September 11, 2017

sedentary behaviour and mortality

Publishing in Annals of Internal Medicine, researchers looked at sedentary behaviour and reported that increased sedentary behaviour, both in total volume and prolonged uninterrupted bouts, was associated with increased risk of death.

A roundup accompanied this analysis.

Title, Date of Publication & Journal

Title: ‘Patterns of Sedentary Behavior and Mortality in U.S. Middle-Aged and Older Adults: A National Cohort Study’ by Diaz et al

Published: Monday 11th September

Annals of Internal Medicine

Study’s main claims – and are they supported by the data

The papers main claims are:

Greater total sedentary and longer sedentary bout each had a dose dependent association with higher risk for all-cause mortality.

This claim is supported. The effects of total sedentary time and sedentary bout duration follow clear dose-dependent trends.

However, total sedentary time and bout length are also correlated so some of the association of bout length with mortality may be due to longer total sedentary time and vice versa.  In Figure 3, comparing high and low sedentary bout duration within people with high sedentary time, and doing the same within people with low sedentary time suggests that overall sedentary time may be the main driver of mortality.

We cannot be certain about this though as some of the confidence intervals (if they had been drawn on Figure 3) would have been wide because of the low numbers of participants in some groups.  The authors do acknowledge that sedentary time may be the more important variable in their limitations section.

Participants with a larger proportion of sedentary bouts of 90+ minutes had higher mortality, and conversely those with a lower proportion of bouts of 1-29 minutes also had higher mortality.

This statement is only partially supported by plots in the main paper.

Confidence intervals are not shown in the relevant plots or reported in the text, so it is difficult to assess the validity of this statement. The means and standard deviations of bout duration in Table 1 suggest that bout lengths of 90+ minutes are likely to occur mostly in individuals who are in the top quartile for sedentary time overall. Therefore the higher risk of mortality in those with the largest proportions of 90+ minute bouts could be partially attributable to longer overall sedentary time.

The press release mostly describes results as stated in the paper. However, the headline Frequent movement breaks throughout the day could potentially mitigate the effects of prolonged sitting is inaccurate as sedentary time in this study includes both sitting and standing periods with low movement.

Strengths/Limitations

Strengths

The study objectively measures sedentary behaviour and physical activity, which reduces reporting bias.

The study looks at all-cause mortality, which is clearly a significant outcome that is objectively measured.  It is also a statistically sound outcome as it’s hard to misclassify death or the date on which it occurs, and death is usually well recorded on death certificates. Nor are there likely to be different definitions.

According to the authors, this study is the largest of its type to date.

The study is based within an existing population cohort study (the REGARDS study).

Limitations

This is an observational study and therefore observed associations cannot infer from these results alone that a cause-effect relationship exists.

Participants are all aged 45 and older so the study cannot determine whether sedentary behaviour is important below the age of 45. These results cannot be generalised to a younger age group without assuming that participants’ sedentary behaviour was the same when they were younger as it was when measured in the study.

Sedentary about duration and overall bout time are positively correlated. Therefore, associations between bout duration time or percent of long/short bouts with mortality could be explained by overall sedentary time, and associations of sedentary time with mortality could be explained by longer bout duration. This is not adequately examined in the paper.

In the main paper, the majority of results are presented as plots of cumulative mortality, which do not include confidence intervals. This makes it difficult to judge the robustness of the conclusions reached.

Participants were followed up for a relatively short time (median of 4 years) to be able to investigate associations with mortality.

The accelerometer was only worn for seven days and it did not distinguish between sitting and standing postures. Using only seven days of activity may not be a representative measure for an individual’s typical pattern of activity throughout their lifetime.

Some risk factors were collected approximately 6 years before the accelerometer data was collected and may have changed by the time the study was started.

Some participants are likely to have developed ill health by the study start, which may affect both sedentary time and mortality. Some potentially important confounding variables were not measured. For example, the study adjusts for an incomplete list of health variables, and the study did not adjust for employment type or socio-economic status.

Glossary

Bout duration: the specific time period during which an individual is sedentary

Dose-dependent relationship: When the effects change when the ‘dose’ of something (e.g. sedentary behaviour) changes

Any specific expertise relevant to studied paper (beyond statistical)?

One reviewer has experience of analysing and designing prospective cohort studies, but no expertise in studies of physical activity or sedentary behaviour. The other reviewer has experience with accelerometer algorithms but no other expertise relevant to this paper.

* ‘Patterns of Sedentary Behavior and Mortality in U.S. Middle-Aged and Older Adults’ by Diaz et al. published in Annals of Internal Medicine on Monday 11 September. 

Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.