Table 1.

Summary of the body of evidence for high temperatures and heatwaves as risk factors of dengue infection.

High temperature studies (n = 102)			Heatwave studies (n = 5)
Quality of evidence assessment
Down grading factors
	Rating	Rationale and summary of evidence	Rating	Rationale and summary of evidence
Risk of bias	0	Only four studies were rated as “definitely high” or “probably high”. The remaining studies (98/103) were rated as “definitely low” and “probably low” risk of bias.	0	All five studies were rated as “definitely low” and “probably low” risk of bias. Thus, we judged there is no substantial risk of bias
Indirectness	0	Dengue cases included in most studies were either laboratory confirmed or clinically diagnosed according to WHO definition of dengue fever. Moreover, the majority of the studies employed direct measurements of exposure.	0	Dengue cases included in most studies were either laboratory confirmed or clinically diagnosed according to WHO definition of dengue fever. Moreover, the majority of the studies employed direct measurements of exposure.
Inconsistency	0	High heterogeneity was observed in the overall pooled estimate (I2 = 99%). However, 80% PI did not contain unity (Appendix pp 50–49). Moreover, sensitivity analysis based on the “leave-one-out” approach did not show considerable difference from overall pooled effect estimates (RR: 1.12–1.15), and there was no change in statistical significance.	−1	The heterogeneity was high (I2 = 89%) and 95% prediction intervals (PIs) contain unity and were wider compared to confidence intervals (Fig. 4). Leave-one-out analysis also showed variation (RR: 1.06–1.16). Thus, the evidence of quality was downgraded for such inconsistencies.
Imprecision	0	Except four studies, the 95% CIs of the studies possessed notably narrow confidence intervals. Moreover, we judge most studies included representative of appropriate proportion of population of interest across different years of time.	0	We judge all five studies included representative of appropriate proportion of population of interest across different years of time. Only two effect (2/13) estimates possessed wider confidence intervals.
Publication bias	0	Despite the observed asymmetrical distribution of studies in the funnel plot and Egger's tests (Appendix pp 56), the pooled effect estimate adjusted after including imputed studies using the Trim and Fill method (Appendix pp 59) did not show difference in statistical significance.	0	Egger's test revealed statistically insignificant result that did not show publication bias (p = 0.156). However, the small number of studies (n <10), makes it uncertain and needs careful interpretation.
Upgrading factors
Large magnitude of effect	0	Effect sizes were small in most studies, and the pooled effect estimate was below 2.	0	Effect sizes were small in most studies, and the pooled effect estimate was below 2.
Dose response pattern	+1	The quality of the evidence was upgraded as majority of the studies indicated a statistically significant increase in the risk of dengue infection with an increase in temperatures above reference values.	−1	Two of the five studies showed a decrease in effect estimate and the random effect contains unity and were statistically insignificant.
Confounding minimizes effect	0	One fourth (26/102) of the studies did not adjust for primary confounders. However, the sensitivity analysis did not show evidence to suggest that possible residual confounders would shift the effect to null.	0	Only one of the studies did not adjust for primary confounders. However, such did not suggest the possible residual confounders would shift the effect to null.
Summary of the quality assessment
Quality of evidence	High	Moderate + (0) + (0) + (0) + (1) = 1. After one upgrade and no down grade in quality factors, the overall quality of evidence was rated as “High”.	Low	Moderate + (0) + (0) + (-1) + (−1) = −2. After down grading two quality factors, the overall quality of evidence was rated as “low”
Summary findings	n/a	Overall, the studies included in the meta-analysis and narrative synthesis resulted in “High” quality of the evidence.	n/a	Overall, the studies included in the meta-analysis resulted in “Low” quality of the evidence.
Strength of evidence assessment
Quality of evidence	n/a	High	n/a	Low
Direction of effect estimates	n/a	The direction of effect estimates largely showed an increasing trend in dengue infection with increasing high temperatures beyond the reference values.	n/a	The direction of effect estimates largely showed an increasing trend in dengue infection associated with heatwave, but the inclusion of unity in the confidence interval makes the result uncertain.
Confidence in effect estimate	n/a	Majority of the studies showed a positive association between dengue infection and increased temperature. Moreover, the prediction interval did not include the null effect and supported the positive association between dengue incidence and temperature in the future studies. The number of studies included in the meta-analysis was also sufficient enough and provide certainty about the future.	n/a	Although an increasing trend was observed in the overall random effect the confidence and prediction intervals contain the null effect. Moreover, the number of studies included in the meta-analysis was small and made the result uncertainty about the future.
Other characteristics of the data that may influence certainty	n/a	Sensitivity analysis based on several factors did not show much variation from the main overall effect estimate and strengthen the certainty of the evidence.	n/a	The leave-one-out sensitivity analysis showed variation from the main overall effect estimates (Fig. S9). The small number of studies included hindered performing certain sensitivity analyses in addition to leave-one-out analysis and increase the uncertainty.
Overall strength of evidence	Sufficient	The quality of evidence was rated as “High”, the prediction interval suggested future studies may not affect the result and several sensitivity analyses showed similar overall estimates. Thus, we concluded that the evidence is sufficient enough to show the effect of high temperature on dengue infection.	Limited	Although a positive association is observed, the “Low” quality of evidence, inclusion of unity in confidence interval and the prediction interval, and the small number of studies included in the meta-analysis resulted a “Limited” evidence to conclude the association between dengue infection and heatwave.