Epidemic dengue transmission in southern Sumatra, Indonesia

TRANSACTIONSOFTHEROYALSOCIETYOFTROPICALMEDICINEANDHYGIENE(2001)95,257-265

Epidemic

dengue

transmission

in southern

Sumatra,

Indonesia

Andrew Lee Corwin’, Ria Purwita Larasati’, Michael J. Bangs’, Suharyono Wuryadi*, Sumarjati Arjoso*, Nono Sukri’, Erlin Listyaningsih’, Sri Hartati’, Rozali Namursa3, Zarkasih Anwar4, Surya Chandra4, ’ US Naval Medical Research Unit No. Benny Loho’, Holani Ahmad6, James R. Campbell’ and Kevin R. Porter’ 2 (US NAMRU-2), a WHO-SEAR0 Collaborating Centre for New and Emerging Diseases, Jakarta, Indonesia; ’ Virology Department, National Institute of Research and Development, Indonesian Ministry of Health, Jakarta, Indonesia; ‘Provincial Health Authority (KANWIL DEPKES DTI), Palembang, Indonesia; 4M. Hoesin Hospital, Palembang, Indonesia; ‘Charitas Hospital, Palembang, Indonesia; 6Directorate General of Communicable Disease Control and Environmental Health (PZMPLP), Ministry of Health, Jakarta, Indonesia Abstract An outbreak of dengue fever (DF), dengue haemorrhagic fever (DHF), and dengue shock syndrome (DSS) in the city of Palembang, south Sumatra, Indonesia was investigated to (i) validate epidemic occurrence, (ii) confirm dengue virus aetiology and associated serotype(s), (iii) provide a demonstrable measure of community impact, and (iv) identify causative relationship (if any) with climatic El Nixio Southern Oscillation (ENSO) influences. Trend analysis based on a 6-year retrospective review of hospital records demonstrates a 3-fold increase in clinical cases for the outbreak period (January-April 1998), relative to historical records. In the 2 hospitals surveyed, the monthly mean number of outbreak-related dengue cases over 4 months was 833 (range 650-995 cases/month); the mean monthly value for the previous 72 months was 107 (range 14-779 cases/month). An apparent trend in epidemic transmission was observed, evolving from a 5-year cyclic phenomenon to an annual occurrence, often indistinguishable from one year to the next. The proportional distribution of clinical outbreak cases into DF, DHF and DSS diagnostic categories was 24%, 66%, and lo%, respectively. The population aged lo-19 years accounted for the largest (35%) proportion ofhospitalized DHF cases, followed by children aged 5-9 years (25%) and children aged 4 years (16%). Serum samples obtained during acute illness from 22 1 hospitalized patients were examined using serology, RT-PCR, and virus isolation in cell culture: 59% of samples had laboratory evidence of a dengue infection. All 4 dengue virus serotypes (DEN l-4) were identified in epidemic circulation, with DEN 3 predominating (43%). DEN 1 was the principal serotype associated with less severe dengue illness, suggesting that virulence may be, in part, a function of infecting serotype. The climatic influence of ENS0 on rainfall and temperature in the months leading up to and during the outbreak was dramatic, and is likely to contribute to favourable outbreak conditions. Keywords: dengue fever, dengue haemorrhagic fever, dengue shock syndrome, epidemic, time trends, age-groups, serotypes, climate, Indonesia Introduction Epidemic dengue fever (DF) has been described throughout South-East Asia, and was documented as earlyas1902inMalaysia (GEoRGE&LAM,I~~~).M~~~ recently, epidemic dengue haemorrhagic fever (DHF) has been recognized in the region: first reported from the Philippines in 1955, Thailand in 1958, and in Indonesia during outbreaks on the island of Java in 1968 (UNGCHUSAK & KUNASOL, 1987; NATHIN et al., 1988). Multiple serotypes of dengue viruses are endemic in most of tropical Asia, including Indonesia (BENENSON, 1995).- In Jakarta (Indone&), all 4 dengue serotypes were isolated from DHF oatients examined from 1975 to 1978 (SUHARYONO et al., 1979). A recent urban outbreak of DFiDHF in eastern Indonesia was attributed to circulating dengue viruses types 1,2, and 3 (RICHARDS et al., 1997). Children appear less susceptible to severe DF compared with adults; however, the risk of DHF is significantly greater for children, with hospitalizations and fatalities 3 times greater than for adults. With appropriate supportive- therapy, case-fatality rates (CFRs) are generallv < 2%. In Indonesia. CFRs for bHF Lave steadily d&lined with improved ilinical care from41% in 1968to 3%in 1983 (NATHIN etal., 1988; BENENSON, 1995). Outbreak trends of DFiDHF are generally characterized as cyclic, outbreak occurrence being a multi-function of seasonality, i.e., rainy versus dry season influencing vector density and competence, availability of susceptible (immunologically) naive populations, and circulating viral strains (BENENS~N, 1995; EPSTEIN et al.. 1995; FOCKS et al.. 1995; HALES et al.,’ 1996). In Inddnesia,*peak incidence of CWDHF cases generally occurs during the months of October through April, usually coinciding with the rainy season Address for correspondence: Captain A. L. Corwin, US NAMRU-2, American Embassy Jakarta, Unit 8 132 NAMRU TMO, FPO Al’ 96520-8132, USA; fax f62 21 424 4507.

(SUHARYONO etal., 1979; GuBLER~~~~., ~~~~;NATHIN et al., 1988; RICHARDS et al., 1997). From the Malaysian experience, the pattern of epidemic DHF has changed from outbreaks every 4 years to an increasing trend of annual occurrence (GEORGE & LAM, 1997). Similarly, epidemic trends of DFiDHF in Indonesia were previously recognized every 5 years; since 1982, that pattern has become somewhat more irregular, with a notable increase in intensity and spread (SUMARMO, 1987; SUROSO,1991). Outbreaks of DF/DHF are generally cyclic in nature, and involve spread from a single epidemic focus. In contrast, multiple epidemic foci are likely to occur against a background of sporadic (continuing), ‘interepidemic’ disease occurrence; outbreaks in endemic a;eas tend to be of a more explosive nature (GUBLER et al.. 1979: DO et al.. 1994: K~o. 1995). Endemic transmission requires critical suscepiible human population size and density measures, often reflective of increased urbanization. This phenomenon has been described in Malaysia, where a dramatic (50%) rise in DHF hospitalizations was attributed to urban migrations, leading to more frequent recrudescence of dengue epidemics (SHEKHAR & HUAT, 1992). Identification ofpossible DF/DHF outbreak scenarios necessitates close monitoring of predisposing factors that may contribute to favourable outbreak conditions, e.g., weather and vector collection data. Most instances of epidemic DF/DHF recognition, however, involve targeted surveillance of hospitalized-case admissions. The large ratio of inapparent to apparent (primary) infections, described as high as 160: 1 in Thailand, results in significant under-reporting of outbreak-related infections from hospital-based surveillance (RUSSELL et al., 1968; BURKE et al., 1988). Also, the generally mild signs and symptoms of dengue can be easily confused with other diseases (upper respiratory illness, influenza, allergy), and therefore preclude early recognition of an outbreak

258

event and may disguise the impact of epidemic dengue in a communitv (DIETZ et aZ.. 1990). Converselv. clinical over-reporting‘of DFiDHF has been documenieh, where only 41% of clinically diagnosed childhood DF/DHF cases in Java, Indonesia, could be confirmed by laboratory testing (CHAIRULFATAH et al., 1995). A reuorted urban outbreak of susnected DFIDHFI dengueshock syndrome (DSS) occu&ed in Palembang, south Sumatra, Indonesia, in 1998. Prompted by Indonesian Ministry of Health initiatives, we investigated this outbreak to (i) validate epidemic occurrence, (ii) confirm dengue virus aetiology and associated serotype(s), (iii) provide a demonstrable measure of community impact, and (iv) identify causative relationship (if any) with climatic El Nifio Southern Oscillation (ENSO) influences. Methods The outbreak Against a background of numerous outbreaks of DHF being reported throughout western Indonesia, a major urban outbreak of susoected DHF was described in local newspaper accounts* from Palembang, Sumatra, in March 1998. Anecdotal reporting based on hospitalcase findings from the affected area suggested significant morbidity and associated disease mortality, beginning in January 1998. Notable was the significant representation of nominally healthv voung adults among the affected population.-An invesiigation was undertaken in April 1998 bv the Indonesian Ministrv of Health. including the Direct&ate General of Communicable Disease CoGtrol and Environmental Health (P2M/PLP), the National Institute of Health Research and Development (LITBANGKES), and US Naval Medical Research Unit No. 2 (NAMRU-2), working with local medical authorities in Palembang. A multiple-design study strategy was used to validate and describe epidemic disease occurrence from an epidemiological perspective using: (i) historical trend analysis, ( ii) hospital-based, clinical case detection, (iii) community-based, cross-sectional study, and (iv) virological diagnostic validation of causal aetiological assumptions. The area. Palembang, located at 3” S, 105” W, is an urban centre in south Sumatra, the most western islandchain in the Indonesian archipelago, with an estimated population of 1333 343 people. The city is situated in a lowland area (O-25 m above sea level) comprising some 4 17 km2, and divided by the Musi river that runs wide and deep through its centre (PALEMBANG MUNICIPALITY ADMINISTRATION, 1998). Fresh water marshlands surround most of the city. Historical trend analysis Historical data were obtained from disease reports inpatient data routinely representing summary (monthly) passed to the provincial health authorities (KANWIL DEPKES DT I) from M. Hoesin General Hospital, and Charitas Hospital, both located in the city of Palembang. Information-was extracted: (i) from the 6: year period prior to the outbreak (7647 DF/DHF/DSS clinical euisodes, Tanuarv 1992-December 1997), (ii) during the actual outbreak period (3331 DFIDHF?DSS clinical episodes, January-April 1998), and (iii) from the post-outbreak period (573 DFIDHFIDSS clinical episodes, May-August 1998), to provide comparative trend analysis of monthly (outbreak versus non-outbreak) episodes of DF/DHF/DSS, by age and sex. The purpose of this study approach was to differentiate outbreak versus non-outbreak occurrence and identify periods of interepidemic activity. Hospital-based, clinical case detection During the January through April 1998 outbreak period 2764 dengue cases were identified, using standardized, clinical investigative selection criteria. Presumptive diagnoses of DF were based primarily on

ANDREW

LEE CORWIN

ETAL.

subjective clinical assessments by hospital physician staff of inpatients presenting principally with fever (a 38”C), with/without intense headache, myalgia, arthralgia, retro-orbital pain, and rash. More severe disease episodes involving haemorrhagic manifestations and/or shock syndrome were clinically diagnosed as DHF or DSS, respectively. This clinical diagnosis acknowledges the fact that in many instances evidence of true plasma leakage, as evidenced by haemoconcentration, was not available owing to early institution of intravenous fluid replacement. Therefore an unknown number of clinically suspected DHF cases could have represented DF with haemorrhage. A total of 2764 inpatient cases of suspected DHFiDSS were recognized as outbreak-related and enrolled for study purposes. Demographic data pertaining to age and sex, place of residence for identifying possible spatial clustering of outbreak episodes, and clinical information descriptive of outbreak case signs and symptoms, were extracted from patient records and/or obtained from personal interviews by the investigative team (using a standardized collection instrument). Community-based, cross-sectional study Four area-specific sampling clusters (a grouping oflike units, e.g., households, villages, districts) were randomly selected for study to provide the following: (i) age and sex representation of the targeted populations for determining age/sex-specific attack rates; and (ii) inapparent-toapparent infection ratios. In addition to the 4 sampled communities (88 households. 478 studv subiects), located in urbanl’alembang, a single rural community (25 households, 127 study subjects) outside the city was selected to compare urban and rural outbreak parameters. Twenty-five households were initially targeted for study inclusion in each selected community, although the actual mean number of study households enrolled per community was 23 (range 13-25), with the family serving as the principal sampling unit to ensure age and sex representation reflective of the larger community. After obtaining informed voluntary consent from study participants, demographic information was collected by trained interviewers from each household member. A responsible family adult served as informant for children aged < 18 years, and adult (2 18 years) household absentees. Overall, the mean age of study participants was 26 years (range l-78 years), with no significant difference (P > 0.05) by community. The mean number of participating family members per household was 5. Laboratory examination A 5-7-mL blood sample was obtained from 221 suspected dengue inpatients (or 8% of the 2764 recognized outbreakcases), selected without bias as to disease severitv. from both M. Hoesin General Hosoital (67 cases) and Charitas Hospital (154 cases) during;he last 7 outbreak weeks, 25 March through 10 May 1998. Additionally, 130 control sera were collected from agesex case-matched (first 130 suspected dengue case sera selected for serological evaluation) hospitalized patients with no clinical evidence suggestive of a dengue viral infection. Specimens were examined for evidence of dengue virus infection using 3 testing methods: (i) virus isolation, (ii) reverse transcriptase-polymerase chain reaction (RT-PCR), and (iii) serology. virus isolation. Serum samples collected from hospitalized cases of suspected dengue virus infection were analysed for the presence of virus as described by SINGH & PAUL (1969) with some modifications. In brief, patient’s serum diluted 1: 10 in Hanks’ balanced salt solution was inoculated on to confluent monolayers of C6/36 cells, a derivative of Aedes elbopictus cells (ATCC Cell Repository Line 1660). Following a l-h incubation at 37°C cell monolayers were washed once in phosphate buffered saline and fresh culture medium was added. Cell cultures were then maintained at 30°C for a total of

EPIDEMIC DENGUE IN INDONESIA

259

14 days. All cultures were monitored daily for the appearance of typical cytopathic effects (CPE). Cultures demonstrating CPE were harvested, the cells applied to immunofluorescence slides and stained with a series of dengue serotype-specific monoclonal antibodies (from a mouse hvbridoma cell line that produced IgG) for virus identification. RT-PCR. Serum aliquots were processed for analysis by RT-PCR. The assay was performed according to the method of LANCIOTIJ et al. (1992). Viral RNA was isolated from serum using a commercially available kit for isolating viral RNA (Qiagen; Qiagen GmbH, Hilden, Germany, following the manufacturer’s instruction). Complementary DNA (cDNA) strands were reversetranscribed using reverse-transcriptase enzyme and dengue-specific oligonucleotide primers (LANCIOTTI et al., 1992). Dengue cDNA was then subjected to nested PCR with the first reaction using universal forward and reverse primers that recognized conserved sequences present in all 4 dengue virus serotvpes. The second nested reaction was performed with dengue serotype-specific forward olirronucleotide mimers. The PCR-nroducts were reso&d by agarose gel electrophoresis and ethidium bromide staining. Serotype identification was determined based on the size/mobility of the PCR-products (LANCIOTTI et al., 1992). Dengue diagnosis by serology. Serum samples collected from hospitalized cases of suspected dengue were tested for the presence of dengue virus antibodies using a microtitre plate IgM/IgG antibody capture (MAC/ GAC) ELISA as described by INNIS et al. (1989). A positive result is obtained when the index value for a sample is 9 40. In the case where IgM is negative, a dengue infection is said to occur when there is a 2-fold rise in acute to convalescent anti-dengue IgG values with an absolute IgG index of 2 100. For the community-based study, a 5-7-mL blood sample was collected from 350 volunteers. Sera were tested for dengue IgM antibodies using a rapid immunochromatographic assay (PanBio Inc., Brisbane, Australia) (VAUGHN et al., 1998). The test is based on an antibody-capture card format using an antigen cocktail made from all 4 dengue virus serotypes. The test was shown to be highly sensitive in diagnosing dengue virus infection in an earlier studv (VAUGHN et al.. 1998). The presence of dengue IgM antibodies in a serum sample by this assay was considered to be indicative of a recent dengue virus infection. Weather data Recorded daily measures of rainfall in millimetres

and

temperatures in degrees Celsius were obtained from the Meteorology and Geophysics Station Talang Betutu, located in the city of Palembang. Statistical analysis x2 tests with Yates’ correction were used to compare multiple proportions from mutually exclusive sample groups. The proportional hypothesis test was used to assess differences between 2 proportions from one group (overlapping categories from a single sample). The 95% confidence intervals (CI) were calculated by the exact and normal methods. Student’s t test and analysis of variance were used to determine the significance of differences between mean values. Results Historical trend analysis Figure 1 shows significant temporal clustering of hospitalized dengue cases during the months of January through April 1998, signifying a clear epidemic phenomenon in contrast to the comparatively low level of endemic disease occurrence in the 6 years prior to the outbreak. A greater than 3-fold rise in number of cases was observed during the outbreak period as compared with the previous 72 months, except for 2 smaller outbreaks recognized in retrospective data analysis. The overall monthly mean number of cases during the outbreak months (January-April 1998) was significantly higher (P < 0.0001; 95% CI for the difference between means, 3 15 -4 12) than for the 6-year period leading up to the epidemic: 416 cases (range, 54 1 in February to 261 in April 1998) versus 53 cases (range, 3 in April 1993 to 509 in January 1995). This dramatic increase in dengue cases was evident in both study hospitals: the monthly mean number of cases at M. Hoesin General Hospital during the outbreak period was 4 18 (range, 26 1 in April 1998 to 541 cases in February 1998), compared with a mean of 53 cases (range, 3 in April 1993 to 509 in January 1995) forthe 6yearspriortotheoutbreak(P < 0:0001;95% CI for the difference between means 285-447). Similarlv. the mean monthly number of outbreak cases at Charitas Hospital (415; range, 337 cases in January 1998 to 481 in March 1998) was significantly higher (P < 0.0001; 95% CI for the difference between means, 306-416) than for the previous 6 years (54; range, 8 cases in May 1995 to 230 in Tanuarv 1997). Finallv, a ureciuitous fall in total hospitahzed cases was evident during the months immediately following the outbreak in May (337 episodes), June (148 episodes), July (65 episodes), and August (23 episodes). A retrospective monthly review of dengue case records

3000’ g: 2500

1992

1993

1994

1995

1996

1997

1998

Fig. 1. Temporal clustering of clinically recognized dengue fever, dengue haemorrhagic fever, and dengue shock syndrome from Charitas and M. Hoesin Hospitals, Palembang, Sumatra, January 1992-August 1998.

ANDREW LEE CORWIN ETAL. for January-April 1992 through 1998 provides an analytical control of possible influencing outbreak determinants, such as seasonal rainfall. Overall, the monthly (representing 28 months) mean number of recorded cases from study hospitals in the pre-outbreak years (for January, February, March, and April) was 65, whereas during the outbreak period the mean was 416 cases (P < 0.0001; 95% CI for the difference between means, 285-418). Similar observations were noted in both M. Hoesin Hospital [a mean of 60 cases during the 24month pie-outbreak period versus 418 during the 4month outbreak period (P < 0.0001; 95% CI for the difference between means, 24 l-474)] and Charitas Hospital [a mean of 54 cases during the 18-month preoutbreak period versus 415 cases during the 4-month outbreak period (P < 0.0001; 95% CI for the difference between means, 268-423)]. There is some evidence of cyclicity in Palembang associated with epidemic dengue occurrence over time. Sporadic, low-level transmission of dengue fever virus was observed for the first 3 years (1992-94), with a monthly mean of 20 episodes (range, 3-98). In contrast, there were 3 distinct peak outbreak periods, the first in 1995, a second in 1996197, and the third and largest in 1998, all occurring in the January-April time frame, although the 1997 outbreak clearly originated and peaked in the latter half of 1996, marking a deviation in the notion of ‘strict seasonality’. The consistency in age-specific disease trends over time (January 1992-August 1998) is highlighted in Figure 2. Children aged 5- 14 years continuously accounted for the largest proportion of all recognized dengue cases. The proportional representation of this population (aged 514 years), relative to the other age-groups, was particularly pronounced during each of the 3 observed (peak) outbreak periods (1995, 1996197, and 1998) at M.

Hoesin and Charitas Hospitals. Young adults, aged 1524 years, were also well represented in the dengue patient populations from throughout retrospective data review. Hospital-based, clinical case detection Clinical epidemiology. The 3 epidemic curves presented in Figure 3 demonstrate gradual rather than dramatic changes, except for the marked decline during week 18 of the outbreak, regardless of clinically suspected diagnosis, signifying an end of epidemic occurrence. Cases of epidemic DF, DHF, and DSS peaked at weeks 8, 9, and 4 of the outbreak, respectively. Overall, a DHF diagnosis was assigned the largest proportion of clinically identified dengue episodes for the 4-month outbreak period, 76% with and 66% without DSS. Children aged lo- 19 years accounted for the largest proportion of all patients with DF (33% of 660 cases) and DHF (35% of 1772 cases), while children aged 5-9 years accounted for the largest portion of DSS (40% of 253) (data not shown). There was no significant difference in mean ages for DF cases (12.9 & 10.5 years), DHF cases (13.9 ZIZ10.6 years), and DSS (8.7 * 7.5 years). Also, the male-to-female case ratios differed little by clinical diagnosis: 0.96: I for DF, 1: 1 for DHF, and 0.81: 1 for DSS. Lastly, the relative proportions of DF, DHF, and DSS cases varied little from month to month during the outbreak period: 24%, 66%, and 11% in January, 23%, 68%, and 9% inFebruary, 26%, 65%, and 9% in March, and 24%, 65%, and 12% in April, respectively. There was notable evidence of spatial clustering of reported cases within urban Palembang, namely those neighbourhoods within close proximity of M. Hoesin and Charitas hospitals (Fig. 4). Some temporal movement of case clusters, marking the progression of epidemic spread over time was observed. Case occurrence in the more peripheral kelurahans* outlying the areas of

.nn.. M. Hoesin Hospital Age-group 800 1

...

(years)

ET-

-T

Months

1992

1993

1994

1995

1996

1997

1998

7ooFaritas Hospital 600

1992

-1

1993

1994

1995

1996

1997

/

1998

Fig. 2. Temporal age distribution of clinically recognized dengue fever, dengue haemorrhagic fever, and dengne shock syndrome by hospital, Palembang, Sumatra, 1992-98. *Arbitrary divides that separate cities, villages, and rural area into political areas.

EPIDEMIC

DENGUE

IN INDONESIA

261

Dengue fever 70 i---p I

60;

Dengue shock syndrome 30

1 2 3 4

5 6 7 8 9 101112131415161718 Week

Dmgue haemowhagic fever

zoor-- ~-

.- ___

1 2 3 4

5 6 7 8 9101112131415161718

Week

1 2 3 4 5 6 7 8 9 101112131415161718 Week Fig. 3. Epidemic curve of clinically recognized dengue fever, dengue haemorrhagic fever, and dengue shock syndrome from Charitas and M. Hoesin Hospitals, Palembang, Sumatra, January 1992-April 1998.

/ Charitas Hospital

Fig. 4. Spatial analysis of clinically recognized dengue fever, dengue haemorrhagic fever, and dengue shock syndrome around the two studied hospitals in Palembang, Sumatra, January-April 1998. Scale: 1 cm = 2.25 km. principal case concentrations was recognized in February, and in March, although less so. Figure 5 depicts the cumulative, geographical clustering of hospitalized recognized dengue cases as function ofpopulation density. The Keluruhans with the highest attack rates (>450 outbreak dengue cases per 100 000) were more dispersed than the geographical mapping of cases (in the absence of census data) would have indicated. Overall, 100 dengue deaths among 2439 known outcomes (4.1%) translated into a CFR of 41 deaths11000 clinically recognized cases. The CFR at

Charitas Hospital (24/1000 cases) was significantly lower (P < 0,001) than for M. Hoesin Hospital (87/ 1000 cases). Age-specific CFRs in the O-4, 5-9, 10-19, 20-29, and 30-49-years categories were 77 (So/), 60 (6%), 21 (2%), 9 (l%), 6 (0.6%) clinical cases, respectively; no deaths were reported among dengue patients aged 2 50 years. CFRs by clinical diagnosis were 9/1000 (< 1%) for DF, 811000 (< 1%) for DHF, and 32811000 (33%) for DSS. There was no significant decline (P > 0.05) in CFRs over time: 6511000 (7%) in January, 4711000 (5%)

ANDREW LEE CORWIN ETAL.

262

Fig. 5. Attack rates of dengue outbreak cases per 100 000 people in 74 keluruhans,January-April recognized cases matched with February 1998 census data by keluruhan (administrative district).

in February, 39/1000 (4%) in March, and 38llOOO (4%) in April. Laborato y-based epidemiology. Laboratory findings provided evidence of recent dengue infections, using either RT-PCR, virus isolation, or MAC/GAC ELISA, in 59% of 221 case sera sampled: 58% of DF, 60% of DHF, and 48% of DSS patients. The percentage of control (afebrile) subjects with laboratory-determined dengue (inapparent) infections (20%) was significantly lower (P < 0.0001) than for cases. Table 1 shows that of 153 case specimens: (i) evaluated by RT-PCR, tissueculture technique, and ELISA IgM capture assay, and (ii) with laboratory evidence of recent infections by RTPCR and/or tissue-culture technique and/or ELISA IgM capture assay,

-

1 (100%) 1 (100%)

“Based on tissue culture, mosquito inoculation, and RT-PCR. bMixed infections reflect RT-PCR testing only; confirmatory evaluation pending.

1 (loo%> -

EPIDEMIC

DENGUE

IN INDONESIA

263

of study participants had dengue IgM antibodies, in the absence of clinical disease. The percentage of subjects surveyed from the case communities with u recognized dengue illness and IgM antibodies was 2% (95% CI for proportion, 0.00458-0.0323), whereas 12% (95% CI for proportion, 0*0889-0.157) of the study population with IgM detectable antibodies reported no outbreak-related dengue illness. Age-specific ARs (clinically recognized) in the 4 urban communities, presented in Table 3, ranged from 0.05) in ARs for males (3%) and females (4%). There was little evidence of familial clustering of dengue illness, although the limited number of actual cases (17) precluded conclusive study findings. Overall, only 13% ofhouseholds from the 4 communities surveyed reported at least a single family member with outbreak-related dengue, of which 82%, 9%, and 9% of households had 1, 2, and 6 affected persons, respectively. Thirty-five percent of households had 1 member with IgM dengue antibody, accounting for 3 1 infections; most reflected single household infections (7 1X) and a smaller proportion with 2 (26%) and 3 (3%) antibody-positive family members.

Weather data Figure 6 clearly highlights the El Niiio Southern Oscillation (ENSO) phenomenon as translated into subnormal rainfall in the months leading up to the outbreak (cumulative rainfall of 132 mm for July-November 1997), followed by relatively heavy rainfall in December 1997 (409 mm) and throughout most of the January-April (1998) outbreak period (cumulative rainfall of 1109 mm). Yet, a comparative review of mean rainfall for the months of July-November 1992 (163mm), 1993 (199mm), 1994 (79mm), 1995 (175 mm), and 1996 (200mm), shows significantly higher measures of rainfall (P < 0.0001) than for the same period in 1997 (26 mm; range, < 1- 124). Temperature readings (“C) from April 1997 through April 1998 were highest for the months of October and November, just prior to the outbreak, and in May 1998 (Fig. 6). Moreover, the mean temperature of 27.7”C for the months of October-December 1997 was higher compared with the same months (October-December) in 1992, 1993, 1994, 1995, and 1996: 26+5,26*8, 26.8, 26.9, and 26.7”C, respectively. During the outbreak months of January-April 1998, the mean temperature of 27.6”C was also relatively higher than for the preceding 6 years: 26.7, 26.3, 26.5, 26.3, 26.2 and 26.5”C, respectively. The differences in mean temperatures in the months leading up to and during the outbreak, compared with the monthly means for 1992-96, were most pronounced during the actual outbreak months of January, February, and March (Fig. 6). Discussion Analysis of trend over time illustrates the epidemic nature of dengue transmission in Palembang, Sumatra,

Table 3. Attack rates of inapparent and apparent Sumatra, in the community-based study

dengue

cases from

Disease”

four

No. surveyed

No. cases

20 48 108

:, 4 2 1 it

2: 61 53

areas,

% 5 13 4 2 1 3

No. tested

No.

252

:

55 63 45 36 41

3 1 0 0 0

Without disease %

No.

;

0 2 8 9 3 5 7

With & without disease %

No.

%

9 15 14 7 14 17

0 2 11 10 3 5 7

9 20 16 7 14 17

“All persons surveyed, regardless of whether a blood sample was collected for testing purposes. bAll persons surveyed from whom a blood sample was collected for testing purposes.

25.

0 -c

Fig.

6.

Palembang,

IgM positiveb With disease

Age-group (years) o-4 5-9 10-19 20-29 30-39 40-49 3 50

urban

mean C”mYlarl”e monrldy ramfall 1992-96~ mean nmxbly remperarure 1992-96 A

Comparison of monthly temperature and rainfall measures in

mean cumukmve monthly rainfall 1997-98 mean monrbly temperature 1997-98

1992-96

and 1997-98,

Palrmbang,

Sumatra.

ANDREW LEE CORWIN ETAL.

264

in January through April 1998. The steep rise and fall of hospitalized dengue cases, observed from both study hospitals, was greater than for any other period over the 7-year historical review. Evident during this analysis is the apparent annualization of outbreak occurrence, beginning in 1995, in that epidemic recognition is observed changing from a periodic (every 3-5 years) to an almost yearly phenomenon. In contrast, historical data from Java, Indonesia (1968-90), showed epidemic cyclicity to -5-year intervals: e.g. 1973, 1977, 1983, and 1988 (SUROSO, 199 1). In Thailand, DHF epidemic activity was first characterized by a 2-year cycle that has also become more irregular and frequent, suggesting increased endemicity (WHO, 1986). Impressive is the steady State trend of proportional age representation among dengue cases over the 7 years leading up to and during the Palembang outbreak, in particular, the disproportionate increase of dengue patients aged 5- 14 years, followed by young adults aged 15-24 years, relative to the other age-groups, specifically during the peak outbreak periods. Notably, the relatively large percentage of dengue patients aged 5-14 years (49%) greatly exceeded their proportional representation in the community (24%). The epidemic curves specific for this outbreak probably reflect significantly increased exposure (mosquito-biting) opportunities for a large proportion of the population living in and around Palembang. The rapid temporal progress of the epidemic, as observed in the number of clinical DF cases seen early on during the outbreak period, suggests a large pool of immunological susceptibles within the general population at risk. A similar outbreak evolution reported from the Americas represents a resurgence of disease after an absence of m&y years (MOURNS, 1982; HAYES & GUBLER, 1992; MORRISON. 19981. In contrast. the 1988 Palembang epidemic occurred in the histoiical context of repeaLf outbreak episodes, against a background of high dengue endemicity. The overwhelming representation of clinically recognized DHF (with and without DSS1 in outbreak-related diagnoses is‘consistent with the DI? to DHF (with and witGout DSS) ratio mix reported in other CHF outbreaks (MORENS. 1982: HAYES & GUBLER. 1992: GUBLER‘ & TREN?, 1995). The literature suggests in: creasing virulence over the course of epidemic transmission, as reflected in a proportional rise in clinical episodes involving haemorrhagic manifestations and/or shock syndrome, and case-fatalities (KOURI et al., 1989; KUNO. 1995: HOLMES. 19981. There was no evidence of a progress&e virulence trend in the Palembang outbreak; the relative proportions of DF to DHF to DSS cases remained similar for each of the 4 months of the epidemic. Indeed, CFRs actually declined over time during the outbreak period. Laboratory testing of samples obtained from hospitalized cases was done using 3 diagnostic methodologies. Seldom were samples positive by all 3 methods. RTPCR, recognized as the most sensitive of the testing methods in that identification is predicated on detection of RNA from the virus, accounted for the largest proportion of dengue confirmed positives (39%). However, RTPCR together with serology proved to be an effective combination of diagnostic assays, detecting 98% of dengue cases. These findings clearly show the need for use of complementary testing strategies in substantiating and describing epidemic dengue occurrence. All 4 dengue serotypes contributed to the epidemic, with DEN-3 accounting for the largest proportion of isolates. Outbreaks of dengue illness throughout Indonesia have been attributed primarily to DEN-3 (SUHARYONO et al., 1979; GUBLER et al., 1981; RICHARDS et al., 1997). The recognized predominance of DEN-3 in dengue outbreaks, however, may have been a function of inherent sampling bias, in that specimen collection is usually restricted as a matter of convenience to hospita-

lized and, by nature, more severe cases (SAMSI et uZ., 1990; HOLMES, 1998). However, study findings from the 1998 Palembane outbreak lend credence to the notion that ‘some &al serotypes or genotypes more often lead to DHF/DSS than others’, suggestive that virulence indices associated with dengue epidemic transmission may be (in part) serotype or genotype specific (SAMSI et aZ., 1990). Analvsis of the Palembanp: data proved DEN-1 principally- responsible for less-severe dengue illness, while more-severe disease (DHF and DSS) was attributed to DEN-3, and to a lesser extent, DEN-2. In Takarta. DEN-3 was shown as the nredominant virus circulating in epidemics, particulirly those involving more-severe clinical manifestations (STREATFIELD et aZ., 1993). Similarly, severe disease (and associated case fatalities) was linked to DEN-3 in dengue outbreaks in Malaysia, while DEN-l has been synonymous with mild illness and low CFRs (GEORGE &LAM, 1997). The outbreak period of January through April 1998 occurred during the annual rainy season. Historically (previous 6 years), the preponderance of DF/DHF/DSS cases has also been observed during the same months. Similarly, the occurrence of DHF in Indonesia in 198 l85 coincided with the normally wetter months of October through April (NATHIN kt uZ., 1988; SUMARMO, 1987). The notable imnact of ENS0 on climatic indices in the months leading up to and during the outbreak was reflected in a marked increase in rainfall and sustained higher than normal temperatures. It appears reasonable to speculate that this phenomenon contributed to, in some part, favourable outbreak conditions. The influence of warming on the dengue virus and vector Ae. uegypti is well documented. Frenzied mosquito biting behaviour, from higher temperatures in the months preceding and during the actual outbreak, probably added to the risk of exposure to infectious Aedes populations. Finally, the rapid accumulation of rainfall after drought-like conditions enhanced larval breeding opportunities (WA-M’S et al., 1987; EPSTEIN et al., 1995; KUNO, 1995; HALES et uZ., 1996). In conclusion, the Palembang outbreak, in the context of a historical perspective, suggests that epidemic dengue transmission in westernmost Indonesia (south Sumatra) has evolved from a 3-5-year cyclic phenomenon to an annual occurrence. Notable was the relatively large proportion of adolescents and young adults among DHFiDSS cases, particularly during periods of heightened epidemic transmission. While DEN-3 was found predominant, less-severe DF illness was principally attributed to DEN-l, suggestive of a virulence component in association with circulating serotypes. Finally, the impact of ENS0 on weather patterns in the months leading up to and during the oitbreak probably contributed to favourable outbreak conditions. The Palembang outbreak reflects the significant impact of DFI DHFiDSS in an urban community setting in tropical South-East Asia. Acknowledgement

We express our appreciation to Dr Krisna from Paediatric Department, M. Hoesin General Hospital, Palembang, who assisted in the coordination of specimen collection, Mr John Master and Dr Masri from LITBANGKES, who contributed in the conduct of both field and clinical investigative efforts. Financial support: This study is under ST0 LB and DoDGlobal Emerging Infectious System (GEIS) funding. Disclaimer: The views of the authors expressed herein do not purport to reflect those of the US Navy, the US Department of Defense or the Indonesian Ministry of Health. The researchprotocol employinghuman subjects in this study has been reviewed and aDmoved bv the US Naval Medical Research Institute’s Corn&tee for ;he Protection of Human Subjects. References

Benenson, A. S. (1995).

Control of Communicable Diseases

EPIDEMIC

DENGUE

265

IN INDONESIA

Manual, 16th edn. Washington, DC: American Public Health Association. Burke,D. S.,Nisalak,A. & Johnson,D. E. (1988).Aprospective study of dengue infection in Bangkok. Amerian Journal of Tropical Medicine and Hygiene, 38, 172- 180. Chairulfatah, A., Setiabudi, D., Ridad, A. & Colenbunders, R. (1995). Clinical manifestations of dengue haemorrhagic fever in children in Bandung, Indonesia. Annals of the Belgian Society of Tropical Medicine, 75, 291-295. Dietz, V., Gubler, D. J., Rigau-Perez, J. G., Pinheiro, F., Schatzmayr, H. G., Bailey, R. & Gunn, R. A. (1990). Epidemic dengue 1 in Brazil, 1986: evaluation of a clinically based dengue surveillance system. American Journal of Epidemiology, 131, 693-701. Do, Q. H., Vu, T. & Huynh, T. (1994). Dengue haemorrhagic fever in the south of Vietnam during 1975-1992 and its control strategy. Tropical Medicine (Nagasaki), 36, 187-201. Epstein, I’. R., Calix, P. 0. & Blanco, R. J. (1995). Climate and disease in Colombia. Lancet, 346, 1243-1244. Focks, D. A., Daniels, E., Haile, D. G. & Keesling, J. E. (1995). A simulation model of the epidemiology of urban dengue fever: literature analvsis, model develoament. ureliminarv validation, and samples of simulation results: American Journal of Tropical Medicine and Hygiene, 53, 489-506. George, R. & Lam, S. K. (1997). Dengue virus infection-the Malavsian exoerience. Annals of the Academv_ of, Medicine. Singapore, 26,-815-819. ” Gubler, D. J. & Trent, D. W. (1993). Emergence of epidemic dengue/dengue hemorrhagic fever as a public health problem in the Americas. Infectious Agents and Disease, 2, 383-393. Gubler, D. J., Suharyono, W., Lubis, I., Eram, S. & Saroso, J. S. (1979). Epidemic dengue hemorrhagic fever in rural Indonesia. I. V&ological and epidemiological studies. American Journal of Tropical Medicine and H>l+ne, 28, 70 1- 7 10. Gubler, D. J., Suharyono, W., Lubis, I., Eram, S. & Gunarso, S. (1981). Epidemic dengue 3 in central Java, associated with low viremia in man. Ameri~anJournal of Tropical Medicine and Hveiene. 30. 1094-1099. Hales: S., Weinstein, I’. & Woodward, A. (1996). Dengue fever epidemics in the South Pacific: driven by El Nino Southern Oscillation? Lancet, 348, 1664-1665. Hayes, E. B. & Gubler, D. J. (1992). Dengue and dengue hemorrhagic fever. Pediatric Infectious Disease 7ouma1, 11, 311-317.Holmes, E. C. (1998). Molecular epidemiology of dengue virus-the time for big science. Tropical Medicine and hternational Health, 3, 855-856. Innis. B. L.. Nisalak, A.. Nimmannitva, S.. Kusalerdchariva, S.. Chongswasdi, V.,*Suntayakom, S.; Put&ri, P. & Hake,-C: H: (1989). An enzyme-linked immunosorbent assay to characterize dengue infections where dengue and Japanese encephalitis co-circulate. AmericanJournal of TropicalMedicine and Hygiene, 40, 418-427. Kouri, G. I’., Guzman, M. G., Bravo, J. R. & Triana, C. (1989). Dengue haemorrhagic feveridengue shock syndrome: lessons from the Cuban epidemic, 198 1. Bulletin of the World Health Organization, 67, 375-380. Kuno, G. (1995). Review of the factors modulating dengue transmission. Epidemiologii Reviews, 17, 321-334. Lanciotti, R. S., Calisher, C. H., Gubler, D. J., Chang, G. J. & Vorndam, A. V. (1992). Rapid detection and typing of dengue viruses from clinical samples by using reverse transcriptasepolymerase chain reaction. JournalojClinicalMicrobiologv, 30. 545-551.

Morens, D. M. (1982). Dengue fever and dengue shock svndrome. Hosbital Practice, 1, 103-l 13. Morrison, A. C., -Getis, A., Santiago, M., Rigau-Perez, J. G. & Reiter, I’. (1998). Exploratory space-time analysis of reported dengue cases during an outbreak in Florida, Puerto Rico, 199 I- 1992. American 7oumal of Tronical Medicine and HYg&e, 58,287-298. . Nathin, M. A., Harun, S. R. & Sumarmo (1988). Dengue hemorrhagic fever and Japanese B encephalitis in Indonesia. Southeast AsianJournal of Tropical Medicine and Public Health, 19,475-481.

Palembang Municipality Administration (1998). Development of dengue hemorrhagic fever disease in Palembang municipality. Palembang, Indonesia: Palembang Municipality Administration, Report 2. Richards, A. L., Bagus, R., Baso, S. M., Follows, G. A., Tan, R., Graham, R. R., Sandjaja, B., Corwin, A. L. & Punjabi, N. (1997). The first reported outbreak of dengue hemorrhagic fever in Irian Jaya, Indonesia. American Journal of Tropical Medicine and Hygiene, 51, 49-55. Russell, P. K., Yuill, T. M., Nisalak, A., Udomsakdi, S., Gould, D. I. & Winter. P. E. 11968). An insular outbreak of denaue hemorrhagic fever. II.‘Virolbgic and serologic studies. American Journal of Tropical Medicine and Hygiene, 17, 600-608. Samsi, T. K., Wulur, H., Sugianto, D., Bartz, C. R., Tan, R. & Sie, A. (1990). Some clinical and epidemiological observations on virologically confirmed dengue haemomhagic fever. Pediatric Indonesia, 30. 293-303. Singh, K. R. & Pat&S. D. (1969). Isolation of dengue viruses in Aedes albopictus cell cultures. Bulletin of the World Health Organization, 40, 982-983. Shekhar, K. 8: Huat. 0. L. (1992). Euidemioloav of denauei dengue haemorrhagic fever in Malaysia-a retrospectiveuepidemiological study 1973-1987. Part I. Dengue haemorrhagic fever (DHF). Asia-Paczji’c~ournal ofPublic Health, 6, 15-25. Streattield, R., Bielby, G. & Sinclair, D. (1993). A primary dengue 2 epidemic with spontaneous haemorrhagic manifestations. Lance& 342, 560-56 1. Suharyono, W., Gubler, D. J., Lubis, I.,Tan, R.,Abidin, M., Sie, A. & Saroso, J. S. (1979). Dengue virus isolation in Indonesia, 1975-1978. AsianJournal of Infectious Diseases, 3, 27-32. Sumarmo (1987). Dengue haemorrhagic fever in Indonesia. Southeast AsianJournal of Tropical Medicine and Public Health, 18,269-274.

Suroso T. (1991). Epidemiologi dengue. Medika, 10, 9-13. Ungchusak, K. & Kunasol, P. (1988). Dengue hemorrhagic fever in Thailand, 1987. Southeast Asian Journal of Tropical Medicine and Public Health, 19, 487-490. Vaughn, D. W., Nisalak, A., Kalayanarooj, S., Solomon, T., Dung, N. M., Cuzzubbo, A. & Devine, P. L. (1998). Evaluation of a rapid immunochromatographic test for diagnosis of dengue virus infection. 3ounal oj Clinical Microbiolopv. 36.234-238. Wa&, D: M., Burke, D. S., Harrison, B. A., Whitmire, R. E. & Nisalak, A. (1987). Effect of temperature on the vector efficiency ofAedes kegvptifor dengue s?virus. AmericanJournal of Tropical Medicine and Hygiene, 36, 143- 152. WHO (1986). Dengue Haemorrhagic Fever: Diagnosis, Treatment and Control. Geneva, Switzerland: World Health Organization. Received 11 May 2000, revised 4 October 2000, accepted for publication 4 October 2000