Correction to Cruse et al., 2013, BMC Medicine
In 2013, we published a study in BMC Medicine titled “Actigraphy assessments of circadian sleep-wake cycles in the Vegetative and Minimally Conscious States”. In that manuscript, we described evidence of circadian rhythms in patients with prolonged disorders of consciousness, which we inferred from wrist actigraphy. The actigraphy data were recorded with devices that continually store time-stamped measures of overt wrist movement, that we know correlate with sleep-wake cycles.
All of the data in the paper was collected in Liège, Belgium, and then transferred to London, Ontario, where it was analysed and written up. Recently, while replicating the analyses in Belgium as a sanity check for a future manuscript, we have discovered that there was a software error during data export for the 2013 paper. Specifically, as the data were collected in Belgium and exported for analysis in Ontario, the software exported the data with a time-zone shift. This means that the time-stamps of all data were shifted by exactly 6-hours (i.e. the time difference between Belgium and Ontario). For example, actigraphy data that appeared to be recorded at 6pm, were in fact recorded at 12pm.
This thankfully has no impact on the analyses or the results, as it is simply a case of a constant value (i.e. 6-hours) being added to all of the patients’ time-stamps. The actigraphy data themselves are entirely unchanged. Nevertheless, it does mean that the acrophases reported in Table 1 of the original study (i.e. the time of day at which each patient’s circadian rhythm peaks) are off by exactly 6-hours.
To ensure that this information is available to those finding our original manuscript, we have published a correction to the original study.
Overall, this experience is a great example of how the chain of data handling and processing can cause subtle errors that may take 5-years to be noticed. Indeed, scientific reporting is likely riddled with subtle errors that may never be noticed. On the positive side, it is also a good example of how sharing code and data with other researchers can help to identify such errors, as we would never have found this out if not for work by others to replicate our results in advance of a new study.