Bulletin of the World Health Organization

National dengue surveillance in Cambodia 1980–2008: epidemiological and virological trends and the impact of vector control

Rekol Huy a, Philippe Buchy b, Anne Conan b, Chantha Ngan a, Sivuth Ong b, Rabia Ali b, Veasna Duong b, Sunnara Yit c, Sophal Ung d, Vantha Te e, Norith Chroeung f, Nguon Chan Pheaktra g, Vithiea Uok h & Sirenda Vong b

a. National Center of Parasitology, Ministry of Health, Phnom Penh, Cambodia.
b. Institut Pasteur–Cambodia, 5 Bld Monivong, POB 983, Phnom Penh, Cambodia.
c. Kantha Bopha Foundation Hospital, Phnom Penh, Cambodia.
d. National Pediatric Hospital, Phnom Penh, Cambodia.
e. Takeo Provincial Hospital, Takeo, Cambodia.
f. Kampong Cham Provincial Hospital, Kampong Cham, Cambodia.
g. Angkor Hospital for Children, Siem Reap, Cambodia.
h. Battambang Provincial Hospital, Battambang, Cambodia.

Correspondence to Sirenda Vong (e-mail: svong@pasteur-kh.org).

(Submitted: 09 November 2009 – Revised version received: 05 January 2010 – Accepted: 07 January 2010 – Published online: 07 April 2010.)

Bulletin of the World Health Organization 2010;88:650-657. doi: 10.2471/BLT.09.073908


Over the past 30 years, dengue fever has emerged as the most important arthropod-borne viral disease of humans worldwide and is a major global public health problem, primarily in the tropics.1 Infection with one of the four serotypes of the dengue virus often produces a self-limited but painful febrile illness. The illness may be asymptomatic or can involve severe manifestations such as dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS), which may rapidly progress to death, particularly in children. To date, no drugs can cure the disease and no vaccine can prevent it. Dengue control and prevention have mainly relied on vector control and community action.

Dengue is considered endemic in Cambodia, a country with poor health and economic indicators.2 The estimated population was 14.6 million in 2008.3 The dengue virus was first detected in Cambodia in 19634 and dengue fever has been reported through passive surveillance since 1980. Surveillance was enhanced in 2000 to include laboratory diagnosis for a sample of patients with suspected dengue and, in 2001, with the introduction of active sentinel surveillance.

This report summarizes surveillance data on dengue collected in Cambodia since 1980. Epidemiological trends were determined primarily using data from recent years. In addition, the impact of a 7-year vector control programme on the incidence of the disease was also evaluated.


Cambodia has a tropical climate, with a rainy season occurring between May and November. Rainfall typically peaks between May and June. Some 80% of the population lives in the southern and north-western parts of the country, which together contain 24 provinces and 185 districts.

National surveillance

National surveillance of dengue was established in 1980 and involved passive reporting of clinically diagnosed cases by public-sector health centres and hospitals. In 2000, virological surveillance was introduced at five hospitals, as described below. Subsequently, in 2001, the system changed dramatically when the National Dengue Control Program (NDCP) implemented sentinel surveillance based on three public hospitals and three non-profit-making private hospitals in four provinces. Cases reported through the sentinel system include those among children in either paediatric hospitals or paediatric wards in sentinel hospitals. Thus, national data collected since 2001 were obtained by both passive and active reporting of cases.

Laboratory testing

Virological and serological surveillance is carried out at three of the public hospitals that serve as sentinel sites, a non-profit making private hospital in Siem Reap and an additional public provincial hospital. Paired serum specimens are collected on admission and at discharge from hospitalized patients with clinically diagnosed dengue. The specimens are centrifuged and sent weekly in liquid nitrogen to the Institut Pasteur–Cambodia for serological, virological and molecular testing. In theory, each site should send 5–10 paired serum specimens taken from a random sample of patients with suspected dengue each week throughout the year. In reality, patients are seldom randomly selected and only two sites regularly send specimens throughout the year. The paired serum specimens are tested using an immunoglobulin M (IgM)-antibody capture enzyme-linked immunosorbent assay (ELISA) and a haemagglutination inhibition assay. Because of possible cross-reactivity, all specimens are systematically tested for anti-dengue virus and anti-Japanese encephalitis virus IgM using an in-house IgM-antibody capture ELISA and a haemagglutination inhibition assay, as previously described.5 The first sample is tested for viral ribonucleic acid using a modified version of the reverse-transcriptase polymerase chain reaction (PCR) procedure described by Lanciotti.6 In addition, the virus is isolated by inoculating sera into C6/36 (Aedes albopictus mosquito) and Vero E-6 cell cultures and identifying the virus serotype by using a direct fluorescent antibody assay employing monoclonal antibodies, as described elsewhere.5

Case definition and data collection

Since 2002, clinical case definitions of dengue fever and its complications have been based on World Health Organization (WHO) definitions7,8 and adapted for health centres and referral hospitals. Because resources were limited, the NDCP gathered data reported passively from referral hospitals and collected actively at sentinel sites on only a weekly basis. Data were collected on individual patients using a standard NDCP form, which recorded each patient’s name, demographic characteristics, disease severity (i.e. dengue fever, DHF or DSS), district of residence, and vital status or status on transfer. The forms were stored centrally at the NDCP office and data were entered into a computerized database using statistical software (Epi Info 2000 version 3.3.1, Centers for Disease Control and Prevention, Atlanta, United States of America (USA)). A system was in place to check patients’ names so that there was no duplication of those who were hospitalized at several different sites for the same illness episode.

Vector control interventions

In theory, since 2001 control of the dengue vector in Cambodia has consisted of biannual larvicide campaigns: 1% temephos sand granules distributed between April and July and between August and October. Medium-to-large water storage containers in households in districts identified by the NDCP as high-risk areas for epidemics were targeted. Targets were mainly in urban centres and densely populated areas. These campaigns were linked to nationwide publicity involving public service announcements on radio and television and in the print media, as well as the use of vehicles with loudspeakers and community meetings before each dengue season. However, because of budgetary constraints, some high-risk districts received only one round of larvicidal treatment between April and July or no treatment at all. Routine vector control activities were also limited and primarily involved community-based clean-up campaigns to remove and destroy small rain-filled containers and insecticide fogging to kill adult mosquitoes around houses close to locations where dengue cases had been reported.

Since the distribution of temephos has not been documented in detail, vector control coverage in each district in the years 2001–2008 was determined by ascertaining whether or not the NDCP intervened in that district in a specific year.

Data analysis

The analysis considered only data recorded and computerized from 2002 onwards because data for 1980–2000 were not collected using a strict clinical case definition for suspected dengue virus infection and data for 2001 were incomplete: 68% of demographic and district-of-residence data were missing. We calculated the age-specific incidence of dengue and the age-adjusted annual incidence per 1000 individuals using population data from the 1998 census.9 Population estimates for other years were obtained from the Cambodian government’s Institute of Statistics.3 The annual number of cases was treated as a time series and the Prais–Winsten generalized linear regression model was used to calculate the significance of any increase or decrease in dengue incidence between 2002 and 2008, both overall and for each district. A change of slope was judged to be statistically significant using the F-statistic if the P-value was < 0.05. We assessed the impact of vector control interventions in individual districts by determining their effect on dengue incidence using a logistic regression model that controlled for the population density in each district. The number of vector control interventions was treated as a continuous variable, with the number per district being the number of years that interventions were used in that district. Results were expressed in odds ratios (ORs) and 95% confidence intervals (CIs). All statistical analyses were carried out using Stata 9.2 statistical software (StataCorp LP, College Station, USA).


Secular trend and seasonality

Of the 194 726 cases of dengue reported to the NDCP between 1980 and 2008, 74 947 (38.5%) were passively reported by public health-care facilities before 2001 using non-standardized clinical definitions of dengue. The secular, or long-term, trend was characterized by a cyclical pattern of epidemics at intervals of about 3–4 years. Since the surveillance system was improved in 2001, the 3–4-year cycle has been less prominent. Two major epidemics occurred after 1997: there were 16 260 cases in 1998 and 39 618 in 2007 (Fig. 1).

Fig. 1. Number of cases of dengue fever reported nationally in Cambodia, 1980–2008
Fig. 1. Number of cases of dengue fever reported nationally in Cambodia, 1980–2008
D2, dengue virus type-2 (DENV-2) was predominant in the year; D3, dengue virus type-3 (DENV-3) was predominant in the year.

Trends in incidence 2002–2008

In the period 2002–2008, the NDCP reported between 9006 and 39 618 cases of dengue per year (annual age-adjusted incidence range: 0.7–3.0 per 1000 population), with the case fatality rate ranging from 0.7 to 1.7% (Table 1). Dengue cases were reported throughout the year, with increases occurring during the rainy season between May and November (i.e. weeks 17–48 in Fig. 2). After taking into account seasonal fluctuations and the major 2007 epidemic, analysis using the generalized linear regression model detected no significant trend in the annual age-adjusted incidence of reported clinical dengue virus infections.

Fig. 2. Incidence of dengue fever in 2007 and mean incidence for 2002–2006, by reporting week, Cambodia
Fig. 2. Incidence of dengue fever in 2007 and mean incidence for 2002–2006, by reporting week, Cambodia
a The first reporting week is the first week in January.

Since the implementation of sentinel surveillance, the proportion of all dengue cases reported that came from sentinel sites has increased from 57.0% in 2002 to 89.1% in 2008 (Fig. 1). For example, in 2008 the two non-profit-making hospitals belonging to the Kantha Bopha Foundation in Siem Reap and Phnom Penh, respectively, accounted for 62.1% of all reported cases. These hospitals provide free medical care to Cambodian children and have large catchment areas.

Overall, from 2002 to 2008, the average proportion of clinical dengue virus infections classified as DHF was 41.5% (range: 20.5–54.0), while 6.6% (range: 3.0–8.7) were classified as DSS and the remainder, as dengue fever (Table 1). The proportion classified as either DHF or DSS peaked in 2006, at 60.6%, and in 2007, at 54.2%.

The highest age-specific incidence of dengue fever occurred in infants aged less than 1 year, followed by those aged 4–6 years (Fig. 3). Some 79.0% of all reported cases were in children aged 9 years or younger (median: 6 years). The age distribution of dengue cases has been consistent since 2002. Moreover, no sex difference in incidence was observed in the period since 2002, during which the median proportion of males was 49.3% (range: 47.7–49.6).

Fig. 3. Age-specific incidence of dengue fever, Cambodia, 2002–2008
Fig. 3. Age-specific incidence of dengue fever, Cambodia, 2002–2008
a There was no requirement to report dengue in those aged 15–19 years.

Virological findings

Between 2000 and 2008, paired serum samples were collected from an annual mean of 715 patients, who comprised 5.2% of all dengue cases reported. Overall, 87.8% of samples were seropositive for dengue and there was little variation across sentinel sites. On average, 70.0% of seropositive samples also tested positive using PCR. Among seropositive patients aged < 1 year, 78% (i.e. 108 of 138) tested positive using PCR. Although most cases occurred during the rainy season, dengue virus infection was also identified during other times of the year, which confirms that dengue is endemic in Cambodia.

Since virological testing started in 2000, all four dengue virus serotypes have been observed to be in circulation each year, with DENV-2 and DENV-3 being predominant (Table 1). The predominant circulating serotype changed from DENV-3 to DENV-2 in 2002 and then switched back to DENV-3 4 years later (Fig. 1). Between 2000 and 2008, both the incidence of dengue and the proportion of cases with DHF were highest in 2006 and 2007, when the predominant serotype was DENV-3.

Impact of vector control

Between 2000 and 2008, dengue vector control interventions based on the distribution of temephos, community participation and the provision of educational messages were undertaken in 94 densely populated districts that the NDCP considered to be most affected by the disease. Of these, only 24 (35%) received interventions for 4 years or more (median: 2; range: 1–7). Linear regression analysis showed that the incidence of dengue declined in only 5 (2.7%) of all 185 districts studied between 2000 and 2008, while it remained unchanged in 162 (86.2%) and increased in 18 (9.6%). Two (40.0%) of the five districts in which the incidence declined had received interventions during the previous 7 years compared with 47.5% of districts where the incidence was unchanged and 33.3% where it increased. Logistic regression models, whether controlling for the district population density or not, failed to find any significant association between the use of interventions and decreased incidence.


This is the first published report of national dengue surveillance data in Cambodia covering a period of 28 years. As the data for 1980–2000 were not collected using a strict clinical case definition for suspected dengue virus infection, we focused on the 2002–2008 period, during which more complete and reliable data on patients and the virus serotype were available.10 The estimated incidence of dengue nationally was high, varying from 0.7 to 3.0 per 1000 population during 2003–2008. Generally there was no change in the overall age-adjusted annual incidence during 2002–2008, although there was a spike in case numbers in 2007. The data also show that dengue remains prevalent among young children in Cambodia, with infants aged < 1 year and children aged 4–6 years being the most affected. The age distribution of dengue cases in other countries in the region showed wide variations. In Thailand and Viet Nam, dengue has become more common in older children.1114 A prospective cohort study of children aged 3–15 years in southern Viet Nam found that the incidence was highest in those aged 6–10 years (L Pollissard, personal communication, 2007). In Malaysia and Singapore, in contrast, most cases were seen in adults aged over 18 years.15,16 The reasons for these differences may include the level of development,17 the effectiveness of vector control programmes,15 the predominance of different virus genotypes18 and a demographic transition or shift.19

Significantly, the alert system for detecting epidemics established by the NDCP and modelled on the early warning system for malaria20 predicted the occurrence of the 2007 epidemic. As shown in Fig. 2, the weekly dengue incidence was consistently above the alert threshold of two standard deviations above the mean in early 2007, while it remained below the threshold in other years. Although the authorities were quickly alerted, the response to the outbreak, which included vector control interventions, educational messages and providing public hospitals with sufficient medical supplies, came too late. Unfortunately, the NDCP has too few human resources and too little funding to implement these interventions in a timely manner.

Although all four dengue virus serotypes were circulating in the country throughout the reported surveillance period, illness was predominantly caused by DENV-2 and DENV-3. The change in the predominant serotype from DENV-3 to DENV-2 in 2002 resulted in only a small increase in incidence, possibly because DENV-2 had been circulating in earlier years and many Cambodians had developed immunity. The incidence declined steadily between 2002 and 2005, until a large-scale epidemic due to DENV-3 occurred in 2006–2007. We speculate that there is a 3–4 year cyclical pattern of epidemics in Cambodia involving different serotypes, with epidemics of the same serotype possibly occurring every 8–9 years (e.g. in 1998 and 2006–2007). Indeed, unpublished laboratory data from the Institut Pasteur–Cambodia and the large-scale DENV-3 epidemic observed regionally in 19982126 indicate that the serious 1998 epidemic which disrupted the Cambodian health system by overloading hospitals27 was due to DENV-3. Moreover, in the 2006–2007 epidemics in Cambodia, the DENV-3 virus was associated with a high proportion of severe complications (i.e. DHF and DSS). In Thailand, DENV-3 was also predominant during the severe dengue years of 1987 and 1998.21 Further studies, which should include full genome sequencing, are needed to explore the association between dengue serotype, virus virulence and disease severity.

Several studies have shown that vector control interventions reduce larval indices though the reduction must be substantial to influence dengue transmission.2830 In Cuba and Singapore, the incidence of dengue was dramatically reduced only after anti-vector legislation was introduced and aggressive vector control measures had been used for years.31 Nevertheless, dengue has re-emerged because new dengue viruses are constantly being introduced from neighbouring countries.15,32 In this study, we evaluated the impact of many years of vector control interventions and educational messages on the incidence of dengue in individual districts. No association between such interventions and disease incidence was observed. Clearly, as dengue transmission is highly localized, a more rigorous assessment of interventions would consider outcomes at the village level, but this information was not available.33 We are not suggesting that temephos does not prevent transmission at the household level, but rather that, given the limited resources, it is unlikely that current interventions in Cambodia will affect disease incidence.

In Cambodia, dengue viruses are transmitted primarily by Aedes aegypti mosquitoes (C Paupy, personal communication, 2002), which are abundant in populated rural areas.34 Over 80% of larval foci for this species are in the ubiquitous, concrete jars filled with rainwater used in most homes. Unfortunately, the quantity of parricides available to the NDCP has been insufficient to cover all high-risk areas thoroughly and there was little reduction in dengue transmission. The long-term strategy for implementing vector control programmes in Cambodia needs to be re-examined.

The use of surveillance data to describe epidemiology and evaluate disease burden has several limitations. In particular, there are weaknesses in the design of the surveillance system resulting from the need to balance limited resources and data quality. For example, surveillance only covered patients hospitalized at major public and non-profit-making paediatric hospitals and paediatric wards to ensure the accuracy of dengue diagnosis. Moreover, clinicians at our surveillance sites often had difficulty in classifying disease severity using standard WHO definitions (Institut Pasteur–Cambodia, unpublished data, 2007). Improving diagnosis by obtaining complete blood counts or carrying out radiographic or ultrasound imaging is often too technically difficult or too expensive for most health-care facilities in Cambodia. The presence of haemoconcentration, suggestive of DHF, is also difficult to detect because Cambodian clinicians tend to administer fluids intravenously as soon as dengue is suspected.35

The size of the patient samples used in virological surveillance was small. Moreover, patients suspected of having dengue were not selected randomly but rather because there was a high level of suspicion that they had severe dengue. Another limitation was that dengue was frequently overdiagnosed during epidemics and underdiagnosed during the intervening periods. The use of laboratory testing in dengue diagnosis is clearly vital when resources permit. We believe that, in the absence of systematic laboratory diagnosis of dengue, surveillance programmes should exclude patients with undifferentiated febrile illnesses to increase the specificity of diagnosis by avoiding the inclusion of those with, for example, influenza, typhoid or leptospirosis.36,37

Despite these limitations, our observation that dengue activity patterns for different ages and genders have remained consistent over time indicates that the surveillance data are reliable. Moreover, no other data available match the completeness or cover the same timescale as the Cambodian national dengue surveillance data.

Another aim of this article was to make the Cambodian surveillance data publicly available for comparison with other surveillance data in the hope that this will lead to better understanding of the pattern of dengue transmission in the region. Currently, however, descriptive national data are difficult to obtain. Differences in the surveillance systems used in other countries must be taken into account. For example, in Malaysia and Singapore, all suspected dengue cases are confirmed by laboratory testing, whereas only hospitalized patients are tested in the Philippines and Thailand. In Viet Nam, as in Cambodia, only a sample of patients suspected of having dengue undergo serological or virological testing and it is not clear whether these patients are representative of the general population. In contrast, in the Philippines and Viet Nam, all clinically diagnosed dengue cases at all health-care facilities, including health centres and hospitals, are reported.38

With the development of dengue vaccines expected in the near future,39,40 there is an urgent need to accurately estimate the true disease burden. Several countries are collaborating with the Paediatric Dengue Vaccine Initiative of the International Vaccine Institute in Seoul, the Republic of Korea, to set up community-based surveillance sites to measure the incidence of dengue accurately.


We are grateful to Ole Wichmann and Paul Kitsutani for their comments in reviewing this manuscript. We also thank Socheat Duong, Director of the National Center of Parasitology, Entomology and Malaria Control, for his continuing support of the national surveillance system.

Competing interests:

None declared.