Background

Under controlled conditions and in some observational studies of runners, airborne fine particulate matter smaller than 2.5 microns in diameter (PM_2.5) is associated with exercise performance decrements.

Objective

To assess the association between event-day fine particulate matter air pollution (PM_2.5) and marathon finish times.

Methods

Using a spatiotemporal machine-learning model, we estimated event-day racecourse-averaged PM_2.5 concentrations for nine major US marathons (2003–2019). We obtained 1,506,137 male and 1,058,674 female finish times from 140 event-years of public marathon data. We used linear and quantile mixed models to estimate the mean and percentile-specific year and heat index-adjusted effect of 1 µg/m³ higher event-day racecourse-averaged PM_2.5 on marathon finish times in sex-stratified samples.

Results

Analyzing all finish times, 1 µg/m³ higher race-day PM_2.5 was associated with 32-s slower average finish times among men (95% confidence limits (CL) 30, 33 s) and 25-s slower average finish times among women (95% CL 23, 27 s). Quantile-specific associations of event-day PM_2.5 with finish times were larger for faster-than-median finishers. While PM_2.5 was generally associated with slower finish times in single-event models, there was effect heterogeneity, and most 95% confidence intervals included the null.

Conclusion

Greater race-day PM_2.5 was associated with slower average marathon finish times, with more pronounced effects in faster-than-median runners. While more research is needed to characterize effect heterogeneity across the performance spectrum, these findings show the impact of PM_2.5 on marathon performance and the importance of considering data from multiple competitions when estimating PM_2.5 effects from event-level data.

中文翻译：

---

在烟雾中奔跑：2003-2019 年美国九场主要马拉松比赛细颗粒物对终点时间的影响分析

 背景

在受控条件下和一些跑步者的观察性研究中，空气中直径小于 2.5 微米的细颗粒物 （PM_2.5） 与运动表现下降有关。

 目的

评估活动当天细颗粒物空气污染 （PM_2.5） 与马拉松结束时间之间的关联。

 方法

使用时空机器学习模型，我们估计了美国九场主要马拉松比赛（2003-2019 年）的赛事日赛马场平均 PM_2.5 浓度。我们从 140 个赛事年的公开马拉松数据中获得了 1,506,137 名男性和 1,058,674 名女性的完成时间。我们使用线性和分位数混合模型来估计 1 μg/m 3 高出³ 个赛事日赛马场平均 PM_2.5 对性别分层样本中马拉松完成时间的平均值和百分位数特定年份和热指数调整后的影响。

 结果

分析所有完成时间，比赛日 PM_2.5 高¹ μg/m 3 与男性平均完成时间慢 32 秒相关（95% 置信限 （CL） 30、33 s）和女性平均完成时间慢 25 秒（95% CL 23、27 s）。对于快于中位数完成者的事件日 PM_2.5 与完成时间的分位数特异性关联更大。虽然 PM_2.5 在单事件模型中通常与较慢的完成时间相关，但存在效应异质性，大多数 95% 置信区间包括零。

 结论

比赛日 PM_2.5 较高与马拉松平均完成时间较慢相关，对快于中位数的跑步者的影响更明显。虽然需要更多的研究来描述整个表现谱中的效应异质性，但这些发现显示了 PM_2.5 对马拉松表现的影响，以及在根据赛事层面数据估计 PM_2.5 效应时考虑多项比赛数据的重要性。