Background
Dengue is a mosquito-borne viral disease in which individuals can suffer up to four times during their lifetime due to the existence of four distinct serotypes (DENV1-4). A primary (first) dengue infection with any serotype induces IgG antibodies that offer protection against homologous serotype infections yet allow subsequent post-primary (secondary, tertiary or quaternary) infections with heterologous serotypes [
1,
2]. Severe dengue disease is associated with, though not limited to, secondary infections due to pre-existing, cross-reactive, IgG antibodies that do not protect against infection, but rather facilitate antibody-dependent enhancement of viral replication [
3]. Without specific therapeutics or a widely available vaccine, costly vector control interventions remain the predominant method of dengue control [
4]. To appropriately deploy these limited interventions, dengue national surveillance operations in endemic countries must accurately monitor force of infection (FOI) over space and time.
In countries where dengue is a notifiable disease, national surveillance efforts typically collate dengue case reports over specified time periods to generate incidence estimates. This readily available data can be used to inform the deployment of limited control interventions, thus maximise their impacts. However, case reporting is heavily influenced by disease awareness and variable healthcare infrastructure, which can distort generated measures [
5]. Alternatively, population-based surveys or cohort studies including seroprevalence [
6] and entomological surveys [
7] can provide more reliable estimates of disease risk within their operational setting and are considered the gold standards for measuring long-term average transmission intensity. Yet, surveys are labour-intensive and difficult to conduct routinely over large geographical areas, so have limited use for routine surveillance proposes.
Recently, age-stratified incidence measures of case reports have been proposed as more suitable indicators of transmission intensity [
5,
8,
9]. In areas of high transmission, the burden of disease is believed to disproportionately impact younger individuals compared to lower transmission areas where individuals likely report with dengue later in life due to the accumulation of immunity over time. This approach helps counter bias introduced by variable health care infrastructures and disease awareness although suffers two major caveats. First, case reporting is dependent on a variety of different diagnostics between and within endemic countries. Surveillance operations rely on non-specific diagnostics or clinical manifestations for notifying dengue [
10]. Consequently, other co-endemic febrile infections, which manifest similarly to dengue and are common among children may prompt more younger individuals to seek care and be misdiagnosed as dengue infections [
11,
12]. Second, among those reporting with true dengue infections, routinely used diagnostics are currently unable to distinguish those experiencing primary or post-primary infections [
1]. Therefore, the age of case reports may be influenced by spatio-temporal imbalances in reported dengue immune status. Stratifying age estimates among those truly experiencing their primary infection may help to overcome these limitations.
In the Philippines, dengue has become a major contributor to mortality and morbidity over the past 50 years with all four serotypes believed to be endemic across urban centres [
13]. Existing dengue surveillance operations consist of collating all reported dengue infections and surveying a subset for subsequent laboratory analysis. The serum is collected from sampled case reports for molecular and serological testing to allow spatio-temporal monitoring of dengue serotypes. Prior to this study, we utilised laboratory data from surveyed dengue patients to develop an algorithm that accurately determines individual primary or post-primary (secondary, tertiary or quaternary) immune status [
14]. Despite this, it remains unknown how surveillance metrics derived from this algorithm can best be used to routinely characterise the FOI. Previous studies have demonstrated that the FOI can be reasonably estimated from age-stratified incidence rates [
5,
9]. However, it remains unknown if other, easily computed, surveillance metrics represent surrogate indicators of the FOI. Here, we investigated whether laboratory/non-laboratory surveillance metrics correlate with the FOI according to age-seroprevalence and be used to routinely predict the burden of dengue across urban centres.
Discussion
We explored methods to routinely describe dengue transmission intensity among cities across the Philippines. Using data from surveillance operations, we found the mean age of laboratory-confirmed primary dengue infections and the mean age of case reports with warnings signs, to a lesser extent, represented suitable surrogate indicators of the FOI. Both these easily computed metrics correlated better than crude and age-stratified incidence with the FOI estimated from seroprevalence data. Our results highlight prominent spatio-temporal heterogeneity in dengue serotype dominance and reported immune status and the FOI across urban centres in the Philippines.
It is well documented that secondary, opposed to primary, dengue infections are at a greater risk of developing severe disease [
20‐
22], yet we found those categorised as post-primary infections were only slightly more likely to present with severe disease than primary infections. This made us suspect that some of those classified as post-primary were tertiary and quaternary dengue infections, given these types of infections are thought to be less severe than secondary infections, although it is still widely assumed post-secondary infections are often asymptomatic [
2,
23]. Furthermore, we identified all four serotypes of dengue in every city almost every year making it plausible for individuals residing in urban centres to suffer multiple dengue infections over their lifetime. Lastly, we found that younger, opposed to older, post-primary dengue infections were at greater risk of presenting with severe disease. A trend possibly attributed to younger and older post-primary infections representing secondary and post-secondary infections, respectively, as shown in [
2], although this may also be attributable to a different disease presentation in older people.
Our city FOI estimates, aggregated over 5 years and generated from catalytic models, were comparable to a 2016 estimate in Cebu (central Philippines) [
23] and revealed a long-term spatial heterogeneity in dengue transmission intensity, similarly shown in China [
24] and Colombia [
25]. We found catalytic models fit with sero-reversion parameters resulted in superior model fits despite sero-reversion being very low across sampled cities. This is consistent with previous findings that documented a minority of dengue cases, who were likely exposed once, experienced IgG waning [
17,
18]. To determine whether the FOI could be estimated from data routinely collected by dengue surveillance operations, we explored the association between FOI, estimated from catalytic models, against various dengue surveillance metrics. Surprisingly, aggregated at the city level and over the study period, all-age reported incidence was higher in lower FOI areas. This trend however has previously been reported in Singapore [
26] and was thought to be attributed to an age shift in disease burden. In low transmission areas, individuals are more likely to experience a dengue infection at a later stage in their life when they may have heightened individual disease awareness and possibly easier access to care. Consequently, low transmission areas may experience increased case reporting. In addition, this trend could also be explained by reduced health-seeking behaviour in high FOI cities wherein individuals refrain from reporting with symptoms as they are aware of many others with this often mild, self-limiting disease. In contrast in low FOI areas, individuals may be more prompted to seek care as they less familiar with the disease and are more concerned by unfamiliar symptoms. However, future population KAP (knowledge attitudes and practises) surveys in high- and low-transmission settings would be valuable in better describing these health-seeking patterns. Finally, this disparity could partially be explained by the FOI and incidence being measures of infection and disease, respectively. FOI estimates also include those asymptomatic with dengue who are thought to account for approximately 75% of all dengue cases and would unlikely seek treatment [
27]. More surprisingly however, after stratifying incidence by age, we observed no association with the FOI which contradicts findings in [
9]. We speculate that variable healthcare access and reporting across the country may account for this irregularity and remains an area of continued investigation. Together, these findings suggest that simply allocating control interventions to areas with elevated reported incidence might exclude regions with higher burdens of infection.
In cities, we found that the average age of reported dengue infections with clinical dengue outcomes increased with decreasing aggregated transmission intensity. Interestingly, the mean age of those reporting with just warning signs of dengue, opposed to severe dengue, proved a superior indicator of FOI. The poor specificity dengue clinical diagnosis is well documented as other febrile infections can present with similar disease manifestations [
28‐
30]. Yet, compared to warning signs, severe disease is a far rarer outcome, particularly among primary and older post-primary infections. Moreover, in Thailand, reported severe dengue manifestations have been shown to be highly spatially and temporally heterogenous due to climactic factors [
31]. Together, these factors likely account for the lack of association between FOI and age of case reports with severe disease.
In contrast to the age of clinically diagnosed dengue patients, the age of laboratory-confirmed dengue cases had a stronger association with the FOI, likely a consequence of excluding other co-endemic febrile infections. Yet, among laboratory confirmed infections, the age of those reporting with primary, rather than post-primary, infections proved a better indicator of FOI. We propose two main factors account for this difference. Age patterns of reporting post-primary infections might be influenced by spatio-temporal imbalances in immune status reporting, observed in this study and previously in Vietnam [
32] and Taiwan [
33]. In high transmission settings, where a higher proportion of the post-primary reporting population include tertiary/quaternary infections, the mean age would be higher than an equally high transmission setting where most of the post-primary infections were secondary infections. In contrast, the age of reported primary infections remains unaffected by imbalances in immune status reporting. Secondly, studies have shown that Zika, a structurally homologous virus to dengue, elicits cross-reactive IgG that primes individuals for secondary dengue infections [
34,
35]. Therefore, spatio-temporal imbalances in potential Zika transmission across the Philippines may similarly impact the age of post-primary, yet not primary, dengue infection reporting. These interactions led us to propose the age at which individuals report with their first dengue infection is the better surrogate of transmission intensity which we used to estimate annual FOI across cities. It should be noted however, given that individuals are classified as primary or post-primary according to antibody titre and ratio thresholds, a minority of the study population with metrics on the cusp of these cut offs may have been misclassified [
14].
Upon estimating city level, annual dengue attack rates according to the age of primary dengue infections, we demonstrated that yearly estimates correlated with overall attack rates according to the catalytic model. Patterns in predicted annual attack rates across cities revealed changing patterns in transmission intensity over time, as seen in Singapore [
26]. We identified cities where transmission is stable, emerging and decreasing. Information that is critical for national dengue surveillance operations to plan for the future [
36]. In Zamboanga, despite a relatively low overall FOI, annual predicted attack rates revealed that transmission intensity increased during the study period, a trend undetected in the collated incidence data. Zamboanga could therefore be earmarked for heightened surveillance activities to better characterise worrying predicted trends in the burden of infection. In Naga, we identified a jump in FOI coinciding with a switch in serotype dominance, yet this was unobserved in other cities. Approximating transmission intensity from the age of reported primary dengue infections has two major benefits. Firstly, existing surveillance operations can generate attack rates quickly and simply. Moreover, as a consequence of the exponential relationship, slight changes in the age of reported primary infections correspond to larger changes in transmission intensity. This enables FOI estimates to be generated at more granular temporal scales, which are more informative than long-term estimates derived from catalytic models. As well as describing dengue transmission intensity patterns, we speculate monitoring age patterns of reported primary cases could assist in dengue vaccine deployment. Currently, the only licenced dengue vaccine, Dengvaxia®, is recommended for individuals aged between 9 and 45 years in endemic areas, presumably as these likely include those with prior dengue exposure [
37]. Our results highlight the spatio-temporal heterogeneity in the age patterns of those reporting with their first infection; therefore, monitoring the age of these infections could potentially help identify suitable age ranges in specific locations to receive vaccination, although individual pre-vaccination testing would still be necessary and this requires legislative changes.
There are limitations associated with these findings. FOI estimates were generated from reporting, non-active, dengue infections with/without previous dengue IgG exposure. Consequently, we may have oversampled individuals who were more prompted to seek health care if they experienced a prior dengue infection. Yet, only 25% of dengue infections are thought to experience symptoms that would prompt them to seek care [
27], and we concluded this health-seeking bias was minimal. Secondly, our fitted catalytic model assumes serotypes contributed equally to transmission over time. Still, given we identified every serotype in each city and that serotype dominancy changed rapidly over the 5-year sampling period, it is reasonable to assume that over the previous decades, from which the FOI is estimated, periodic serotype dominancy may have resulted in serotypes contributing equally to overall transmission. Lastly, our analysis confirms the age of primary dengue infections is associated with FOI at the city level, yet it remains unconfirmed as to whether the age of reported dengue cases accurately represent transmission intensity at lower administrative levels where patients are more likely to cross administrative borders for care. Similarly, our analysis focused on revealing annual trends in dengue transmission intensity. It remains unknown whether seasonal biases may influence the reporting age of primary dengue infections at more granular temporal scales.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.