity is the probability of the test giving a titre less than the cutoff in samples from persons without the disease. These probabilities depend upon the distribution of titre in true patients and normal individuals. Since the agglutinins stay for prolonged period of time after infection- a proportion of the healthy individuals will have detectable levels of antibodies. At the same time, in true patients it takes some time for the antibodies to reach detectable levels. These two sets of persons i.e., healthy individuals having detectable antibodies and the true patients who have not acquired antibodies in detectable levels account for the false positive and false negative results of the test respectively. The former depends upon the endemicity of the disease in an area and hence the endemicity is a factor that has to be taken into account while fixing a cut-off titre for optimal accuracy of the test.
Reference laboratories often face this problem as they receive samples from areas of different endemicity. Hence we undertook this study to estimate the best cut-off titres for samples from areas of high and low endemicity.
Material & Methods
A total of 1944 serum samples collected during 1999-2000 from healthy individuals and confirmed patients during different stages of disease were used in the study. The patients and healthy individuals were from Andaman Islands, where leptospirosis is highly endemic4,5 and from four other states viz., Kashmir, Kerala, Maharashtra and Orissa representing low endemicity areas. There were 181 samples from patients, 122 from Andamans (61 acute and 61 convalescent) and 59 from other states (30 acute and 29 convalescent). There were 1763 samples from healthy individuals (1064 from Andamans and 699 from other states). Isolation of the organism from blood or a four fold rise in titre or seroconversion in paired serum samples by MAT was the basis of confirming current leptospiral infection in patients.
Serological techniques: MAT was performed in micro titre plates using 10 live leptospiral strains as antigens following standard procedure6. The strains belonged to serovars grippotyphosa, australis, icterohaemorrhagiae, pomona, rachmati, canicola, poi, hebdomedis, hordjo and patoc. These serovars represented serogroups Grippotyphosa, Australis, Icterohaemorrhagiae, Pomona, Autumnalis, Canicola, Javanica, Hebdomedis, Sejroe and Semaranga respectively. MAT was performed on all samples at dilutions 1:50, 1:100, 1:200 and 1:400. Those found positive at 1?400 were titrated up to end titres.
The proportions of patients giving positive results in individual MAT at different cut-off titres were calculated. Statistical significance of the difference in the positivity rates and the distribution of titres among patients from Andaman and other states were tested using Pearson ;2 test. The percentage of patients giving positive results at any given cut-off titre is the sensitivity of the test at that cut-off. Similarly the percentage of healthy individuals giving positive results at any given cut-off titre is the false positivity rate. Using this method sensitivity and specificity of single MAT, at different cut-off titres, were calculated separately for high endemicity areas and low endemicity area.
Likelihood ratio positive (LR+), likelihood ratio negative (LR-) and LR+ LR- ratio (LR+/LR-) at different cut-off titres were calculated. These ratios were calculated for the test done on samples during the first week of illness and second to fourth week of illness. Receiver operating characteristic (ROC) curves were drawn for each of the areas for the test done on early and late stages of the disease and the best cut-off titres were chosen based on the distance of the data point from the upper left corner of the chart.
Results
During the first week of disease, the highest titre observed among patients from Andamans was 1:800 and that in patients from other states was 1:1600. 47.5 per cent of the patients in Andamans and 56.7 per cent in other states gave a titre of 1:50 or above (P>0.05). Sensitivity at different cut-off titres was calculated from the proportion of patients giving a titre equal to or greater than the cut-off titre. During the first week the sensitivity dropped from 47.5 per cent at 1:50 to 14.8 per cent at 1:200 in the case of Andamans and from 56.7 to 26.7 per cent in the case of other states. Sensitivity at all cut-off titres was slightly lower in the case of patients from Andamans compared to patients from other states.
During second to fourth weeks, the highest titre observed among patients from both Andamans and other states was 1:3200. All the patients from both areas gave a titre of 1:50 or above. In the case of both the areas the sensitivity.dropped only to around 93 per cent from 100 per cent when the cut-off titre was raised from 1:50 to 1:200. However, the drop was steeper for further increase in cut-off. At 1:800 it was 54.1 per cent for Andaman Islands and 69 per cent for other states.
The distribution of titres among the healthy population in Andamans and other states is shown in Table 1. 47.8 per cent of the healthy individuals from Andamans and 90.6 per cent of those from the other states were seronegative. Thus the seroprevalence was 52.2 per cent in Andamans and 9.4 per cent in other states. The highest titre observed was 1:3200 in the case of samples from both Andamans and other states.
The cumulative percentage against each titre represents the specificity for a cut-off equal to the next titre. For example in Andamans, 63.7 per cent gave a titre of 1:50 or less. If the next titre i.e., 1:100, is taken as the cut-off, the.test would have correctly identified this 63.7 per cent as negative. Sensitivity, specificity, likelihood ratios and LR+ LRratio are shown in Table II. The specificity in Andamans was very low at 47.8 per cent at a cut-off titre of 1:50, whereas it was 90.6 per cent in other states. By raising the cut-off titre to 1:400, the specificity almost doubled in the case of Andamans. However, in the other states, the gain was only 7 per cent.
In the first week of illness, LR+ and LR+/LRwere below one for all cut-off titres in the Andamans, which indicate that the test does not yield any useful information during the early phase of the disease and the separation between a positive test and negative test was narrow. However, in the other states, LR+ was above one for all titres from 1:50 to 1:1600 with the highest value of 10.98 for the cut-off of 1:200. The highest LR+ LR- ratio was also for the titre 1:200. However, the value 14.61 was much below the desirable level of 507 indicating that even the maximum separation between a positive and negative test was low. During the later phase of the disease, LR+ was above 1 for all titres in Andaman and the highest value was for 1:3200. LR-, for which the lowest value will be the desired one, was lowest at 1:50 and increased with the cutoff titre. LR+/LR- could not be calculated for titres 1:50 and 1:100 as the false negative error rate was 0 per cent at these titres. The highest value was 41.82 observed for the titre 1:200. This was much closer to the desired level of 50.
ROC curves for Andaman and other states are shown in Figs I and 2. The curve for early phase of the disease in Andaman Islands lies completely below the no benefit line. So at no cut-off titre did the test give useful information. The curve for the later phase of illness is above the no benefit line and rises sharply with false positivity rate during the initial part of curve, which is a characteristic of a good curve. The point on the curve nearest to the upper left corner of the chart corresponds to the titre 1:200, indicating that in the Andaman islands during the second to fourth week of illness this is the ideal cutoff titre. The curves for both first week of illness and second to fourth week of illness in other states lie above the no benefit line. Both the curves do not extend beyond 10 per cent false positivity rate. The point nearest to the upper left corner on the curve for first week of illness corresponds to the cut-off titre of 1:50 and that on the curve for second to fourth week corresponds to 1:100, which will be the ideal cut-off during the early and late phase of the illness respectively.
Discussion
The present study was an attempt to find out the cut-off titres most suitable for diagnosis in high and low endemic areas. Andaman Islands is a highly endemic area for leptospirosis 4,5-8 with more than 50 per cent seroprevalence. It has been observed that about 29 per cent of the population acquires leptospiral infection during the harvest season9. In the other states, from where samples were obtained, leptospirosis is known to occur but not as commonly as in the Andamans. Although there is no recent information, there had been reports of leptospirosis from Kashmir earlier’O. In Kerala leptospirosis has been occurring in certain areas I though the exact seroprevalence is not known. Leptospirosis is known to occur in Mumbai (Bombay) in Maharashtra]2. Recently there was an outbreak of leptospirosis in Mumbai following heavy rainfall and flooding (unpublished data). Outbreak of leptospirosis has been reported from Orissa after a cyclone and flooding in October 199913.
Seroprevalence in the Andamans was more than five times that in other states. The distribution of titres in healthy individuals also showed distinct patterns in the two areas. In the Andamans a higher proportion of the population had high. titres. Because of this the specificity was low at lower cut-off titres. It started at 47.8 per cent at the lowest cut-off of 1:50 and reached 98.8 per cent at 1:3200. The specificity was 74 and 87.1 per cent at cut-off titres of 1:200 and 1:400 respectively. In the other states, specificity was above 90 per cent for all cut-off titres.
The test did not show any desirable characteristic during the first week of illness in high endemicity area. The likelihood ratio positive was below one and the likelihood ratio negative was above one for all titres. The LR+ LR- ratio was very small indicating that the separation between a positive and negative test result was small. During second to fourth week, the test had better characteristics. LR+ was higher than one at higher cut-off titres and LRwas near zero at all titres. The highest calculated LR+ LR- ratio at 1:200 was close to the desired level of 50. The ROC curve, which was good characteristics, indicated an ideal cut-off of 1:200. At this cut-off the test had a sensitivity of 93.4 per cent and a specificity of 74.1 per cent. Although it would be desirable -to have a better specificity, at the next dilution of 1:400 the sensitivity was only 75.4 per cent.
In the case of mildly endemic area, the test had better characteristics for both first week and second to fourth weeks. During the first week, LR+ was above I for all cut-off titres and LR- was somewhat lower than in the case of the Andamans. However, the highest value of LR+/LR- observed at a cut-off titre of 1:200 was much below the desired minimum level of 50. The best cut-off titre indicated by the ROC curve i.e., 1:50 yielded a sensitivity of only 56.7 per cent, hence the usefulness of the test during the first week is limited. The test had the best characteristics during the second to fourth weeks in mildly endemic area. LR+ was much above one and LR- was close to zero. LR+ LR- ratio was above 500 for titres 1:100 and 1:200 indicating very good separation between a positive and negative test. The ROC plot has an ideal shape with a steep rise in the beginning. The best cut-off, as observed in the curve was 1:100, where the sensitivity was 94.8 per cent and specificity 96.6 per cent. In the case of nonendemic areas like Kashmir, demonstration of antibodies at a titre of 1:50 itself would be near confirmatory proof of recent leptospiral infection in the presence of symptoms 14 as the specificity was above 90 per cent.
In the present study the number of’patients was lower compared to the number of healthy individuals. Hence the estimates of titre distribution in patients are less accurate than the estimates in the general population. This implies that the assessment of sensitivity would be less accurate than specificity, as sensitivity depends upon the distribution of titres in the patients.
A few IgM based screening tests-5-17 have come into use more recently. Although these tests are sufficient for initiating treatment in an emergency, because of the broadly reactive nature of IgM antibodies the specificity of these tests is comparatively lowl7,18. MAT has a place in the diagnosis of leptospirosis because of its better specificity and the additional information it gives about the infecting serogroup. However, the results of MAT have to be interpreted cautiously, particularly when paired samples are not available. Information about the background seroprevalence of the area from where the sample was obtained is crucial. When samples from a particular area have to be processed routinely, it is better to determine the ideal cut-off by carrying out serosurveys and finding out the distribution of titres in the community.
Acknowledgment
The authors acknowledge the help and cooperation of Dr Azra Shah, Dept. of Pathology and Microbiology, Sher-lKashmir Institute of Medical Sciences, Srinagar, Dr B. Ravindran, Regional Medical Research Centre, Bhubaneswar, Dr S. Sajith Kumar, Dept. of Community Medicine, Medical College, Kottayam, Dr Preeti Mehta, and Dr Chaya Kumar, Dept. of Microbiology, K.E.M. Hospital, Mumbai. The authors thank Dr Sameer Sharma, Sh. Umapati and Sh. Paritosh De, for technical assistance.
References
1. Faine S. Guidelines for the control of leptospirosis. WHO Offset Publication No. 67. Geneva: World Health Organization, 1982 p. 76-8.
2. Merien F, Baranton G, Perolat P. Comparison of polymerase chain reaction with microagglutination test and culture for diagnosis of leptospirosis. J Infect Dis 1995; 172: 281-5.
3. Blackmore DK, Schollum LM, Moriarty KM. The magnitude and duration of titres of leptospiral agglutinins in human sera. N Engl J Med 1984: 97: 83-6.
4. Sehgal SC, Murhekar MV, Sugunan AP. Seroprevalence of leptospirosis in North Andaman. Indian J Med Microbiol 1994; 12: 289-91.
5. Murhekar MV, Sugunan AP, Vijayachari P, Sharma S, Sehgal SC. Risk factors in the transmission of leptospiral infection. Indian J Med Res 1998; 107: 218-23.
6. Wolff, JW. The Laboratory diagnosis of leptospirosis Charles C Thomas, Publisher Springfield, Illinois lJ.S.A., 1954 p. 39-62.
7. Jekel JF, Elmore JG, Katz DL. Epidemiology, biostatistics and preventive medicine. Philadelphia: WB Saunders Company; 1996 p. 87-97.
8. Sehgal SC, Murhekar MV, Sugunan AP. Outbreak of leptospirosis with pulmonary involvement in North Andaman. Indian J Med Res 1995; 102: 9-12.
9. Sehgal SC, Sugunan AP, Murhekar MV, Sharma S, Vijayachari P. Randomised controlled trial of doxycycline prophylaxis against leptospirosis in an endemic area. Int J Antimicrob Agents 2000/ 13: 249-55.
10. Franklin GD. A mild epidemic of jaundice. Indian Med Gazette 1913; 48: 218-21.
11. Kuriakose M, Eapen CK, Paul R. Leptospirosis in Kolenchery, Kerala, India: epidemiology, prevalent local serogroups and serovars and a new serovar. Eur i Epidemiol 1997; 13: 691-7.
12. Dalai PM. Leptospirosis in Bombay: Report of five cases. Indian J Med Sci 1960; 14 : 295-301.
13. Leptospirosis, India. Report of the investigation of a postcyclone outbreak in Orissa, November 1999. Wk1Y Epideniol Rec 2000; 75: 217-23.
14. Faine S. Clinical laboratory diagnosis of leptospirosis. In: Leptospira and leptospirosis. London: CRC Press 1994. p. 17-40.
15. Terpstra WJ, Ligthart CS, Schoone GJ. ELISA for the detection of specific IgM and IgG in human leptospirosis..7 Gen Microbiol 1985: 131: 377-85.
16. Arimitsu Y, Kmety E, Ananyina Y, Baranton G, Ferguson IR, Smythe Ld, et al. Evaluation of the one-point microcapsule agglutination test for diagnosis of leptospirosis. Bull World Health Organ 1994, 72: 395-9.
17. Gussenhoven GC, van der Hoorn MAWG, Goris MGA, Terpstra WJ, Hartskeeri RA, Mol BW, ei al. LEPTO Dipstick, a Dipstick assay for detection of leptospiraspecific Immunoglobulin M antibodies in human sera. J Clin Microbiol 1997; 35: 92-7.
18. Sehgal SC. Vijayachari P, Subramaniam V. Evaluation of Leptospira micro capsule agglutination test (MCAT) for serodiagnosis of leptospirosis. Indian J Med Res 1997; 106: 504-7.
P. Vihayachari, A.P. Sugunan & S.C. Sehgal
National Leptospirosis Reference Centre, Regional Medical Research Centre, (ICMR) Port Blair, India
Received June 15, 2001
Reprint requests: Professor S.C. Sehgal, DIrector, Regional Medical Research Centre (ICMR) Post Bag No 13, Port Blair 744101, India
Copyright Indian Council of Medical Research Sep 2001
Provided by ProQuest Information and Learning Company. All rights Reserved
