Daylight Saving Time Transitions and Risk of Heart Attack

A Systematic Review and Meta-Analysis

Background: The health risks of daylight saving time transitions are intensely debated. Disturbed circadian rhythms and lack of sleep after transitions might increase the risk of acute myocardial infarction (AMI). The only meta-analysis on the risk of AMI has now been considerably expanded.

Methods: In this systematic review and meta-analysis (including meta-regressions and sensitivity analyses), we examine the frequency of AMI in the first few weeks after daylight saving time transitions (OSF registration www.doi.org/10.17605/OSF.IO/7CFKS). Eight databases were searched for pertinent literature up to September 2023. Authors were contacted for additional information. Study quality was rated using the Newcastle–Ottawa Scale.

Results: Twelve studies from ten countries were included in the meta-analysis. Nine were of adequate quality, and three were of good quality. The pooled relative risk (RR) of AMI after daylight saving time onset (spring) was 1.04 (95% confidence interval [1.02; 1.07], I²: 57.3%), and 1.02 ([0.99; 1.05], I²: 51.6%) after daylight saving time offset (autumn). Recalculation after the exclusion of one study with inconsistencies yielded pooled RR values of 1.04 [1.01; 1.06] and 1.00 [0.99; 1.02], with the spring results being heterogeneous (I²: 56.9%) and the autumn results homogeneous (I²: 17.1%).

Conclusion: Current evidence suggests that there may be an increased risk of AMI after the spring transition, although there is moderate to marked heterogeneity among the studies that support this conclusion. More easily interpretable studies, such as those already conducted in the field of economics, should clarify associations with the aid of discontinuity regression and placebo tests. To this end, comparative risk analyses using years or places wherein daylight saving time was not practiced would be suitable.

The statutory time transitions, i.e. the onset of daylight saving time (DST) in spring and the offset back to standard time in autumn, are the subject of much debate since it seems plausible that they are associated with adverse health effects (1, 2). Such time transitions can result in the disruption of circadian rhythms and have a negative effect on, for example, sleep duration (3). As they are not controlled by research, daylight saving time transitions provide the conditions for a natural experiment (4) of potentially enormous statistical power, given the exposure of populations in about 70 countries (5). Due to DST transitions, a significant part of the world’s population is exposed to light, work or other activities as well as meals at biologically unanticipated times.

The health effects of DST transitions have been the subject of many studies. One meta-analysis (6) reviewed the total available evidence based on seven epidemiological studies and found an increased risk of acute myocardial infarction (AMI) following the daylight saving time transition in spring, but not in autumn. A study published in the New England Journal of Medicine in 2008 sparked interest in the question of whether there is an association between DST and AMI (7). It found a significant increase in AMI risk in Sweden (incidence 1.051 [1.032; 1.071]) after transition to DST. Possible explanations for the observations included a lack of sleep with increased activity of the sympathetic nervous system and elevated levels of pro-inflammatory cytokines (7). In discussions about the abolition of daylight saving time, these results are cited as an example of health effects (1, 3) associated with the transition to DST in spring.

The aim of this systematic review was to present the currently available evidence on associations between DST and AMI. The following research questions were posed:

• Is the evidence from epidemiological studies consistently showing an association between DST and AMI and what overall effects can be estimated?
• What are the factors that could be responsible for differences between individual studies?

Methods

Our systematic review extends the only meta-analysis on DST and adverse health effects (6) to find answers to the research questions (OSF-registered: www. doi. org/10. 17605/OSF.IO/7CFKS – 14 August 2023; registered during the project).

Systematic review—PICO framework

Regarding the PICO (population, intervention, comparison, and outcome) framework we did not impose any inclusion restrictions on the population. The intervention investigated was DST transitions. The outcome was the incidence of and/or mortality due to AMI. The frequencies of AMI during the index and reference periods were compared. While the first week after daylight saving time transitions was commonly used as the index period, some studies used longer index periods (eSupplement). The reference periods (comparison periods during which no DST effect is assumed) vary too.

Search strategies

The systematic electronic search for pertinent literature was initially conducted in 2020 in the databases PubMed, Web of Knowledge/Web of Science, WHOLIS, Cochrane Library, Open Grey, and from 2022 additionally in the databases Econlit and SCOPUS (and updated on 1 September 2023 based on a search in PubMed, Web of Science, Econlit, and Scopus). The Google Scholar search engine was also used. The electronic literature searches were supplemented by manual searches in the references of included publications and in reviews. Additional information about the search strategy (for example, search strings) is provided in eTable 1.

Table

13 individual studies on daylight saving time transitions and acute myocardial infarction

Inclusion criteria

Studies were included if they reported week and/or day estimates of the risk of AMI before and after the annual DST transitions and provided sufficient data to allow comparisons of AMI incidences and/or AMI mortality. The registration of the various studies was not an inclusion criterion.

Data extraction

PICO-relevant information as well as special features and limitations of the studies were extracted by at least two authors (in particular AH and PL) in consultation with additional authors (in particular PM). If the information was incomplete or if there were any questions, the authors of the included publications were approached for additional information or clarification.

Quality assessment

Three authors (AH, PL and PM) assessed the quality of the included studies using a Newcastle-Ottawa scale (NOS) that was modified for “natural” experiments to rate the quality of cohort studies based on the following three dimension: selection, comparison, and outcome (eSupplement and eTable 2). Additionally, the team of authors critically evaluated individual studies, with a special focus on the consistency of the reported results.

Meta-analysis

Estimates of the relative risk (RR) of AMI after DST transitions for the respective populations and the corresponding 95% confidence intervals (CIs) were extracted from the individual studies (taking into account that seasons are reversed in the northern and southern hemispheres). Missing information was obtained from other information in the publications; for example, results reported as daily values were converted to weekly values. Meta-RRs with 95% CIs (8) for the spring transition and for the autumn transition were calculated using random effects (RE) models. In addition, heterogeneity was determined (I²: the percentage of variation explained by heterogeneity, and p_het: the heterogeneity p-value using Cochran’s Q statistics)(9). Egger’s test (10) was used to address possible small study publication bias, and the trim-and-fill method was used to test for funnel plot asymmetry (11).

Random-effects meta-regression models were fitted using restricted maximum likelihood estimation (8). The predetermined co-variables were:

(a) Country/continent where the study was conducted, representative of population- and location-related differences

(b) Time of sunrise at the study location, and

(c) Combinations of (a) and (b).

We performed sensitivity analyses for suitable study groups and recalculations of the only existing meta-analysis by Manfredini et al. (2019) based on corrected estimates. The statistical software package Stata (12) was used for all analyses.

Results

Selection process of the literature search—individual studies

Figure 1 shows a PRISMA flow diagram (PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses) of the literature search. Of the 15 publications included, twelve studies were identified as suitable for our meta-analysis (eTable 1). The three studies not included in the meta-analysis were unsuitable for the following reasons:

Figure 1

PRISMA flow diagram of the systematic literature search and the study selection process

• From two studies, no comparable AMI RRs could be derived (13, 14).
• One study by Janszky et al. (15) was excluded because of data overlapping with the study by Janszky and Ljung (2008) (7).

The Table provides an overview of region and latitude, study years as well as RRs and CIs for spring and autumn. Further information about participants, index and reference periods as well as special features/limitations of the 15 publications is provided in eTable 1.

The twelve studies evaluated in our main analysis were conducted in ten countries (Europe: Germany, Finland, Croatia, Netherlands, Sweden, Spain; outside Europe: USA, Brazil, Iran, Mexico) between 2008 and 2023. Of these studies, nine were from the field of health science and three from the field of economics (16, 17, 18); observation periods ranged between one (19) and 26 years (20). The studies included in the analysis investigated diverse endpoints based on data from a variety of sources: There were three hospital studies (19, 21, 22) and five hospital registry studies (17, 23, 24, 25, 26). Two studies used data from population-based morbidity registries to determine both hospital cases and out-of-hospital deaths (7, 20); two mortality studies analyzed data from official death registries (16, 18). The studies differed with regard to the selected index periods (always immediately after a daylight saving time transition) and the reference periods (in some cases shortly before the index period, in other cases including the periods before and after the index period).

Critical assessment of methods

The assessment of study quality using the adapted Newcastle-Ottawa Scale (NOS) revealed quality-relevant differences between the studies (for details refer to eTable 2). These differences were related to:

• the representativeness of the cases included;
• the time point used as the time of manifestation of AMI; and
• the index and reference periods selected.

Nine individual studies were found to be of adequate quality and three studies of good quality (16, 17, 18). Some published study estimates were corrected (e.g., following correspondence with the authors); comparisons of original estimates with those used here are provided in eTable 2 .

Meta-analysis and sensitivity analyses

Our meta-analysis provided results on risks of AMI due to DST onset (spring) and offset (autumn) based on all twelve studies included. For the spring transition, a pooled RR of 1.044 (95% CI: [1.015; 1.073], p = 0.003) was determined, with significant heterogeneity in the individual results (I²: 57.3%, p = 0.007; Figure 2).

Figure 2

Forest plots of relative risk (RR) and 95% confidence interval (CI) from individual studies, the main meta-analysis and a meta-analysis without the study by Culic (2013) for spring transition

The sensitivity analyses showed that the heterogeneity was not attributable to the two structurally different studies by Sandhu et al. (23) and Mofidi et al. (19), but apparently due to the heterogeneity between individual studies, such as Janszky and Ljung (2008) (7), Čulić (22) and Rodríguez-Cortés et al. (25) (eTable 3). For the autumn transition, the meta-analysis yielded an RR of 1.018 ([0.989; 1.048], p = 0.220; I²: 56.9%) which was thus not statistically significant (Figure 3).

Figure 3

Forest plots of relative risk (RR) and 95% confidence interval (CI) from individual studies, the main meta-analysis and a meta-analysis without the study by Culic (2013) for autumn transition

Together with Čulić (22) we identified inconsistencies in his study. These were calculation errors and biased results, not all of which could be corrected by Čulić (eTable 1, eTable 2). Without the study by Čulić, the meta-analysis yielded lower RRs and considerably lower heterogeneity (spring: Figure 2, I²: 51.6 %; autumn: Figure 3, I²: 17.1%). In re-analyses of the 2019 meta-analysis by Manfredini et al. (6) and in sensitivity analyses, almost no heterogeneity was found after exclusion of the study by Čulić. All analysis results are listed in eTable 3.

For both the spring and autumn transitions, there was little evidence of bias due to the small-study effect, i.e. publication bias (eFigure). None of the many meta-regression models found a significant effect of the investigated co-variables, whether in combination or alone. No differences were found for age and sex, but the dataset was limited (data not shown, but available upon request from the authors). The random effects estimates were comparable to the fixed effects estimates in all analyses.

Discussion

After comprehensive searches for pertinent literature, twelve studies were included in our meta-analysis on the risk of AMI after the DST transitions in spring and autumn in ten countries worldwide. This significantly expanded the dataset compared to the only meta-analysis performed in this subject area so far, which was conducted in 2019 (6). While one of the seven studies of the previous meta-analysis was excluded due to overlapping data, an additional six studies were newly included to help answer the research questions.

Consistency and overall effect estimates

Overall, the results of the twelve included studies showed an increased risk of AMI after DST onset in spring, but no change in risk after the transition in autumn (27). The effect of the divergent study by Čulić (2013) (22) was further examined in sensitivity analyses. Publication bias did not appear to be a relevant issue. Indeed, the epidemiological data obtained since 2008 provided further evidence to support the observations by Janszky and Ljung (7) of an increase in AMI risk after the spring transition.

Heterogeneity and generalizability

The pool of all twelve included studies was heterogeneous. After exclusion of the study by Čulić (22), our results for the DST transition in spring were less heterogeneous and those for the autumn transition were homogeneous. Consequently, there are fewer statistical concerns about generalizing the risk result of 1.00 for the autumn transition across the eleven pooled studies. For the co-variables analyzed, no relevant effects were identified. Neither the region in which the study was conducted, nor the time of sunrise at the study location caused differences in the effects of the spring transition. Similarly, the exclusion of studies due to study design characteristics did not provide an explanation for the heterogeneity observed. Latitude could not be used as an additional co-variable due to collinearity.

Comparison with the 2019 meta-analysis

The two studies by Janszky et al. (2008, 2012) contributed significantly to the results of the 2019 meta-analysis (combined weights of 65.3% and 57.8% to the spring and autumn transitions, respectively). We excluded the study by Janszky (2012) from our meta-analysis due to overlap with the dataset of the study by Janszky and Ljung (2008) (7, 15). Our current meta-analysis included a significantly broader study basis of twelve studies in ten countries—thus comprising six additional epidemiological studies—and, in addition, study data were corrected. Moreover, inconsistencies in the results of the study by Čulić (22) were identified; consequently, this study was excluded in sensitivity analyses. Remarkably, the re-analyzed 2019 meta-analysis with six studies showed no heterogeneity and in our 2023 meta-analysis with eleven studies we found far less heterogeneity. Hence, our current results provide a more robust confirmation of the results published in 2019 (6) on myocardial infarction after daylight saving time transitions in spring and autumn.

Strengths and limitations of this systematic review

What sets this systematic review apart from an epidemiological perspective is its very comprehensive systematic search for pertinent literature in a wide range of different databases, including those with economic content, such as SCOPUS and Econlit. As a result, our review offers the most comprehensive dataset in comparison to other reviews on this topic (6, 28, 29). In particular, the identification of the high-quality economic study by Toro et al. (2015) (16) is worth mentioning, as it was not included in the 2019 meta-analysis or in any of the other reviews. Studies with an economic focus on DST, such as the studies by Toro et al. (16), Tanaka and Koizumi (17, and Goodwin et al. (18), are not listed in medical databases, such as MEDLINE. Another strength of our review is the thorough critical evaluation of the individual studies by the team of authors, which resulted in the identification of inconsistencies in some study results.

From a chronobiological perspective, one possible limitation could be that postulated long-term effects over several months were not investigated (28). The assumption that there are effects on the risk of AMI which extend beyond the comparatively short observation period of one week is supported by studies that found longer index periods after DST transitions to be relevant for the respective populations studied (16, 17, 18).

Future studies with sufficient informative value

The studies by Toro et al. (16), Tanaka and Koizumi (17) as well as Goodwin et al. (18)can serve as a model in terms of their methods. They use discontinuity regression models specifically designed to assess the effects of abruptly changing exposures. Stata, for example, is one of the statistical software packages that offer these methods (12). Such analyses are also useful with regard to the index periods examined: In contrast to the other included studies, the index and reference periods did not have to be predefined, but were determined from the data, using statistical methods. These time windows, identified as optimal, were consistently longer (up to four weeks) than the arbitrarily set index periods of usually only one week in the other studies. In addition, these studies looked more closely at possible causal relationships by performing comparative discontinuity regressions at the time points of DST transitions (for example, last Sunday in March) in consecutive years (17) or neighboring locations (16, 17, 18) without DST transitions (so-called placebo tests), i.e. under conditions where no effects from time transitions could possibly have occurred. Placebo tests were also performed within a few weeks after the date of DST transitions (16, 18). With this approach, potential residual confounding, e.g. due to seasonal effects, can be identified and taken into consideration. In addition, Toro et al. (16) looked at neoplasia, viral infections and parasitic diseases, i.e. conditions for which no evidence is available to suggest that the risk of dying from any of these conditions is changed by DST transitions. In contrast to the actual spring transition, none of these studies provided any indication of a sudden increase in risk. So far, this type of study approach has not been pursued in medical research.

Another geographic region of interest for DST studies could be Australia, where some territories implement DST transitions, while others do not (30). In addition, data of regions could be analyzed where daylight saving time has been newly introduced or abolished.

Conclusion

The results of this review with meta-analysis based on studies with mostly adequate and in some cases good quality show an increase in the risk of AMI after DST onset in spring with some moderate to marked heterogeneity between the individual studies. After excluding one study due to inconsistent results (22), no heterogeneity between studies and overall no evidence of a change in risk was found for the autumn transition.

Following the spring transition, people may experience a sleep deficit with abrupt changes in biological rhythms, but not after the autumn transition which provides more time to sleep (7). The existing studies should be supplemented by studies of better methodological quality.

Acknowledgement

The authors would like to thank the study authors Čulić, Goodwin, Tanaka, Tigre, Lyons, and Forbes for further information on their publications as well as Dr. Valérie Groß, who initiated this project and supported the 2020 searches for pertinent literature. This study is part of the doctoral thesis of AH. TCE would also like to acknowledge the stimulating working conditions he enjoyed as a visiting scholar at UC Berkeley in 2023.

Conflict of interest statement
The authors declare no conflict of interest.

Manuscript received on 20 October 2023, revised version accepted on 16 April 2024.

Translated from the original German by Ralf Thoene, M.D.

Corresponding author
Philip Lewis, PhD MPhil

Institut und Poliklinik für Arbeitsmedizin, Umweltmedizin und Präventionsforschung

Universitätsklinikum Köln, Kerpener Straße 62, 50937 Köln, Germany

Cite this as:
Hurst A, Morfeld P, Lewis P, Erren TC: Daylight saving time transitions and risk of heart attack—a systematic review and meta-analysis. Dtsch Arztebl Int 2024; 121: 490–6. DOI: 10.3238/arztebl.m2024.0078

References

1. Roenneberg T, Winnebeck EC, Klerman EB: Daylight saving time and artificial time zones—a battle between biological and social times. Front Physiol 2019; 10: 944

2. afp/aerzteblatt.de: Ein Viertel klagt über gesundheitliche Beschwerden nach Zeitumstellung. 23. März 2023. www.aerzteblatt.de/nachrichten/141951/ (last accessed on 12 August 2023).

3. Roenneberg T, Wirz-Justice A, Skene DJ, et al.: Why should we abolish daylight saving time? J Biol Rhythms 2019; 34: 227–30

4. Erren TC, Lewis P, Shaw DM: The COVID-19 pandemic: ethical and scientific imperatives for „natural“ experiments. Circulation 2020; 142: 309–11

5. Craig P, Cooper C, Gunnell D, et al.: Using natural experiments to evaluate population health interventions: new Medical Research Council guidance. J Epidemiol Community Health 2012; 66: 1182–6

6. Manfredini R, Fabbian F, Cappadona R, et al.: Daylight saving time and acute myocardial infarction: a meta-analysis. J Clin Med 2019; 8: 404

7. Janszky I, Ljung R: Shifts to and from daylight saving time and incidence of myocardial infarction. N Engl J Med 2008; 359: 1966–8

8. Sutton AJ, Abrams, KR, Jones DR, Sheldon TA, Song, F: Methods for meta-analysis in medical research; 2000: Wiley, Chichester, U.K.

9. Sterne, JAC, Newton HJ, Cox NJ: Meta-analysis in Stata: an updated collection from the Stata Journal. 2009: Stata Press: College Station, Texas.

10. Egger M, Smith GD, Schneider M, Minder C: Bias in meta-analysis detected by a simple, graphical test. BMJ 1997; 315: 629–34

11. Duval S, Tweedie R: Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis. Biometrics 2000; 56: 455–63

12. StataCorp: Stata Statistical Software: release 14. Texas, USA: StataCorp LP, College Station 2015.

13. Lyons S, Forbes KF, Brick A: Revisiting the link between daylight savings time and acute myocardial infarction. SSRN Electronic Journal 2023; DOI:10.2139/ssrn.4329776

14. Jin L, Ziebarth NR: Sleep, health, and human capital: evidence from daylight saving time. Journal of Economic Behavior & Organization 2020; 170: 174–92

15. Janszky I, Ahnve S, Ljung R, et al.: Daylight saving time shifts and incidence of acute myocardial infarction—Swedish register of information and knowledge about swedish heart intensive care admissions (RIKS-HIA). Sleep Med 2012; 13: 237–42

16. Toro W, Tigre R, Sampaio B: Daylight saving time and incidence of myocardial infarction: evidence from a regression discontinuity design. Economics Letters 2015; 136: 1–4

17. Tanaka S, Koizumi H: Springing forward and falling back on health: the effect of daylight saving time on acute myocardial infarction. medRxiv 2022: doi: https://doi.org/10.1101/2022.07.06.22277274

18. Goodwin MB, Gonzalez F, Fontenla M: The impact of daylight saving time in Mexico. Applied Economics 2024; 56: 22–32

19. Mofidi M, Kianmehr N, Foroghi Qomi Y, et al.: Daylight saving time and incidence ratio of acute myocardial infarction among Iranian people. J Med Life 2019; 12: 123–7

20. Kirchberger I, Wolf K, Heier M, et al.: Are daylight saving time transitions associated with changes in myocardial infarction incidence? Results from the German MONICA/KORA Myocardial Infarction Registry. BMC Public Health 2015; 15: 778

21. Jiddou MR, Pica M, Boura J, Qu L, Franklin BA: Incidence of myocardial infarction with shifts to and from daylight savings time. Am J Cardiol 2013; 111: 631–5

22. Čulić V: Daylight saving time transitions and acute myocardial infarction. Chronobiol Int 2013; 30: 662–8

23. Sandhu A, Seth M, Gurm HS: Daylight savings time and myocardial infarction. Open Heart 2014; 1: e000019

24. Sipilä JOT, Rautava P, Kytö V: Association of daylight saving time transitions with incidence and in-hospital mortality of myocardial infarction in Finland. Ann Med 2016; 48: 10–6

25. Rodríguez-Cortés FJ, Jiménez-Hornero JE, Alcalá-Diaz JF, et al.: Daylight saving time transitions and cardiovascular disease in Andalusia: time series modeling and analysis using visibility graphs. Angiology 2023; 74: 868–75

26. Derks L, Houterman S, Geuzebroek GSC, van der Harst P, Smits PC, PCI Registration Committee of the Netherlands Heart Registration: Daylight saving time does not seem to be associated with number of percutaneous coronary interventions for acute myocardial infarction in the Netherlands. Neth Heart J 2021; 29: 427–32

27. Rothman KJ: Significance questing. Ann Intern Med 1986; 105: 445–7

28. Čulić V, Kantermann T: Acute myocardial infarction and daylight saving time transitions: is there a risk? Clocks Sleep 2021; 3: 547–57

29. Fansa A, Fietze I, Penzel T, Herberger S: Does daylight saving time lead to more myocardial infarctions? Somnologie 2023; 27: 233–40

30. Ellis WA, FitzGibbon SI, Barth BJ, et al.: Daylight saving time can decrease the frequency of wildlife-vehicle collisions. Biol Lett 2016; 12: 20160632