We found that leptospirosis was a common, but often clinically unsuspected, cause of fever among unselected patients seeking care in southern Sri Lanka. Farming and rice paddy work were associated with increased risk for leptospirosis, as was exhibiting acute febrile illness during the harvesting season (July–October). In our setting, testing acute-phase serum specimens alone for IgM was less sensitive and specific for diagnosing acute leptospirosis than was diagnosis by observation of clinical features.

Isolation of Leptospira spp. confirms acute infection, but requires special media that must be incubated for up to 13 weeks, and has low sensitivity (1). Therefore, the diagnostic standard for acute leptospirosis is a definitive rise in titer between paired serum specimens (1). Historically, these results have been obtained by the microscopic agglutination test (MAT). Serum specimens are first reacted with live antigen suspensions of different leptospiral serovars. After incubation, the serum-antigen mixtures are examined and titers determined. For paired serum specimens, the highest dilution of serum at which 50% agglutination occurs must be determined, a laborious and inherently subjective task (1).

Sensitivity is compromised if all locally relevant serovars are not represented, and live cultures of all serovars tested must be maintained whether live or formalin-killed antigens are used. Subculturing many Leptospira spp. weekly is hazardous for personnel, and laboratory-acquired infections occur (1). Reading a MAT requires a dark-field microscope, which is unavailable in most laboratories, including Karapitya Teaching Hospital. Furthermore, a MAT detects both IgM and IgG and lacks sensitivity and specificity when early acute-phase serum specimens alone are tested rather than paired specimens (1). Patients with fulminant illness may die before seroconversion occurs. A MAT may also be less sensitive than an IgM ELISA, even for convalescent-phase specimens. Relative to isolation of Leptospira spp., the reported sensitivities of MATs for acute, late acute-phase, and convalescent-phase serum specimens were 30%, 63%, and 76%, respectively, and of IgM ELISA, 52%, 89%, and 93%, respectively, in 1 study in Barbados (8). In another study, results of MAT and IgM ELISA for a single early acute-phase specimen were comparable (49%) (9).

To overcome the practical pitfalls of MATs, we chose to test paired serum specimens by IgM ELISA, which requires only an inexpensive plate reader, is relatively easy to perform, and provides objective, reproducible results as demonstrated by a parallel comparison of results of multiple commercial assays (9). Furthermore, results of performing IgM ELISA on paired serum specimens from patients from various geographic regions have compared favorably (sensitivity 86.5%, specificity 97.0%) with MAT results (10). These data suggest that at least 13.5% of febrile illnesses in our cohort were acute leptospirosis. We assayed paired specimens for IgM instead of IgG, because the kinetics of IgG are more variable (11,12). IgM generally appears within 1 week of symptoms and persists for months to years after infection (13) with titers higher than those of IgG throughout (11,14). Additionally, in some patients for whom leptospirosis is confirmed by culture and MAT, IgG never develops (11,13).

We chose a commercially available IgM ELISA that has performed comparably to others in detecting serovars likely present in southern Sri Lanka (9). IgM is inherently cross-reactive, and thus serovars themselves are not detected. In a recent study from central Sri Lanka, the predominant serovars were Mednensis and Hardjo, but others included Australis, Ballum, Canicola, Celledoni, Cynopteri, Pomona, and Robinsoni (15). Previously, identification of the serovar Icterhemorrhagiae in Sri Lanka led to control of the rodent vector (16). The assay we used has reliably reacted with serovars Icterohemorrhagiae, Canicola, Grippotyphosa, Bataviae, Pomona, Tarassovi, Copenhageni, Bratislava, Hebdomadis, Sejroe, Australis, Panama, Pyrogenes, Patoc, Hardjo, and Cynopteri (7,17).

Notably, detection of acute-phase IgM did not predict which patients had acute leptospirosis, despite its widespread use as an acute diagnostic test. Retrospective studies suggest sensitivities and specificities of 36%–53% for single acute-phase IgM and 90%–99% for MAT on paired serum specimens, respectively (9,10). The varied sensitivity likely reflects different case definitions and control groups, timing of acute-phase specimen collection (up to 42 days after onset), geography and serovar distribution, platforms and protocols (e.g., ELISA ± use of RF absorbent, indirect hemagglutination, and dot-ELISA and IgM dipsticks), and convenience sampling. Notably, ELISA of single (acute-phase) serum specimens has performed as well or better than MAT or indirect hemagglutination of single serum specimens, so those strategies are not advised (10).

The most widely recognized problem with using acute-phase IgM to identify acute leptospirosis is that many persons in disease-endemic areas are expected to have preexisting antibodies. Some have advocated higher cut-offs to discriminate between acute infection and preexisting antibodies (1), because patients may be harmed as much by incorrectly attributing fever to leptospirosis as by falsely excluding it. However, data to support this approach are lacking, and misclassification could occur both early in acute infection (impaired sensitivity because antibody is not yet present) and later (impaired specificity because antibody is persistent). In our rigorous comparison of single vs. paired serum specimens, we found acute-phase IgM had especially poor sensitivity (17.5%), since patients sought treatment early (≈3 days), and more acute infections were identified by seroconversion than by a definitive rise in titer. The median duration of illness in those diagnosed by rise in titer versus seroconversion tended to be longer (4 days [IQR 3–5 days] vs. 3 days [IQR 2–5 days], respectively; p = 0.09). Requiring a higher cutoff titer would further impair sensitivity. Hence, acute-phase IgM testing alone has multiple limitations for diagnosis of acute leptospirosis, regardless of the cut-off.

Only a few studies have evaluated the use of serologic testing for identifying leptospirosis in febrile cohorts. In Laos, 372 febrile patients were evaluated with ELISA (Panbio Ltd., Brisbane, Queensland, Australia) and immunochromatographic testing (ICT), which was compared with the MAT; acute leptospirosis (single titer >400 or 4-fold rise in titer) was identified in 23 (12.4%) of 186 patients (18). The sensitivity of ELISA and ICT was relatively high (60.9% and 47.3%, respectively), which could be explained by a long duration of fever (median 9 days). The sensitivity of ELISA for acute-phase versus convalescent-phase serum specimens was comparable (60.9% and 65.2%, respectively), but convalescent-phase serum specimens were obtained 4.5 days after acute-phase serum specimens. The specificity of both assays was similarly poor (65.6% for ELISA and 75.5% for ICT). In Thailand, Cohen et al. identified acute leptospirosis in 67 (9.5%) of febrile subjects using 2 rapid assays, a dipstick, and latex slide agglutination test (19). Patients sought treatment after a mean of 3.4 days of fever and returned 22 days later. Compared with MAT on paired serum specimens, the sensitivity of testing acute-phase serum with the dipstick and latex slide agglutination tests was 22% and 13%, respectively, which is similar to our findings.

Strengths of our study include the rigorous, prospective design, uniquely large sample size, inclusion of an unstudied population believed to be at high risk, an unusually high rate of follow-up, and clinical correlation. To minimize selection bias, we used standardized criteria to sequentially enroll a large cohort (≈900 patients) with thorough follow-up to enable assessment of acute-phase IgM testing versus clinical impression and relevant epidemiologic and clinical features. Those patients from whom paired serum specimens were not available differed only slightly from the included population. We excluded the few with equivocal results to avoid possible misclassification with resultant potential failure to identify significant predictive clinical features. By rigorously distinguishing acute from recent leptospirosis, we were able to confirm that myalgias and arthralgia were frequent symptoms and that the presence of conjunctivitis or conjunctival suffusion is diagnostically helpful. Leptospirosis in this cohort was relatively mild, as evidenced by stable vital signs, the absence of jaundice, and complete blood counts within nearly normal ranges. The sparse laboratory data reflect standard clinical practice in which automated testing is largely unavailable in the public sector, expensive in the private sector, and thus infrequently obtained.

Our results might have differed if we had used a different diagnostic standard, but culture, for example, is insensitive, labor intensive, unlikely to be available soon at Karapitya Teaching Hospital and many similar hospitals, and too slow to guide clinical management. We may have misclassified the number of days with fever, since temperatures are infrequently taken at home or in the hospital; however, duration of fever correlated well with that of symptoms, and no systematic bias would be expected, since etiologic diagnoses were not known when patients sought treatment at the hospital. Our estimate of leptospirosis may be low if a wider array of serovars is circulating in southern Sri Lanka than were detected by the ELISA used; however, no other available commercial assay would have been expected to be more sensitive.

We conclude that leptospirosis causes substantive illness in southern Sri Lanka. Furthermore, we found that testing acute-phase serum specimens for IgM has multiple limitations for the diagnosis of acute leptospirosis, because a positive result more often denoted past infection than an acute infection, and results were negative early in infection. Clinical impression is comparatively better without added cost (20). Paired serum specimens can provide rigorous diagnosis, but patients and clinicians need rapid diagnosis to guide clinical management. A few antigen-based or nucleic acid–based rapid tests have been described, but prospective clinical validations are limited (17,21–25). Rapid, pathogen-based tests for early diagnosis need to be developed.