edit: this started about statistics and ended up about smoking. Why? Who knows?!
I’ve started writing a post on the anti-vaccination movement, but I’ve realized it relies on an understanding of some epidemiological and statistical jargon / concepts, so I thought I’d throw together a quick glossary. These are kind of useful terms and concepts to understand anyway, and they’ll probably come up again in subsequent posts on this blog, so I thought it would be good to have something to link back to.
Incidence rate: newspaper reports often occur to the absolute numbers of cases during an outbreak, and they may compare it to the number of cases in the same community the previous year or decade. This isn’t necessarily useful if you don’t know how many people were in the community at both points in time. If the number of cases increases by 10% (say from 8000 to 8800), that has different implications depending on what happened to the size of the community, which the newspaper frequently doesn’t tell you. If the population also increased 10% over the same period of time, the disease is not actually any more common. Epidemiologists therefore report the “incidence rate” wherever possible – usually the number of cases per 100,000 people per year, or “person-years”. For example, Australia confirmed 37,000 cases of swine flu in 2009 in a population of 22 million, so the incidence rate was 1,680 / 100,000 person years. That’s a lot of swine flu.
Risks, risk ratios and absolute risk increases: “Risk” refers to what is colloquially referred to as the chance of something happening. The risk of an event is always between 0% (never occurs) and 100% (always occurs), which are also expressed as 0.0 and 1.0. The risk of an individual Australian getting swine flu in 2009 can be expressed as having been 1.68% , 0.0168, or 1 in 59.
Saying something increases the risk of a negative outcome by 50%, or 1.5-fold, sounds quite undesirable, however that statement doesn’t tell you anything about the absolute risk of either event. You might be comfortable taking a drug which increased your risk of a rare disease from 1 in 10,000 to 1 in 7,500, or from 0.01% to 0.015%. This is a 50% increase, or what epidemiologists would call a “risk ratio” of 1.5 – the risk of disease is 1.5-fold greater in those taking the drug as those not taking it, but the “absolute risk increase” is only 0.005%. You probably wouldn’t be willing to take a drug which increased your risk of a common disease, such as ischemic stroke, from 1 in 20 to 1 in 15, or 5% to 7.5% – another risk ratio of 1.5, but an absolute risk increase of 2.5%.
This is something to be wary of in newspaper articles and even public health campaigns – saying something increases the risk of an event or disease by x-fold or x% often makes the absolute increase in risk sound far more serious than it truly is. Anti-smoking campaigns are the best example of this. A recent meta-analysis of cohort studies observed rates from 9 to 20-fold higher in female smokers, and 22 to 36-fold higher in male smokers. Risk ratios this high are extremely rare in chronic disease epidemiology – a lot of modern research focuses on risk ratios less than 2. A tutor of mine who works in prostate cancer tells me they’d be excited to find a factor which caused a risk ratio of 1.1.
So smoking does dramatically increase your risk of lung cancer, but it doesn’t bring your absolute risk up to the level a lot of people imagine. I know smokers who claim to simply accept that lung cancer is inevitable for them, because that’s the impression you get from the campaign. It’s not entirely accurate, especially for women.
Lung cancer was the 4th leading cause of death in Australia in 2010, causing 8,100 deaths to Alzheimer and Dementia’s 9,001, stroke’s 11,200 and cardiovascular disease’s 21,700. The observed incidence of lung cancer in people aged 40 – 79 who have never smoked in that meta-analysis ranged from 5 up to 14 per 100,000 person years in men, and from 14 up to 21 per 100,000 person years in women. The rates in current smokers ranged from 174 to 363 / 100,000 person years in men, and 149 to 293 / 100,000 person years in women. At worst, then, 3-4 smokers per thousand will get lung cancer in any given year. For men, this is significantly lower than the observed rates of heart attacks, from 346 to 486 / 100,000 person years. The rates for women are from 190 to 354 / 100,000 person years for women, slightly higher than the rate of lung cancer in female smokers. The estimated lifetime risk of lung cancer for Canadian smokers is 17.2% or roughly 1 in 6 for men, and 11.6% or around 1 in 9 for women.
A one in nine chance of a frequently inoperable, fatal condition is obviously significant, and I personally decided it wasn’t worth it a few years ago. None the less, while well-intentioned, anti-smoking campaigns which are designed to give the impression that a majority of smokers will eventually get lung cancer are simply dishonest. If those meta-analyses are to be believed, the average American is more likely to have a heart attack than the average American smoker is to get lung cancer. Given that smoking also significantly increases your risk of cardiovascular disease, if I were a smoker, especially an overweight male smoker, I’d be a lot more worried about the heart attack.