COVID-19 and dengue are two clinically similar entities, especially within the first 24 to 48 hours from symptom onset [10]. In a context of co-epidemics, our cohort study, conducted within a SARS-CoV-2 testing center upon mild to moderate cases of COVID-19 and non-severe cases of dengue identified several key distinctive features for both infections. Among the clinically discriminant variables at presentation, retro-orbital pain, body ache and headache were strong predictors of dengue while anosmia was the only predictor of COVID-19 and URTI symptoms were indicative of OFIs. To a lesser extent, gut symptoms other than diarrhea, dysgeusia and fatigue were suggestive of dengue whereas cough referred to another diagnosis (OFIs or COVID-19), albeit found in nearly a third of dengue. Among the epidemiological variables, the contact with a COVID-19+ case and a delayed presentation beyond three days of symptom onset were predictive of COVID-19, a rapid presentation within three days was suggestive of dengue, while active smoking was less likely observed with COVID-19 or associated with OFIs. These elements are summarized in the S1 Figure.

Our findings reveal several unexpected differences at the presentation to hospital between COVID-19 or dengue as compared to OFIs, and between COVID-19 and dengue, dengue appearing at first glance more symptomatic and with a more abrupt onset than COVID-19 or OFIs in the setting of a SARS-CoV-2 testing center.

These discrepancies might reflect first a selection bias, the more symptomatic cases of both infections having been referred primarily to the emergency units, these redirecting the COVID-19 cases towards the Saint-Denis referral hospital for quarantine purpose. This could be arguably deduced from weighing on the inverse probability of hospitalization, which was on average 2.5-fold higher than from the UDACS, all through the study period. Doing so abrogates, for instances, the effects of a delayed presentation and the protection of active smoking for the prediction of COVID-19 (data not shown). Together with the fact that the dengue epidemic was more active in the southern part of the island, this fuels the idea that time to presentation in our study partly stemmed from differences in recruitment driven both by the organization and access to care. Importantly, weighing the analysis strengthened also the odds ratios of a contact with a COVID-19+ case for the same, as well as those of headache and retro-orbital pain for the prediction of dengue. These elements suggest that this putative selection bias was more pronounced on epidemiological than on clinical variables.

Second, our results might also be affected by a misclassification bias, which may arise both from the poor sensitivity of SARS-CoV-2 RT-PCR and rapid NS1 antigen, rather than from the false positive rates.

Third, given the fear of COVID-19 at that time, we cannot rule out the possibility of a reporting bias, as some patients may have declared URTI symptoms or cough in excess just to be tested for SARS-CoV-2. Consistent with this, are the relatively high percentages of cough and URTI symptoms among dengue cases, as well as the totality of anxiety cases upon testing observed within the OFI group (data not shown), for instances.

Together with the abovementioned sources of bias, a lack of power might have reduced the capability to shed light on other discriminating factors. However, we believe that this study faithfully reflects the real epidemiological situation on Reunion island at that time, given diagnostic practices and means that were commonly used in this era of uncertainty, which is unlikely to have biased the overall sense of our results.

This being said, our findings are also in agreement with the literature.

First, the fact that dengue was more symptomatic than COVID-19 fulfills both the concept of “force of infection” and the trade-off model according to which, the time spent in the susceptible group to an infectious disease is inversely correlated to its incidence [11], and the virulence (i.e., ability to cause illness, lethality) grows with the transmission rate until it reaches a plateau [12]. Consistent with these assumptions, according to Santé Publique France reports, the attack rate observed over the study period was 22-fold higher for dengue (≈905 per 100.000 inhabitants) than for COVID-19 (≈41 per 100.000 inhabitants). This was explained by the recent introduction of DENV-1 serotype (March 2019) complicating five years of DENV-2 circulation [7], cases of secondary dengue, the effectiveness of the lockdown to slow the progression of COVID-19 and the fact that SARS-CoV-2 impacted at that time mainly “healthy” individuals (travelers and their relatives). In this framework, the relevance of body ache, headache and retro-orbital pain at presentation for the differential diagnosis between COVID-19 and dengue accounts for the involvement of dengue in the general and digestive spheres, as proposed by Nacher et al. in a recent opinion paper, COVID-19 being more pronounced in the respiratory sphere [10]. Interestingly, we also found one COVID-19+ case who was tested negative for dengue suffering retro-orbital pain, as previously reported in Taiwan [13].

Second, our cohort study supports the high positive predictive values and specificities of the contact with a COVID-19+ case and anosmia for the diagnostic of COVID-19, which is congruent with risk prediction models developed for healthcare workers in Italy [14] and findings from the Coranosmia cohort study in France [15], respectively.

Together with the abovementioned putative selection bias, the delayed presentation to hospital of COVID-19 cases, as compared to dengue, might also illustrate the mild (“pauci-symptomatic”) character of COVID-19 illness during the first pandemic surge on Reunion island, as well as some consecutive lags in contact tracing. Overall, the individuals who did not feel or only slightly sick with COVID-19 might not have felt the need to be tested.

However, this hypothesis does not stand the absence of clear association between the proportions of asymptomatic infections and time to presentation, which was observed upstream to the analysis (i.e., with 3-fold more asymptomatic cases, OFIs cases presented faster than COVID-19 cases; data not shown).

Interestingly, active smoking was less likely to be observed with COVID-19 as compared to OFIs or dengue, but this apparent protective effect was not robust as suggested above. Moreover, it was not replicated for asymptomatic SARS-CoV-2 infections, nor was it shown to protect from contracting illness with COVID-19 (data not shown). This finding seems paradoxical given recent evidence shows that Angiotensin-converting enzyme 2 (ACE2), the SARS-CoV-2 entry receptor, is overexpressed in smoker’s bronchial and alveolar epithelia, which should increase the risk of infection [16-18]. Whether this finding results from abovementioned misclassification or reporting bias deserves further studies. Notwithstanding, this fuels the smoker’s paradox according to which active smokers were first underreported among the patients hospitalized for COVID-19 in several countries [19], then less likely to be infected, as shown from a recent meta-analysis [20].

In conclusion, our cohort study identified several factors distinguishing non severe dengue from COVID-19 at clinical presentation in a context of recent dengue endemicity and first introduction of SARS-CoV-2. Although prone to potential biases, these data suggest that non severe dengue may be more symptomatic at presentation than COVID-19 in a co-epidemic setting with higher dengue attack rates, a pattern that might also result from different forces of infection (lesser exposure to SARS-CoV-2 than to DENV). Whether these findings may serve other regions facing co-epidemics, deserves more investigations, development, and validation of more accurate diagnostic tools. These findings highlight also the need not to jeopardize dengue control wherever COVID-19 dengue co-epidemics have the potential to wrought havoc to the healthcare system [21].