We woke up on July 18th 2012 to the news “Inactivity ‘killing as many as smoking‘”. The headlines were from an article in The Lancet. It is free to view the article, but you need to register an email and password with The Lancet to be able to see the article.
It has taken a bit longer than usual to dissect this article, as it was not at all easy to understand the process that was undertaken. A 22 page appendix, with tables and little explanation, added to the problem. (In what follows, I’ve put quotations from studies in italics as well as “quotation marks” to make it easier to distinguish direct quotes.)
The findings of the study are as follows:
“Worldwide, we estimate that physical inactivity causes 6% (ranging from 3·2% in southeast Asia to 7·8% in the eastern Mediterranean region) of the burden of disease from coronary heart disease, 7% (3·9—9·6) of type 2 diabetes, 10% (5·6—14·1) of breast cancer, and 10% (5·7—13·8) of colon cancer. Inactivity causes 9% (range 5·1—12·5) of premature mortality, or more than 5·3 million of the 57 million deaths that occurred worldwide in 2008. If inactivity were not eliminated, but decreased instead by 10% or 25%, more than 533 000 and more than 1·3 million deaths, respectively, could be averted every year. We estimated that elimination of physical inactivity would increase the life expectancy of the world’s population by 0·68 (range 0·41—0·95) years.”
Let us go through the study step by step – looking at coronary heart disease (CHD) as an example (the process being the same for CHD, type 2 diabetes and the two cancers chosen – breast and colon) and see what assumptions have been made along the way.
Assumption 1 – The World Health Organisation (WHO) physical activity guidelines are correct.
In 2010, the World Health Organisation published a report called “Global Recommendations on Physical Activity for Health“.
The report says that “Adults aged 18–64 years should do at least 150 minutes of moderate-intensity aerobic physical activity throughout the week, or do at least 75 minutes of vigorous-intensity aerobic physical activity throughout the week, or an equivalent combination of moderate- and vigorous-intensity activity.”
The report gives exactly the same advice for over 64 year olds.
The reference cited for the recommendation of 150 minutes a week is the US Department of Health Physical Activity Guidelines Advisory Committee Report, 2008.
The executive summary of this report says: “Data from a large number of studies evaluating a wide variety of benefits in diverse populations generally support 30 to 60 minutes per day of moderate to vigorous intensity physical activity on 5 or more days of the week.”
That’s not very definitive. The report goes on to say: “increasing evidence suggests that participating in no more than 1 hour per week of moderate-intensity physical activity is associated with lower risk of all cause mortality and the incidence of coronary heart disease.” So one hour is OK?
And there are many concerns expressed throughout the 683 page report as to the lack of evidence and information available to the committee to make recommendations: “…it is apparent that major unanswered issues still exist in response to the question, ‘How much of what type of activity is enough to improve health?'”
The report concludes by saying: “To fill the gap in our knowledge about dose response, investigators should design and conduct studies that evaluate effects of the following variables at fixed volumes of physical activity: intensity, frequency, duration, and multiple bouts. Details related to these variables would allow more precise physical activity guidelines to be developed across the breadth of activity-related health outcomes.”
150 minutes or “five-a-week” as the UK Department of Health likes to call it (to accompany five-a-day presumably) is not a robust scientific absolute standard. It is a guideline. As the closing conclusion of the committee report says – we do not know the optimal duration, frequency, intensity and type of activity that is going to provide optimal benefit. We do not know the lowest point of benefit or harm; we do not know the highest point of benefit or harm. “Five-a-week” is memorable and achievable – like five-a-day – but it is not an absolute marker upon which to then base such a comprehensive study.
Assumption 2 – 150 minutes activity a week is the only thing that matters. 149 minutes counts for nothing. 300 minutes is not better or worse.
The entire study was about the WHO measure of inactivity – defined as the percentage of each population NOT meeting the 150 minutes of exercise each week. If 50% of population A were meeting this guideline and the other 50% were doing 149 minutes of exercise each week, this would have the same inactivity score as population B with 50% doing 150 minutes of exercise each week and the other 50% doing nothing at all. There is no allowance for how much activity is done by a population overall – just what percentage of the population tick this recommended 150 minutes box.
Assumption 3 – Activity is what people do during leisure time.
P220 of the Lancet report (this is the second page of the article) states: “Data for prevalence of inactivity depended on the instrument used for assessment and varied according to whether a study assessed physical activity during leisure only (most commonly) …”
Given that the study is global – all countries from Ethiopia to America, Malta to Australia and Argentina to Iran, the lifestyles that are naturally active (the ‘third world’ farming communities) could hardly be more different to the 4am gym generation of New York.
The study also relies on self reported activity, which is notoriously unreliable. Japan recorded one of the highest levels of INactivity, which surprised me. Did the Japanese people not think of their walk or cycle to work as activity?
We need to spend a bit of time on Step 1 because it is the crux of the study. This step is not even in the body of the article – it is the appendix – the 22 pages of tables with next to no explanations, as mentioned in the introduction to this post.
The team found 10 epidemiological studies (studies of a particular population) where data had been collected over a period of time for activity and incidence of coronary heart disease, type 2 diabetes, colon cancer and breast cancer – the factors being considered by this study.
For coronary heart disease there were 8 studies that recorded “persons eventually developing or dying from CHD” during the study. Please note here that this stage of the process is therefore recording incidence and deaths from CHD, not just deaths. The conclusions of the overall study, which rely fundamentally on this stage, are all about deaths.
The 8 studies were:
|Summary Table||Genders||Dates||A (%)||B (%)||C|
|Aerobics Center Longitudinal Study||Both||1978-2006||45.5||55.8||1.23|
|EPIC – Norfolk Study||Both||1993-2011||30.7||39.4||1.28|
|Harvard Alumni Health Study||Men||1988-2008||22.8||27.0||1.18|
|Shanghai Women’s Health Study||Women||1997-2004||45.4||50.2||1.11|
|Scottish Health Study||Both||1998-2008||62.1||72.7||1.17|
|Women’s Health Study||Women||1992-2010||47.4||53.4||1.13|
|Average of adjustment factors||1.20|
Column A is inactivity for the overall population
Column B is inactivity for “persons eventually developing or dying from CHD”
Column C is the Adjustment Factor = Column B divided by Column A
The arithmetic average of the 8 adjustment factors for each study is 1.20.
The FINRISK study had approximately 7,000 participants; the Aerobics Center Longitudinal Study had over 50,000 participants. I found this part of the process incredible – the straight average of all the studies was taken – not a weighted average. Three of the studies were American, three British, one from Finland and one of Shanghai women. These eight were averaged to apply a global factor to then be applied to every country in the study.
The study took inactivity overall in each study and inactivity in “persons eventually developing or dying from CHD.” This was used to derive an “adjustment factor”. The article explains the process as follows:
“For example, in the Shanghai Women’s Health Study, the prevalence of inactivity in all women at baseline was 45·4% versus 51·6% in women who died, yielding an adjustment factor of 1·14 (51·6 / 45·4 = 1·14). For each outcome, we calculated the adjustment factor in every study, and averaged this factor across studies. We applied the average adjustment factor to the prevalence of physical inactivity, by country, to estimate the prevalence of inactivity in cases of coronary heart disease, type 2 diabetes, breast and colon cancer, and death from any cause.”
The straight average (add the 8 numbers in column C and divide by 8) is 1.20 – this is the adjustment factor applied to every country in the world for the increased ‘risk’ of CHD by being inactive.
All the usual factors that are wrong with most studies are wrong with this:
– This is association not causation and therefore nothing can be said about risk. Does inactivity cause ill health or are people inactive because they are ill? The direction of possible causation is as plausible either way round.
– This is just looking at two factors that cannot be looked at in isolation. It is assuming that an observation about inactivity/activity and people going on to develop CHD is the whole story. What about smoking? stress? processed food consumption? alcohol? An active person is more likely to have a healthy lifestyle – less likely to smoke, drink and eat junk. Hence an even stronger association could have been found in these eight studies between smoking and going on to develop CHD or processed food consumption and going on to develop CHD and so on. No risk factor can be assumed by looking at two factors in isolation.
– Taking the Aerobics Center Longitudinal Study as an example, baseline activity levels as far back as 1978 are assumed by this study to determine, working through the steps, global deaths in 2008.
– This is about relative risk, not absolute risk. The absolute risk is likely extremely small – as it is in all these types of studies. Let’s take the Whitehall study as an example. It had 10,000 participants. 40.2% of these were inactive (4,020 people); the remaining 5,980 were active. Let us keep the maths simple and assume that 100 participants developed CHD
|Whitehall study as an example|
|Inactive in total study (%)||40.2|
|Inactive in people who went on to develop CHD (%)||45.4|
|Assume 100 people developed CHD||100|
|Inactive people who went on to develop CHD||45.4|
|Active people who went on to develop CHD (*)||54.6|
|Overall inactive people||4,020|
|Overall active people||5,980|
|% of inactive who developed CHD||1.13%|
|% of active who developed CHD||0.91%|
(*) This is quite amusing in itself – more active people went on to develop CHD than inactive people. This is the case for every figure below 50% for inactive people who went on to develop CHD. This applies in half of the eight studies.
Slight digression here just to add the data from the study to confirm this point: Table 1 in the paper confirms more active than inactive people went on to develop each condition. Table 1 uses data for all countries included in the study:
– The prevalence of inactivity in people eventually developing CHD was 42.2%. This means that 57.8% of people who eventually developed CHD were active;
– The prevalence of inactivity in people eventually developing type 2 diabetes was 43.2%. This means that 56.8% of people who eventually developed type 2 diabetes were active;
– The prevalence of inactivity in women eventually developing breast cancer was 40.7%. This means that 59.3% of women who eventually developed breast cancer were active;
– The prevalence of inactivity in people eventually developing colon cancer was 42.9%. This means that 57.1% of people who eventually developed colon cancer were active.
I could present the data from the Whitehall study table above as 20% more active people developed CHD than inactive people [(54.6-45.4)/45.4]
I could present this data above as 24% more inactive people developed CHD than active people [(1.13%-0.91%)/0.91%]
Lies, damned lies and statistics.
The team admitted that data for inactivity according to the WHO guidelines was available for North America and Europe, but not the rest of the world. They divided the study into seven regions. The number of countries in each region is in brackets after the region name: Africa (34); Latin America and Caribbean (13); North America (2); Eastern Mediterranean (9); Europe (36); Southeast Asia (9) and Western Pacific (19). The study is impressive in its coverage. However, this is its biggest fault in terms of accuracy and common sense. Data for North American and Europe has been extrapolated to apply to the other five regions and lifestyles could not be more different.
The paper describes the lengths that the team undertook to get missing data for five regions. Given that the inactivity data from the WHO was comprehensive for North America and Europe and given that the studies that were being used to create an adjustment factor were from the same regions, with one exception, surely the study should have focused on regions where the data was more robust? Notwithstanding all the errors made in Step 1 already.
Using best data/estimates found for five regions, the bulk of the paper presents Table 2 – a large table with all 122 countries listed with an estimated PAF. “The Population Attributable Fraction (PAF) is a measure used by epidemiologists to estimate the effect of a risk factor on disease incidence in a population. It estimates the proportion of new cases that would not occur absent a particular risk factor.” (quotation from The Lancet paper).
This table has applied the 1.20 CHD factor taken in Step 1 from the eight studies and applied it to the inactivity level for each country to indicate the percentage of “new cases that would not occur” if people were active. Please note again – this should still only be about incidence of CHD – the numbers leap from these percentages to life expectancy assuming that everyone who developed CHD (or type 2 diabetes, colon cancer or breast cancer for the other conditions in the study) dies.
Step 3 gave the summary numbers for the “findings” paragraph. “Worldwide, we estimate that physical inactivity causes 6% (ranging from 3·2% in southeast Asia to 7·8% in the eastern Mediterranean region) of the burden of disease from coronary heart disease…”
The summary for Africa is 3.9%; Latin America and Caribbean 7.1%; North America 6.2%; Eastern Mediterranean 7.8%; Europe 5.5%; Southeast Asia 3.2% and Western Pacific 7.2%. Overall the median is 5.8% just for CHD.
Some of the numbers seem extraordinary at a glance – Greece is apparently the most active (least inactive) country in Europe (really? Greece is too hot to move, let alone to beat the rest of Europe). The Cook Islands in the Pacific and Malta in Europe are apparently the most inactive countries in the world with Argentina (Latin America) and Swaziland (Africa) not far behind. Bangladesh is apparently the most active (least inactive) country of the 122 documented. The North American estimated PAF is 6.2% – close to the world average, so the North Americans are not any more or less ‘lazy’ than average.
Remembering that this is about disease incidences, not deaths, these PAF estimates are nonetheless extrapolated into Table 3 – the gain in life expectancy that could be enjoyed by each country if only they ‘did five-a-week’. The inactive Cook Islands are claimed to stand to gain an extra year and a half if only they got off their arses; Bangladesh could gain about a month. Overall the median life expectancy to be gained is 0.68 of one year.
I couldn’t resist this back of the envelope:
If someone does the 2.5 hours per week from the age of 18-64 as recommended, this equates to 5.4 days a year or 249 days in total. The life expectancy ‘to be gained’ is one day fewer – 248 days!
|Time spent exercising age 18-64|
|Life expectancy increase 0.68 years||248.2||days|
|Exercise from age 18-64 (46 years)||249.2||days|
Back to assumptions now, as this is the right time to introduce a 4th assumption that has been made:
Assumption 4 – Even if a country does not suffer a condition it is assumed that lives can still be saved!
This is another astonishing one. By applying the 1.20 factor (step 1) to all countries of the world to the estimated inactivity level for that country, it is assumed that CHD deaths (should be incidences) can be avoided. A country might have no, or relatively very few, incidences of some of these more typically Western conditions. Here are Africa’s worst countries by famine . Only two of these countries are NOT included in the Lancet study (Angola and Lesotho). I suspect people in Ethiopia and the rest of these countries are dying from starvation and infectious disease, not from type 2 diabetes or CHD.
War torn regions are also included in this study – Sierra Leone, Zimbabwe, The Congo – does anyone live long enough in such regions to develop heart disease or type 2 diabetes?
The study applies the PAF percentages for each region and concludes that “15,000 deaths from coronary heart disease in Africa could have been averted in 2008 by removal of physical inactivity. 60 000 could have been avoided in the Americas, 44 000 in the eastern Mediterranean region, 121 000 in Europe, 59 000 in southeast Asia, and 100 000 in the western Pacific region.”
“For all-cause mortality, the overall median PAF was 9%. Applying this figure to the 57 million deaths worldwide in 2008, we estimated that more than 5·3 million deaths (ranging from 525 000 in the eastern Mediterranean to 1·5 million in the western Pacific region) could be averted every year if all inactive people become active.”
There was a much quicker back of the envelope… Remember the adjustment factor assumed from the eight studies back in Step 1? It was 1.20 for CHD and the paper noted a factor of 1.22 to be applied for all-cause mortality (calculated from a slightly different eight studies). Overall inactivity (table 1) was noted as 42.9%. Multiply inactivity (42.9%) by what has been assumed to be the risk factor for inactivity (0.22) and this gives 9%. Then declare that 9% of the 57 million deaths worldwide, which occurred for reasons from war to famine to infectious disease to non communicable disease, could have been avoided if only people did a magic number of 150 minutes of activity each week, as recorded in eight studies back as far as 1978.
PostScript: I do think that natural activity is a good thing to do. I don’t include marathon running, step aerobics and repetitive gym activities in this. Walking, tending the land, building, cleaning, lifting, playing, climbing – the natural activities resurrected by the Paleo movement – are what humans should be doing. I suspect the most populous countries in the world, China and India, are doing a great deal of this and probably not considering it exercise. I do think that such natural activity (especially outdoors) will be beneficial for human health – for conditions as diverse as heart disease and anxiety. However, there are far too many errors, assumptions and unreasonable extrapolations in this study for this claim to be credible: “physical inactivity seems to have an effect similar to that of smoking.”
Yet another study more about headline grabbing than evidence based genuine discoveries.