Rickettsioses as Major Etiologies of Unrecognized Acute Febrile Illness, Sabah, East Malaysia

Because acute rickettsioses are common, underrecognized, and untreated etiologies of these illnesses, empirical doxycycline treatment should be considered.

R ickettsioses and related infections, including scrub typhus, caused by Orientia tsutsugamushi; spotted fever group rickettsioses (SFGR), often called tick typhus; and typhus group rickettsioses (TGR), also called murine typhus, are reemerging and neglected causes of acute febrile illness (AFI) in Southeast Asia (1)(2)(3)(4). Serologic confirmation of acute rickettsioses, however, requires demonstration of a 4-fold rise in IgG titer by indirect fluorescent antibody (IFA) because acute-phase IgG might represent prior infection and IgM might indicate recent or cross-reactive antibody (1). Furthermore, serologic testing in the early phase of illness is not helpful because antirickettsial antibodies are often not present within the first 2 weeks.
Rickettsioses have been reported in Malaysia historically (2)(3)(4). Malaysia state tertiary hospital data for the period 1994-1999 identified antirickettsial IgG or IgM in 6,442 (10.6%) tested samples (IgG or IgM titer >400 by the indirect immunoperoxidase test), including 4.9% O. tsutsugamushi, 3.1% TGR, and 2.6% SFGR (5). Although rickettsioses are notifiable diseases according to Malaysia Ministry of Health guidelines (6), routine testing is not conducted, paired acute-phase and convalescent-phase serum samples are rarely available, and only 53 infections were reported nationally during 2009-2015 (7). In the East Malaysia state of Sabah, 2 cases were reported over that period despite a seroprevalence of 5.5% among 11,037 patients tested in previous years at the state tertiary referral hospital (5) and high (>50%) seropositivity reported in febrile patients from certain rural areas (8). We prospectively enrolled patients and performed reference standard diagnostic testing to document and describe the epidemiology and clinical features of confirmed rickettsioses among patients with nonmalarial AFI in Sabah, East Malaysia.

Study Sites and Referral System
We conducted the study during December 12, 2013-July 15, 2015, in 2 adjacent district referral hospitals, Kudat and Kota Marudu, in northwest Sabah, Malaysia, on the island of Borneo. These districts cover an area of 3,204 km 2 and had an estimated population in 2015 of ≈169,000 (9). According to Malaysia Ministry of Health divisional health facility structures, each district has a single central referral hospital and subdistrict health clinics. The climate is tropical, has no defined dry season, and experiences increased rainfall during November-March. The region consists of both coastal and inland areas with elevations up to 1,000 m and extensive recent human land-use change from intact forest to oil palm plots and plantations (10).

Study Population
Patients >4 weeks of age with a documented temperature of >38°C within 48 hours of admission to the medical or pediatric wards and a negative malarial blood film were eligible. We obtained consent from the patient or a parent or guardian if the patient was <18 years of age; we also obtained assent from persons 7-17 years of age.

Study Procedures
Trained research nurses used standardized case record forms to collect structured epidemiologic, clinical, and laboratory data, including the admitting clinician's findings on examination, provisional diagnosis, and management. At enrollment, venous blood was obtained for routine and research-related testing after consent. Patients were requested to return for routine hospital outpatient follow-up and convalescent-phase serum sampling 14 days after enrollment.

Laboratory Procedures
Hematologic and biochemical tests, blood cultures, and malaria smears were performed in the hospital laboratory. Anemia was classified according to World Health Organization age-and sex-based criteria (11). Acute kidney injury was defined per KDI-GO criteria (12).
We conducted IFA testing at the Uniformed Services University for the Health Sciences (Bethesda, MD, USA) to confirm rickettsial infections; 2 experts read each slide, as described (13,14). We screened convalescent-phase serum samples using IFA at 1:80 dilution for IgG to O. tsutsugamushi (Karp strain), SFGR (Rickettsia conorii Malish 7 strain), and TGR (R. typhi); for reactive samples, we tested acute-and convalescent-phase paired serum samples together and titrated positives to 2,560. To evaluate the utility of acute-phase IgM for identifying paired IgGconfirmed acute infections, we tested a subset with paired IgG results for IgM in addition to those without convalescent-phase serum. We screened acutephase serum samples for IgM at 1:40 and titrated positives to 1,280.
We performed multiplex real-time PCR to detect O. tsutsugamushi (56-kDA major outer membrane protein gene), SFGR (consensus ompA sequence), and TGR (R. typhi 17-kDa lipoprotein precursor gene) on ED-TA-anticoagulated blood or buffy coat stored at -70°C as described (15). We tested 5 µL of DNA extracted from a 200 µL sample of EDTA-anticoagulated whole blood or buffy coat; we required 2 out of 2 or 3 replicates to confirm a positive result by PCR.

Case Definitions
We used stringent criteria to define rickettsial infections (16), as reported previously (13,14). We defined confirmed rickettsioses as any IgG titer >160; confirmed past rickettsial infection as stable or declining IgG titer >160; confirmed acute rickettsial infection as a >4-fold rise in IgG titer by paired IFA to convalescent-phase titer of >160, positive PCR result, or both; confirmed probable acute rickettsial infection as a 2-fold rise in IgG titer with convalescent-phase titer >160 or single IgG titer >160 without paired serum samples; and possible rickettsioses as any IgG titer of 80 (acute infection if IgG seroconversion). We distinguished SFGR from TGR when we observed a >2-fold difference in SFGR versus TGR titer; if titers were equal, we categorized the infection as group indeterminate.

Statistical Analysis
We analyzed data by using Stata (StataCorp, https://www.stata.com). We performed 2-group comparisons by using the Student t-test for continuous variables with normal distributions or the Wilcoxon-Mann-Whitney test for skewed distributions. For categoric variables, we used a χ 2 or Fisher exact test. We calculated odds ratios (ORs) and 95% CIs by using the Mantel-Haenszel method.

Ethics
We obtained ethics approval for this study. The study was approved by relevant institutional review boards in Malaysia (National Medical Research Ethics Committee), Australia (Menzies School of Health Research), and the United States (John Hopkins University and Duke University).

Diagnosis by PCR
An additional 11 of 319 patients with acute-phase whole blood or buffy coat available had acute rickettsial infections by PCR. These infections were 1 O. tsutsugamushi, 8 SFGR, and 2 TGR (Table 1).

Clinical and Laboratory Features of Acute Rickettsial Infections
Patients with confirmed acute rickettsial infections (n = 49; median age 39 years) were equally likely to be male and were of similar age (p = 0.32) to those with probable acute rickettsioses (median age 43 years); both groups were slightly older than those with no rickettsial infection (median age 29 years; p = 0.007) ( Table  2). The median age of those with acute infections did not differ significantly (p = 0.69): 36 years (IQR 21-51 years) for O. tsutsugamushi, 49 years (IQR 12-58 years) for SFGR, and 38 years (IQR 27-56 years) for TGR. Eight children (<15 years of age) had confirmed acute rickettsial infections (5 SFGR, 2 O. tsutsugamushi, and 1 TGR), of whom 5 (63%) had their infections confirmed by PCR; the youngest was a 6-month-old infant. The median duration of hospitalization for those with confirmed acute rickettsial infections was 5 days (IQR 3-7 days), which was comparable to those without rickettsioses. No deaths were reported.
Anemia occurred in 24 (51%) patients with acute rickettsioses and hematologic results available, and we noted a hemoglobin level of <80 g/L in 3 adults (Appendix Table 4). Peripheral leukocyte counts, renal function after controlling for age, and acute kidney injury, observed in 6/33 patients (18%), were similar in patients with and without acute rickettsioses.
Doxycycline was administered to 2 (4%) patients with confirmed acute rickettsioses (both O. tsutsugamushi infections), albeit in both cases for a provisional diagnosis of gastroenteritis. None of the 6 rickettsial case-patients with a provisional diagnosis of acute undifferentiated fever were treated with doxycycline.

Epidemiologic Features of Patients with Confirmed Acute or Past Rickettsial Infections
Patients with acute or past rickettsial infections were older than those without (median age 43 vs. 24 years; p<0.001); however, the distribution by sex was similar (57% male among those with past rickettsial infection vs. 52% among those without rickettsial infection). Most patients (67%) reported rural residence (Appendix Table 5). Recently spending time in forest areas was more common (OR 2.1 [IQR 1.0-4.3]; p = 0.037) in patients with acute or past rickettsial infection (14%) than patients with neither (7%), including staying overnight in the forest for those with scrub typhus (11%; p = 0.047). Patients were more likely to report a primary occupation as a rubber tapper (OR 5.9 [IQR 1.9-18.5]; p = 0.002) for those with acute or past rickettsial infections (10%) than for patients without rickettsioses (2%). Of the 13 rubber tappers with seroprevalent rickettsioses, 3 (23%) had confirmed acute scrub typhus and 5 (38%) past scrub typhus. Farmers were also more likely (OR 2.

Discussion
Although rickettsial infections were first documented in Malaysia in 1925 (17), few cases are currently confirmed and reported nationally (18). Paired serum samples are infrequently obtained, and IFA requires expertise; both these factors limit diagnostic efforts.
By confirming acute infection based on a 4-fold rise in IgG titer, PCR positivity, or both, we documented and characterized rickettsioses as an important cause of nonmalarial AFI in Malaysia (Figure 2). We found definitive serologic evidence of acute or past rickettsioses in 36% of patients and confirmed acute infections in 26% of patients. Scrub typhus (O. tsutsugamushi infection) was the most common rickettsiosis, especially among rural residents.
Acute-phase IgM demonstrated insufficient sensitivity for identifying acute rickettsioses. Moreover, acute-phase IgG, which cannot distinguish acute from past rickettsioses, is an imperfect measure of seroprevalence because antibodies decline over time even in patients who are not treated (19). PCR also detected few acute infections compared with paired serologic tests. Rickettsial infections were unsuspected clinically, and nearly all were untreated. Therefore, increased clinical awareness, improved diagnostic tools, and empirical use of doxycycline are essential to reduce illness and death attributable to rickettsioses.
Confirmation of acute rickettsioses as a major cause of AFI in Sabah, Malaysia, is to be expected because rickettsioses have been identified elsewhere 1414 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 26, No. 7, July 2020 in the region. A large multicenter study in Thailand, Vietnam, and Indonesia showed that 6% of 1,578 enrolled patients had serologically confirmed acute rickettsioses (20). Other rigorous clinical and crosssectional studies have demonstrated marked spatial heterogeneity in seroprevalence of rickettsioses (21), including 1% in Thailand (22), 33% in areas bordering Myanmar (23), 7% to 27% in Laos (24)(25)(26), and up to 10% in parts of Indonesia (27,28). Our study suggests that rickettsioses are at least as common (26%) in Malaysia as elsewhere in the region.
Rickettsioses are likely to be reemerging in East Malaysia. Despite the presence of a common vector of scrub typhus, Leptotrombidium (L.) deliense mites, being documented in coastal areas for decades (29), a single statewide study (including Kudat District) in the 1980s reported no acute rickettsioses among 383 febrile patients with paired serum samples tested by IFA (30). Sero-epidemiologic findings from 837 persons at that time demonstrated a low seroprevalence of 0.8% for O. tsutsugamushi, 8.6% for SFGR, and 0% for TGR (30). At the Sabah state tertiary referral hospital during 1994-1999, a total of 11,037 samples from febrile patients sent for acute IgG or IgM testing showed a comparatively low seroprevalence of 2.6% for O. tsutsugamushi, 1.3% for TGR, and 1.6% for SFGR (5). However, the inability to confirm acute rickettsioses, even retrospectively, if paired serum samples are not obtained has constrained broader public health understanding of their epidemiology and disease prevalence among the at-risk population in Sabah.
The epidemiology of rickettsial infections in this study is consistent with previous reports of scrub typhus and spotted fever rickettsioses described in rural areas in Southeast Asia (31); those reports indicated that infections were more commonly found in male adults, in agricultural, plantation, or forestry workers (8,28,32,33), and in persons with lower socio-economic status (21,34). The presence of typhus group rickettsioses in rural residents in our study population, despite generally being reported more frequently in urban populations (31), highlights the need for further studies to better understand the epidemiology of rickettsioses in Malaysia and to design appropriate control measures. Because SFGR and TGR cross-reactivity by IFA is expected, testing for all 3 rickettsioses (O. tsutsugamushi infection, SFGR, and TGR) is important.
Factors explaining a potential increase in rickettsioses in Sabah could in part be related to changes in human land use (10) and the relationship with occupational or travel behavior and vector bionomics (35). The effect of habitat on O. tsutsugamushi seroprevalence has previously been evaluated in peninsular Malaysia; higher seropositivity overall was observed in indigenous adults with exposure to intact forest areas (73%) compared with forest fringe (48%) and village areas (8%) (4). These findings are broadly consistent with the association in our study between persons with acute or past rickettsioses and a history of extended forest exposure. Rubber plantation workers in peninsular Malaysia have been reported to have higher SFGR prevalence (40%), compared with O. tsutsugamushi (8%) and TGR (1%) (36). In contrast, although those identified as rubber tappers in our study had high seroprevalence to rickettsioses overall, most of these infections were attributable to O. tsutsugamushi.
We found that patients with acute rickettsioses had clinical findings similar to those without acute rickettsioses, with the exception of more frequent dizziness and hearing loss and less frequent vomiting. Detection of hearing loss might assist in clinical recognition of rickettsioses and has been reported in up to one third of patients with acute scrub typhus (37) as well as in those with SFGR (38). Eschar occurs most frequently on the anterior chest in women and the groin in men with scrub typhus (39); the relative rarity of eschar and higher frequency of hepatomegaly found in our study was also previously described in Malaysia (2). We observed severe cases of acute kidney injury in a small number of patients; however, acute kidney injury remains clinically important because of the subsequent increased risk for chronic kidney disease (40).
Common laboratory markers, such as abnormalities in leukocyte count, thrombocytopenia, anemia, and raised liver transaminases, were all nonspecific (31). Neutropenia occurred in 1 patient with acute scrub typhus, despite this sign being reported as a potentially useful clinical tool to differentiate scrub typhus from dengue (41). AFI was primarily attributed to dengue by treating clinicians, who initiated dengue fluid management protocols. In other cases, empiric treatment of suspected typhoid or leptospirosis was initiated with cephalosporins or penicillins, which are ineffective against rickettsioses. Treatment with doxycycline is recommended by the Malaysia Ministry of Health (42) because an estimated 6% of untreated scrub typhus cases in Asia are fatal (43); however, we found that rickettsioses were unsuspected and untreated.
Tools to support early diagnosis and local evidence-based clinical algorithms to guide directed and empirical treatment of undifferentiated AFI are needed. Diagnosis of acute rickettsioses remains problematic. Orientia and Rickettsia spp. are obligate intracellular bacteria predominantly targeting endothelial cells (44,45); hence, real-time PCR on peripheral blood is insensitive for the detection of acute rickettsioses (15,19,46). Incomplete follow-up meant the sensitivity of PCR could not be directly measured against gold-standard IFA serologic diagnosis for SFGR and TGR, although low PCR sensitivity of 18% has been reported (19). Longstanding technical difficulties with IFA also include the need for extensively trained readers and expensive equipment. Conventional use of antibody detection against only 3 major O. tsutsugamushi antigen groups and serologic cross-reactivity further limits accurate diagnoses; however, high antigenic variation has been observed across Southeast Asia (47,48). We found that acute-phase IgM was a poor predictor of acute rickettsial infection even in the setting of acute scrub typhus, underscoring the importance of increased clinical awareness, improved diagnostic tools, and empiric treatment of those with epidemiologic or clinical risk factors.
The main limitation of this study was the relatively low proportion (41%) of patients with both acute-phase and convalescent-phase serum samples compared with our previous studies (17,18). For this reason, we also performed PCR and tested for IgM, even though sensitivity of the former and specificity of the latter are limited for identifying acute rickettsioses (19). Our estimates of the seroprevalence of rickettsioses and incidence of confirmed acute rickettsioses are conservative. First, only half of patients returned for convalescent follow-up, yet most confirmed acute rickettsioses were seroconversions rather than based on 4-fold rises in titer alone. Second, over half of patients who did have a convalescent-phase serum sample obtained had it obtained earlier than the recommended >14 days. In experimental human SFGR infection, >3 weeks were required before the geometric mean titer of infected volunteers reached diagnostic IgG titers (49). Third, antibody detection of different antigen groups such as that observed in O. tsutsugamushi was based on cross-reactivity to the single Karp strain. Finally, patients with seroconversions indicative of past infection might be missed by a screening strategy that entails testing convalescentphase serum samples first.
To better estimate the actual prevalence of acute rickettsioses, we additionally delineated patients with confirmed rickettsioses with likely acute infection (patients with rising or single IgG titers of >160) as having probable acute infections. The likely underestimation of acute infections caused by diagnostic uncertainty is supported by the similarity in initial clinical and laboratory findings between confirmed and probable acute rickettsioses. The burden and clinical features of rickettsioses are likely underestimated in our study not only because of underreporting of laboratory-confirmed cases but also because persons with mild illness, nonspecific features, and limited physical or economic ability to travel to hospital and persons living in remote areas might not have sought care. Similarly, persons with severe disease might have sought care at or been transferred to a tertiary hospital before study enrollment or died before care they sought or obtained care. TGR prevalence is also potentially underestimated because our population was mainly rural rather than urban. However, our data are likely generalizable to similar populations in Sabah and potentially neighboring areas.
In conclusion, rickettsioses are prevalent among persons with nonmalarial AFI in Sabah, Malaysia. A comprehensive prospective study of rickettsial infections across Malaysia is warranted to define the totality of regional risk. Improved diagnostic tools are urgently needed; in their absence, increased clinical suspicion and empirical doxycycline might avert disease and death attributable to rickettsioses.