This is the opening to Chapter Nine in my book The Obesity Epidemic: What caused it? How can we stop it?
Extract from The Obesity Epidemic
“There are lies, damn lies and statistics”, Mark Twain.
Have you ever wondered why you constantly hear “one in three people die from heart disease” and yet you don’t lose one third of your friends and family every year? The USA death rate from all causes for 2006 was 0.78%.[i] That means 777 people, per 100,000 residents of the population, died in 2006. Death certificates recorded 200 of these deaths as heart disease. So 0.2% of the USA population died from heart disease in 2006. If you have 500 friends, you are likely to lose one of them to heart disease during a year. If most of your friends are female, heart disease is so weighted towards males that you could have 620 female friends, or 405 male friends, with the same risk of losing one friend during a year. That’s still one lost friend too many, but there is also an age dimension to consider. You would need to know 166,667 children aged 5-14 to have a likelihood of one dying from heart disease (which would in turn most likely be a rare hereditary condition) and yet, if you know 100 people aged 85 and over, 22 of them are likely to have heart disease on their death certificate over the next year.
In 1950, staying with the same American data, the death rate from all causes was 1.45%. Hence, this has been virtually halved with advances in medical treatment. (I would argue that, had we maintained our natural eating heritage and advanced our treatment options in the way that we have, death rates could have been reduced further still). Of these 1,446 deaths per 100,000 people, 589 (41%) were recorded as heart disease. This is interesting per se as the World Health Organisation was only formed in 1948 and heart disease was a little recognised condition before then. By 1960, the death rate had fallen to 1.34% with 42% of those deaths recorded as heart related.
In 1970, the year that the Seven Countries Study was published, the overall death rate had fallen further to 1.22% and heart deaths had fallen to 40% of these. This could be interpreted as fewer than six people in 1,000 were dying from heart disease, or four in ten – depending on lies and statistics. The four in ten positioning provides the context for the impetus for change that preceded the 1977 Dietary Goals for the United States announced by Senator George S. McGovern, chair of the Senate Nutrition Committee.
“Heart deaths have halved”
I was reminded of these statistics with the headlines of 26 January 2012:
The Telegraph: Heart attack deaths halve in a decade.
Daily Mail: Heart attack deaths halve in eight years.
and The Heart.org: Heart attack deaths plummet in three EU nations
as just three examples.
It’s a really excellent article – well written, well laid out, well explained, a great concept for a topic to research – far better than the sensationalist headlines that emanated from it. However, it did invite these headlines with the conclusion “From 2002 to 2010 in England, the age standardised total mortality rate fell by about half”.
This is where the extract from The Obesity Epidemic helps to show the two ways in which numbers can be presented. If we think (as many people currently do) that one in three people dies from heart disease, these headlines will have us thinking only half a person in three or one in six will now be dying from heart disease. “I’ve got six friends, so only one will die, instead of two!”
It also reminded me of the section in Dr Malcolm Kendrick’s The Great Cholesterol Con where he showed how he can double our chances of winning the lottery – buy two tickets! We now have a one in seven million chance of winning whereas, with one ticket, we had a one in fourteen million chance. Most sensible people would realise that we still have naff all chance of winning the lottery, but our chances have doubled.
Looking at the section in the BMJ paper entitled “Trends in Mortality”, we can see that “From 2002 to 2010, the age standardised mortality rate from acute myocardial infarction per 100 000 population decreased in men from 78.7 to 39.2 and in women from 37.3 to 17.7.” That is, indeed, almost halved in men and slightly more than halved in women. However, look at those lottery numbers.
In 2002, not even 1 man in 1,000 died from acute myocardial infarction (a heart attack) and in 2010 not even 1 man in 2,000 died from acute myocardial infarction. Put another way, if you knew 1,271 men in 2002 you would likely know one who died of a heart attack. In 2010 you would need to know 2,551 men to have one likely die of a heart attack. These are still really low odds. Yes, they have halved (as they did in the USA between 1950 and 2006), but they are still really low odds. Thinking you need to know 2,551 men before one will die is rather different to thinking you will now lose one friend in six, rather than one friend in three.
Event rate and fatality rate
The death rate is a function of the event rate and the event fatality rate i.e. does a heart attack occur and is the heart attack fatal. The real contribution of the paper was the data analysis to see that the improvements in mortality rates were approximately split between these two factors. About half the reduction came from the event rate (fewer incidents) and the other half came from the fatality rate (those incidents that proved fatal). However, this was not the case across all age groups: “Changes in the event rate played a bigger part in middle aged people, whereas change in case fatality played a slightly bigger part in young and very old people. This can be explained by the fact that , while the average annual changes between 2002 and 2010 in case fatality rate did not differ substantially between age groups, the declines in event rate did.”
a) Improvements in the event rate were attributed to “During our study period, national survey data for England indicates that the prevalence of smoking, high blood pressure, and high cholesterol levels decreased, while the prevalence of obesity and diabetes increased.”
I would disagree with two out of three of these. Declines in smoking – and particularly the ban on smoking in public places from March 2006 (Scotland) to July 2007 (England) – should have had a difference in event rate. High blood pressure is a symptom, not a problem – the underlying problem needs to be explained and how this changed between 2002-2010 illustrated. I do not consider cholesterol to be a risk factor in heart disease – quite the opposite – it is one of the major repair tools at our body’s disposal. Hence there is insufficient explanation for change in event rate, unless smoking can explain the whole.
Also, the paper notes that “The annual declines in event rates were not statistically significant for men and women aged 30-54 and 85 and older, which raises the possibility that there was no real improvement in the rate of occurrence of acute myocardial infarction in these age groups over the past decade in England.”
b) Improvements in the fatality rate have long been attributed to faster response times for initial treatment and general improvements in treating people in those first critical few minutes when fatalities are most likely.
This is all about numbers – “Lies, damn lies and statistics”.
The headlines scream “Heart attack deaths halved” and indeed they have. But your chance of winning the lottery can still double if you buy two tickets and this doubling is pretty meaningless. A man’s chance of dying from a heart attack was 0.079% in 2002 – not even a tenth of one percent. In 2010 this chance was 0.039% – not even half of a tenth of one percent. For women the chances were 0.037% and 0.018% in 2002 and 2010 respectively. Your personal chances of losing a friend from a heart attack were small and are now smaller. Your chance of winning the lottery is small and could be smaller. The headlines did not reflect this.
[i] Centres for Disease Control and Prevention, (Using age adjusted data), http://www.cdc.gov/nchs/hus.htm (data page), ftp://ftp.cdc.gov/pub/Health_Statistics/NCHS/Publications/Health_US/hus09tables/09contents_tables.pdf, (detailed list of data available).