Lunar cycle effects on sleep and the file drawer problem

Abstract

Popular beliefs about the influence of the full moon on humans exist, although no solid evidence has so far confirmed these ideas [1Foster R.G. Roenneberg T. Human responses to the geophysical daily, annual and lunar cycles.Curr. Biol. 2008; 18: 784-794Abstract Full Text Full Text PDF PubMed Scopus (240) Google Scholar]. Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] recently presented fascinating data on lunar cycle effects on human sleep. However, in a re-analysis of sleep electroencephalography (EEG) data in three large samples, we were unable to replicate their findings. In addition, we identified further mostly unpublished null findings, suggesting that the conflicting results might be an example of a publication bias (i.e., the file drawer problem). Popular beliefs about the influence of the full moon on humans exist, although no solid evidence has so far confirmed these ideas [1Foster R.G. Roenneberg T. Human responses to the geophysical daily, annual and lunar cycles.Curr. Biol. 2008; 18: 784-794Abstract Full Text Full Text PDF PubMed Scopus (240) Google Scholar]. Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] recently presented fascinating data on lunar cycle effects on human sleep. However, in a re-analysis of sleep electroencephalography (EEG) data in three large samples, we were unable to replicate their findings. In addition, we identified further mostly unpublished null findings, suggesting that the conflicting results might be an example of a publication bias (i.e., the file drawer problem). In a post hoc analysis of 64 sleep nights of 33 healthy volunteers, Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] found that nights recorded around the full moon were characterized by a 20 minute reduction of total sleep time, reduced time spent in sleep stage 4 (S4) as well as increased sleep latency and rapid eye movement (REM) sleep latency. Furthermore, delta activity during NonREM (NREM) sleep was 30% lower in nights around full moon compared to nights around new moon. Post hoc estimation of effects sizes revealed medium to large effects of lunar class on these sleep parameters (all η2 > .10). As data were recorded in a light-controlled sleep laboratory setting, this study might be the first solid evidence confirming folk beliefs regarding lunar influences on objective measures of sleep. However, the findings by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] rely on a small number of participants and hence require replication. Here we re-analyzed three large data sets in search of lunar effects on sleep. Data set 1, recorded at the Max Planck Institute of Psychiatry, Munich, included 470 sleep EEGs from 366 healthy participants in a light-controlled sleep laboratory (see Experimental Procedures in Supplemental Information, published with this article online). Lunar classes were defined as previously [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar], with lunar class 1 representing full moon ± 4 nights, lunar class 2 covering the waxing and waning moon periods (full moon ± 5–9 nights) and lunar class 3 representing new moon (full moon ± 10–14 nights). In contrast to the results reported by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar], we did not find any evidence for an influence of lunar class on objective sleep parameters (all P >0.1, Table 1).Table 1Sleep parameters for lunar classes 1 (full moon), 2 (waxing/waning moon) and 3 (new moon) for three different samples (Munich1, Munich2 and Basel).VariableLunar class 1Lunar class 2Lunar class 3Sig. for lunar classMeanSEMeanSEMeanSEMunich1 (n = 470)n = 153n = 150n = 167F 2, 465PTST415.9±3.6420.7±3.5420.7±3.7.42.66WASO12.5±1.011.5±1.211.4±1.1.20.82SL221.3±1.619.9±1.519.7±1.5.61.55RL108.5±4.898.4±3.2103.0±3.61.79.17Stage 19.7±0.59.6±0.49.4±0.5.06.94Stage 252.6±0.853.3±0.752.5±0.6.26.77Stage 39.5±0.49.1±0.310.0±0.32.07.13Stage 47.6±0.67.7±0.68.0±0.6.01.99SWS17.1±0.716.7±0.618.0±0.7.69.50NREM69.7±0.670.1±0.570.5±0.5.31.73REM20.6±0.420.4±0.420.1±0.4.27.77Munich2 (n = 757)n = 238n = 260n = 259F 2,745PTST440.0±2.6437.7±2.4436.7±2.11.41.24WASO2.1±0.31.9±0.22.6±0.3.89.41SL214.9±1.015.2±0.815.9±0.8.21.81RL78.6±2.575.8±1.973.6±1.8.50.61Stage 16.6±0.27.0±0.27.5±0.3.02.98Stage 248.0±0.448.2±0.548.9±0.5.94.39Stage 310.9±0.210.8±0.210.7±0.22.48.08Stage 410.9±0.410.5±0.59.6±0.53.05.05∗SWS25.6±0.524.6±0.623.4±0.62.37.09NREM60.9±1.061.4±0.960.6±0.92.04.13REM21.9±0.322.0±0.321.9±0.3.35.71Basel (n = 870)n = 260n = 291n = 319F 2, 863PTST450.1±5.0452.2±4.3453.7±4.5.05.95WASO4.0±0.53.6±0.33.4±0.3.72.49SL220.5±1.623.7±2.120.5±1.81.46.23RL87.6±2.688.5±2.287.3±2.6.12.89Stage 13.4±0.13.3±0.13.3±0.1.56.57Stage 249.5±0.549.5±0.549.6±0.4.20.82Stage 311.3±0.311.3±0.311.3±0.3.02.98Stage 417.1±0.416.3±0.416.5±0.4.49.61SWS28.5±0.627.6±0.527.8±0.5.31.74NREM77.9±0.477.1±0.377.4±0.31.45.24REM18.7±0.319.6±0.319.3±0.32.47.09TST: total sleep time (in minutes); WASO: wake after sleep onset (in percent of TST); SL2: sleep latency to stage 2 (in minutes); RL: REM sleep latency (in minutes), Stage 1–4: sleep stages 1–4 (in percent of TST); SWS: slow wave sleep (sum of stages 3 and 4 in percent of TST); NREM: non-REM sleep (sum of stages 2, 3 and 4, in percent of TST), REM: REM sleep (in percent of TST). As indicated by an asterisk, sleep stage 4 in the Munich2 sample is the only variable yielding significance for lunar class (P = .05). Planned comparisons for this variable revealed a significant increase in stage 4 sleep during full moon (lunar class 1) as compared to new moon (lunar class 3, P = 0.04), which is opposite to the result reported by Cajochen et al. 2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar. None of the other planned comparisons for sleep parameters reported to be influenced by lunar cycle (TST, SL2, RL, stage 4, see 2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar) were significant. For further analyses see Supplemental Information. Open table in a new tab TST: total sleep time (in minutes); WASO: wake after sleep onset (in percent of TST); SL2: sleep latency to stage 2 (in minutes); RL: REM sleep latency (in minutes), Stage 1–4: sleep stages 1–4 (in percent of TST); SWS: slow wave sleep (sum of stages 3 and 4 in percent of TST); NREM: non-REM sleep (sum of stages 2, 3 and 4, in percent of TST), REM: REM sleep (in percent of TST). As indicated by an asterisk, sleep stage 4 in the Munich2 sample is the only variable yielding significance for lunar class (P = .05). Planned comparisons for this variable revealed a significant increase in stage 4 sleep during full moon (lunar class 1) as compared to new moon (lunar class 3, P = 0.04), which is opposite to the result reported by Cajochen et al. 2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar. None of the other planned comparisons for sleep parameters reported to be influenced by lunar cycle (TST, SL2, RL, stage 4, see 2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar) were significant. For further analyses see Supplemental Information. Data set 2 was also recorded in the Max Planck Institute of Psychiatry; however, in contrast to the first sample this data set included 757 sleep EEGs of 29 volunteers each of whom slept for 13–60 consecutive nights in the sleep laboratory, i.e. across all phases of the lunar cycle. Also in this data set we were not able to replicate any of the findings by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] (Table 1). Data set 3 represents a sample recruited at the University of Basel. It includes sleep EEGs of 870 young, healthy volunteers who slept at home. Again, there was no indication of a significant effect of the lunar cycle on sleep variables (Table 1). In addition, oscillatory activity during NREM sleep did not differ between lunar classes in this data set (Figure S1). In particular, no effects were observed in the EEG delta band as reported by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar]. In sum, we were not able to replicate the large effects of lunar cycle on sleep EEG reported by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] in three large samples consisting of 470, 757 and 870 sleep recordings: none of the variables that were reported to be affected by lunar cycle by Cajochen et al. yielded significant results in any of our samples. This also holds for alternative analyses with outlier exclusion and interaction analyses. Please note that for other sleep variables that were not significant in the data of Cajochen et al., some small and inconsistent effects and interactions with lunar class emerged (for details and discussion, see Supplemental Discussion). Even though exact values of power calculations have to be interpreted with caution due to differences in research designs, in our three samples we can exclude the existence of similarly large lunar cycle effects as reported by Cajochen (e.g., effect sizes η2 >0 .10) with high confidence: in the first data set, the statistical power to detect such effects was above 99%; in the second and third data set, we can exclude the existence of even smaller lunar cycle effects (η2 > 0.03) on the respective sleep parameters with 99% confidence. The present analysis suggests that the significant and consistent lunar cycle effects on sleep reported by Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] might be the result of other sources of variance. For example, as a post hoc re-analysis of an earlier study with a different focus, assignment of the subjects to lunar classes was not randomized for age and sex. These two factors had significant main effects on most of those sleep variables that also showed significant lunar cycle effects. However, age and sex were unevenly (even though not significantly so) distributed across lunar classes in the Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] data set (e.g., almost three times as many older subjects in lunar class 1 compared to class 3 according to their Table S3). In addition, publications of re-analyses of data already ‘exploited’ otherwise are particularly prone to publication biases such as the ‘file drawer problem’ [3Rosenthal R. The file drawer problem and tolerance for null results.Psychol. Bull. 1979; 86: 638-641Crossref Scopus (5438) Google Scholar]. Besides Cajochen et al. [2Cajochen C. Altanay-Ekici S. Münch M. Frey S. Knoblauch V. Wirz-Justice A. Evidence that the lunar cycle influences human sleep.Curr. Biol. 2013; 23: 1485-1488Abstract Full Text Full Text PDF PubMed Scopus (93) Google Scholar] and three authors of this correspondence (F.W.B., B.R., M.D.), several other researchers independently from each other had also re-analyzed sleep data from earlier studies in search of lunar cycle effects: while the post hoc analysis of 1,145 sleep diaries of 31 subjects reported similar results as Cajochen et al. [4Röösli M. Jüni P. Braun-Fahrländer C. Brinkhof M.W.G. Low N. Egger M. Sleepless night, the moon is bright: longitudinal study of lunar phase and sleep.J. Sleep Res. 2006; 15: 149-153Crossref PubMed Scopus (45) Google Scholar], neither a re-analysis of 9,778 sleep diary nights of 196 subjects [5Unpublished data from Schredl, M., Fulda S., and Reinhard, I. (2006). Dream recall and the full moon. Percept. Mot. Skills 102, 17–18.Google Scholar] nor a re-analysis of a sleep diary study additionally using actigraphy to assess 5,152 nights of 391 participants found any lunar cycle effects on sleep [6Zeitlhofer J. Klösch G. Saletu B. Barbanoj M.J. Danker-Hopfe H. Kunz D. Himanen S.-L. Kemp B. Penzel T. Roeschke J.D.G. Is there a lunar effect on subjective and objective ratings of sleep quality.J. Sleep Res. 2004; 13: 822Google Scholar]. Also, three prospective studies analyzing sleep diaries of cumulatively 6,575 nights of 147 participants failed to find evidence for lunar cycle effects on sleep [7Binkley S. Tome M.B. Crawford D. Mosher K. Human daily rhythms measured for one year.Physiol. Behav. 1990; 48: 293-298Crossref PubMed Scopus (32) Google Scholar, 8Pandey J. Grandner M. Crittenden C. Smith M.T. Perlis M.L. Meteorologic factors and subjective sleep continuity: a preliminary evaluation.Int. J. Biometeorol. 2005; 49: 152-155Crossref PubMed Scopus (18) Google Scholar, 9Krebs, S.K. (2010). Is there a connection between synodic lunar cycle and subjective sleep duration and quality? MD thesis University of Tübingen. http://nbn-resolving.de/urn:nbn:de:bsz:21-opus-522989.Google Scholar]. In sum, eight data sets with cumulatively 23,600+ nights led to negative results, compared to two data sets with a total of 1,209 nights reporting lunar cycle effects on sleep. Emphasizing the possibility of a file drawer problem, three of the studies with negative findings (among them the two largest samples) have never been published as full journal papers. However, we do not know how many more sleep researchers have already screened their sleep EEG data for lunar cycle influences without publishing their results. Thus, we would like to encourage others to report their findings. However, to overcome the obvious limitations of retrospective data analyses, carefully controlled studies specifically designed to address the question of lunar cycle influences on sleep using a within-subjects design in large samples are required for a definite answer. We would like to thank Sue Binkley, Gerhard Klösch, Michael Schredl and Jürgen Zulley for their helpful comments. B.R. received grants from the Swiss National Science Foundation (PP00P1_133685) and from the University of Zurich Clinical research priority project ‘Sleep and Health’. Download .pdf (.6 MB) Help with pdf files Document S1. Experimental procedures, one figure, and one table