The largest single study in the Lowe meta-analysis is Banks (2010), which had 159/1688 of the participants (9.4%). In table 2, Lowe reports that Banks measured 1 day of sleep restriction. But Banks actually did 5 days of 4hr sleep restriction/4hr time in bed, followed by 1 “recovery” day of 0, 2, 4, 6, 8, or 10 hours of time in bed. It’s unclear why Lowe coded it as 1 day of cumulative sleep restriction rather than 5-6. In addition, I don’t understand how they would have incorporated effect sizes into their meta-analytic procedure.
Given that this single study is a large proportion of their total participants, and that the meta-analysis analyzes the mediating effect of days of cumulative sleep restriction on cognitive function, I think this is potentially important to understand better. I’ve emailed the author to ask about this. If and when I hear back, I’ll let you know.
In table 2, Lowe reports that Banks measured 1 day of sleep restriction. But Banks actually did 5 days of 4hr sleep restriction/4hr time in bed, followed by 1 “recovery” day of 0, 2, 4, 6, 8, or 10 hours of time in bed. It’s unclear why Lowe coded it as 1 day of cumulative sleep restriction rather than 5-6.
Banks (2010) conducted neurobehavioral assessments on every day of sleep restriction (including the first one). See Figure 2 in their paper.
[In case you’re wondering how Lowe et al extracted data from Figure 2, apparently there’s software for that. From their article:
A Plot Digitizer program (Rohatgi, 2017) was used to estimate means and standard deviations (SD) when data was available in chart form only.
Banks (2010) conducted neurobehavioral assessments on every day of sleep restriction (including the first one). See Figure 2 in their paper.
I’m not sure why that’s relevant to my question? They still did 5-6 days (depending on whether you count the recovery day, on which many SR subjects got even less sleep than before) of sleep restriction in total. Unless Lowe is only counting the results from the first day of sleep restriction, which is possible but seems unlikely, describing this as 1 day of cumulative sleep restriction duration seems inaccurate. However, I will withhold judgment until I hear back from Lowe.
Huh. I had assumed they might have been only including the first day, but you’re right that that’s unlikely. Probably a typo.
That wouldn’t change their overall meta-analytic effect sizes, though, since they aggregate across different durations of sleep restriction when calculating it.
If it was indeed a typo (there is at least one more in the same table—Saleh (2003) is duplicated when there should be one entry for 2003 and one for Saleh 2011). On the other hand, it’s possible it was miscoded or misinterpreted. I’d be surprised if that were the case, but it’s hard to see, for example, how they came up with a number for “mean hours of sleep deprivation” given the 6 different cases for the recovery day sleep in the experimental group, or whether they code an effect size for the group that was sleep deprived for 5 days and then got 10 hours of sleep on the recovery day.
Another odd discrepancy.
The largest single study in the Lowe meta-analysis is Banks (2010), which had 159/1688 of the participants (9.4%). In table 2, Lowe reports that Banks measured 1 day of sleep restriction. But Banks actually did 5 days of 4hr sleep restriction/4hr time in bed, followed by 1 “recovery” day of 0, 2, 4, 6, 8, or 10 hours of time in bed. It’s unclear why Lowe coded it as 1 day of cumulative sleep restriction rather than 5-6. In addition, I don’t understand how they would have incorporated effect sizes into their meta-analytic procedure.
Given that this single study is a large proportion of their total participants, and that the meta-analysis analyzes the mediating effect of days of cumulative sleep restriction on cognitive function, I think this is potentially important to understand better. I’ve emailed the author to ask about this. If and when I hear back, I’ll let you know.
Banks (2010) conducted neurobehavioral assessments on every day of sleep restriction (including the first one). See Figure 2 in their paper.
[In case you’re wondering how Lowe et al extracted data from Figure 2, apparently there’s software for that. From their article:
]
I’m not sure why that’s relevant to my question? They still did 5-6 days (depending on whether you count the recovery day, on which many SR subjects got even less sleep than before) of sleep restriction in total. Unless Lowe is only counting the results from the first day of sleep restriction, which is possible but seems unlikely, describing this as 1 day of cumulative sleep restriction duration seems inaccurate. However, I will withhold judgment until I hear back from Lowe.
Huh. I had assumed they might have been only including the first day, but you’re right that that’s unlikely. Probably a typo.
That wouldn’t change their overall meta-analytic effect sizes, though, since they aggregate across different durations of sleep restriction when calculating it.
Dr. Lowe did get back to me briefly to say she’s looking into it. I’ll post an update when I hear back.
If it was indeed a typo (there is at least one more in the same table—Saleh (2003) is duplicated when there should be one entry for 2003 and one for Saleh 2011). On the other hand, it’s possible it was miscoded or misinterpreted. I’d be surprised if that were the case, but it’s hard to see, for example, how they came up with a number for “mean hours of sleep deprivation” given the 6 different cases for the recovery day sleep in the experimental group, or whether they code an effect size for the group that was sleep deprived for 5 days and then got 10 hours of sleep on the recovery day.