^{1}

^{2}

^{3}

^{3}

^{3}

^{4}

^{1}

^{5}

^{3}

^{1}

^{6}

^{7}

^{1}

^{5}

^{*}

Conceived and designed the experiments: PH NTH HQC TVP TND NNC JF VSN. Performed the experiments: HQC PH TVP NNC. Analyzed the data: HQC PH KTDT. Contributed reagents/materials/analysis tools: NNC TVP. Wrote the paper: PH HQC KTDT.

The authors have declared that no competing interests exist.

An estimated 2.4 billion people live in areas at risk of dengue transmission, therefore the factors determining the establishment of endemic dengue in areas where transmission suitability is marginal is of considerable importance. Hanoi, Vietnam is such an area, and following a large dengue outbreak in 2009, we set out to determine if dengue is emerging in Hanoi.

We undertook a temporal and spatial analysis of 25,983 dengue cases notified in Hanoi between 1998 and 2009. Age standardized incidence rates, standardized age of infection, and Standardized Morbidity Ratios (SMR) were calculated. A quasi-Poisson regression model was used to determine if dengue incidence was increasing over time. Wavelet analysis was used to explore the periodicity of dengue transmission and the association with climate variables. After excluding the two major outbreak years of 1998 and 2009 and correcting for changes in population age structure, we identified a significant annual increase in the incidence of dengue cases over the period 1999–2008 (incidence rate ratio = 1.38, 95% confidence interval = 1.20–1.58, p value = 0.002). The age of notified dengue cases in Hanoi is high, with a median age of 23 years (mean 26.3 years). After adjusting for changes in population age structure, there was no statistically significant change in the median or mean age of dengue cases over the period studied. Districts in the central, highly urban, area of Hanoi have the highest incidence of dengue (SMR>3).

Hanoi is a low dengue transmission setting where dengue incidence has been increasing year on year since 1999. This trend needs to be confirmed with serological surveys, followed by studies to determine the underlying drivers of this emergence. Such studies can provide insights into the biological, demographic, and environmental changes associated with vulnerability to the establishment of endemic dengue.

Dengue is the most common vector-borne viral disease of humans, causing an estimated 50 million cases per year. The number of countries affected by dengue has increased dramatically in the last 50 years and dengue is now a major public health problem in large parts of the tropical and subtropical world. It is of considerable importance to understand the factors that determine how dengue becomes newly established in areas where the risk of dengue was previously small. Hanoi in North Vietnam is a large city where dengue appears to be emerging. We analyzed 12 years of dengue surveillance data in order to characterize the temporal and spatial epidemiology of dengue in Hanoi and to establish if dengue incidence has been increasing. After excluding the two major outbreak years of 1998 and 2009 and correcting for changes in population age structure over time, we found there was a significant annual increase in the incidence of notified dengue cases over the period 1999–2008. Dengue cases were concentrated in young adults in the highly urban central areas of Hanoi. This study indicates that dengue transmission is increasing in Hanoi and provides a platform for further studies of the underlying drivers of this emergence.

Dengue is caused by infection with one of four genetically related but serologically distinct dengue virus serotypes, which are transmitted by the bite of an infected female

Dengue epidemiology is a determined by a complex interaction of vector, pathogen, and host biology; macro and microclimate; the physical environment; and social factors

The study used public health surveillance data routinely collected through the Ministry of Health notifiable diseases surveillance program. The study did not use patient medical records and all data were analyzed anonymously. The analysis of the surveillance data was approved by the National Institute of Hygiene and Epidemiology, a specialized institute of the Vietnam Ministry of Health.

Hanoi is Vietnam's capital and its second largest city. It is located in the north of the country in the low lying and densely populated Red River delta. In the 2009 decennial census, the population of Hanoi was estimated at nearly 6.5 million, with a population density of 1943 persons per km^{2}. In contrast to South Vietnam, which is always hot but with dry and rainy seasons, North Vietnam has four distinct seasons with hot and humid summers, receiving the majority of rainfall, and cool and relatively dry winters.

Dengue is one of 24 infectious diseases in Vietnam for which there is monthly mandatory reporting from all administrative levels, which, in increasing size, are communes, districts, provinces, and regions. The detection and reporting of dengue cases follows the Ministry of Health 1999 guidelines on surveillance, diagnosis and treatment of dengue

Anonymous individual case reports of all clinically diagnosed dengue cases of any disease severity in Hanoi residents reported through the public health surveillance system were obtained from Hanoi Preventive Medicine Center (PMC) from1st January 1998 to 31^{st} December 2009. Data available on every case were gender, commune and district of residence, and month and year of diagnosis. Age was missing for two subjects only.

Age stratified population estimates for Hanoi by district were derived as follows. Data for 1999 and 2009 were obtained directly from the decennial national Population and Housing Census' conducted by the General Statistic Office of Vietnam (GSO.

Daily records of total rainfall; mean wind velocity; mean, maximum and minimum temperature; and relative humidity from a central Hanoi meteorology station were obtained from the National Centre for Hydrometeorological Forecasting (

To adjust for the potential confounding effect of changes in the age structure of the population of Hanoi over the study period, we applied age specific incidence rates for each year to a standard population (direct age standardization)

To determine if dengue incidence was increasing over time, a quasi-Poisson regression model was used with the annual count of dengue cases as the outcome variable, year as the independent variable, and (log-transformed) mid-year population size as an offset. The age adjusted count per year, rather than per month, was used since monthly counts are non-independent because of seasonal patterns in dengue incidence, and monthly counts would therefore not fulfill the Poisson assumption of independence of outcomes. A quasi-Poisson model was used as the data were over-dispersed. Linear regression was used to test whether there was a linear time trend in the median age of infection. Statistical analysis was conducted in R 2.9.0 (R Foundation for Statistical Computing, Vienna, Austria).

Standardized Morbidity Ratios (SMR) for the 2009 dengue outbreak were calculated using the age-stratified population distribution from the 2009 census and age-stratified dengue incidence in Ba Dinh District in 2009 as the reference population (SMR = 1). SMR's were displayed using the statistical software R 2.9.0 (R Foundation for Statistical Computing, Vienna, Austria).

Wavelet analysis was used before to explore the periodicity of dengue transmission and the association with climate variables. Wavelet analysis is especially suited to time series where the average, variance, or relationship with co-variates change over time (i.e. non-stationary), and provides the possibility to explore linkages between multiple non-stationary time series

Between 1998 and 2009, 25983 dengue cases were notified to Hanoi PMC. Large outbreaks were observed in 1998 and 2009, with smaller annual outbreaks in intervening years (

(A) Incidence of notified dengue cases per 100,000 persons, 1998-2009. (B) Log number of notified dengue cases, 1998–2009. Red line is the mean monthly temperature in degrees centigrade. Blue line is the total monthly rainfall in millimeters.

In the quasi-Poisson model there was no statistically significant yearly increase in age adjusted incidence when all years were included (

Time period | Parameter Estimate (log rate ratio) | Standard error (of log rate ratio) | P-value | Estimated incidence rate ratio | 95% CI of incidence rate ratio |

1998–2009 | 0.23 | 0.16 | 0.18 | 1.25 | 0.92–1.71 |

1999–2008 | 0.32 | 0.07 | 0.002 | 1.38 | 1.20–1.58 |

The unadjusted median and mean age of notified cases were 23 and 26.3 years respectively, and 53% of cases were male. The majority of notified cases (85%) occurred in individuals aged over 15 years (

(A) Histogram of age of notified cases, all years combined. (B) Box and whisker plot of unadjusted age of notified dengue cases by year. The black bar denotes the median, the open box the upper and lower quartiles, and the whiskers the maximum and minimum values. The open triangles denote the median age of notified cases after age standardization to the 2009 census data. The T symbol denotes the mean age of notified cases after age standardization to the 2009 census data.

IgM ELISA results were available from 3678 cases notified between 2002 and 2009. The proportion of cases with an available ELISA result varied from 68.2% in 2002 to 4.5% in 2009. The overall, proportion of ELISA tests that were anti-dengue virus specific IgM positive was 23.7%, ranging from 13.6% in 2006 to 49.8% in 2003. The lowest annual test positivity rates were observed in 2006 (13.6%), 2007 (17.6%), and 2008 (16.1%).

Dengue SMRs were highest in the central districts of Hanoi, with the districts with an SMR greater than three all being highly urbanized and densely populated areas in the centre of the city (

Continuous wavelet analysis of the aggregated time series showed a significant annual cycle of dengue transmission (

(A) Cases per month (x100) in Hanoi from 1998 to 2009. (B) Log transformation of cases in 14 central districts of Hanoi. (C) Continuous wavelet transformation of the aggregated time series. Colors code indicate increasing intensity, from blue to red; black broken lines show statistically significant areas (threshold of 5% confidence interval); the curved black solid line delimits the area within which inferences are robust because of sufficient time series data (the cone of influence), (D) Power wavelet spectrum solid line: the mean spectrum; dotted line: threshold value of 5%.

All local climatic variables show a statistically significant association with dengue in the 1-year mode (

(A) Rainfall, (B) Mean temperature, (C) Humidity and (D) Wind velocity. The left panel represents the wavelet coherency in which blue and red indicate low and high coherence respectively. The black broken lines show α = 5% significance level. The curved black solid line delimits the area within which inferences are robust because of sufficient time series data (the cone of influence), The right panel represents the oscillations of the time series in the phase analysis. Red lines: dengue time series, blue lines: climatic variables.

In Hanoi dengue transmission demonstrates clear annual cycles that are associated with a lag of around two months with seasonal increases in mean temperatures and rainfall. This pattern reflects climate driven changes in vector abundance, survival, and biting frequency; and the time it takes for a newly infected mosquito to become infectious to humans (the extrinsic incubation period)

Large outbreaks occurred in 1998 and 2009, and wavelet analysis indicated possible 2–3 years periodicity. A 2–4 years cycle has been described in Thailand, Cambodia, and South Vietnam. Moreover, a meta-cycle between 8–10 years has been observed in Thailand

Excluding the two major outbreaks (in 1998 and 2009), the annual age adjusted incidence of notified dengue increased significantly. However, over the same period the average age of infection has not decreased, as would be expected if the per capita rate of infection in susceptible people (the force of infection) was increasing

The average age of notified dengue cases in Hanoi is high and 85% are aged over 15 years. Although the average age of clinical cases in Thailand has increased over recent years, the clinical burden of dengue in much of Southeast Asia still primarily falls on children aged less than 15 years

Our study is limited by its reliance on passively notified cases of clinically apparent dengue infection. Only a small proportion of all notified cases were dengue IgM positive and the lowest IgM test positivity rates were in 2006, 2007, and 2008. More cases were notified in each of these three years than in any other years except the outbreak years of 1998 and 2009. If the size of the 2006–2008 outbreaks were inflated by the reporting of an unusually large number of non-dengue illnesses, the increase in dengue incidence we have identified may be false. The data reported here are of illness episodes that met the surveillance case definition for dengue, which includes both fever and rash. Measles also presents with fever and rash, and although included in the Expanded Program of Immunization (EPI), there was a large national epidemic in 2009

Since anti-dengue virus IgM is not detectable until 3–5 days after infection, and is frequently negative in secondary dengue infections, asays to detect anti-dengue virus IgM antibodies have limited sensitivity (perhaps <20%) when used alone to diagnose dengue infections during the acute illness

As dengue disease severity is modulated by age, prior dengue exposure history, and the infecting dengue serotype, the incidence of clinically apparent dengue cases is unlikely to have a linear relationship with transmission intensity

We thank all reporting laboratories, physicians, epidemiologists, entomologists as well as Preventive Health Center and National Institute of Hygiene and Epidemiology public health departments, for their cooperation and advice in collecting these surveillance data.