RESUMO
INTRODUCTION: Ongoing controversy surrounds the role of atmospheric temperature in the incidence of intravaginal testicular torsion (iTT). This debate may be attributed to inadequate research methodology. As environmental risk factors have been successfully investigated with distributed lag non-linear model regression (DLNM), we applied this methodology to investigate the association between daily mean atmospheric temperatures (Tmean) and daily incidences of intravaginal testicular torsion (iTT) in our region. STUDY DESIGN: We analyzed time series consisting of the daily incidences of surgically confirmed iTT according to Tmean, in a circumscribed region in central Brazil from 2012 to 2015, with non-parametric tests, unadjusted and seasonally and long-term trend adjusted time series regression, as well as with DLNM. RESULTS: We recovered 218 cases of iTT in 1125 days of study. Most patients were teenagers (median 15.8 years, interquartile range 14.1-18.5 years). Within the 188 days with events, a single event was recorded on 161 days, two events were recorded on 24 days, and three events were recorded on 3 days. Tmean was lower in days with iTT compared with days without iTT (median 21.4 °C vs. 20.9 °C, p = 0.0002). We found decreasing magnitude and uncertainty of the direction of the effect of Tmean as a risk factor for iTT as the time series regression model was adjusted for seasonal and long-term trends. DLNM indicated a more complex exposure-response relation, with a proportional increase in risk when Tmean fell below 19.4 °C at the day of exposure (for 18.0 °C, RR 4.35) and a protective effect, for similar temperatures, after 1-2 and 7-12 days of exposure (RR 0.44 and 0.78, respectively). DISCUSSION: The association between lower Tmean and higher incidences of iTT at first observed with conventional non-parametric tests and unadjusted time series regression disappeared with adjusted time series regression models, reproducing the conflicting results of the literature. In contrast, DLNM revealed both a proportional effect of Tmean with decreasing temperatures and a delayed decrease in risk, suggesting a harvesting effect, seen when the pool of susceptible patients is depleted at exposure leading to a subsequent decrease in the incidence of the disease. CONCLUSION: According to DLMN, exposures to lower Tmean were associated with immediate greater risk and delayed reduction in risk for iTT. This pattern, indicating a harvesting effect, strongly argues that low temperatures do constitute a risk factor for iTT.