A Study of a Study: Lifestyle Changes Add 14 Years your Life

The news headlines come out:

But what does this actually mean?

Where do these claims come from?

There are many studies that have headlines like this.

Let's unpack this a bit, it's a bit of a ride!
I am going to be using a Harvard Study from 2018.

Here is the context that leads to the headlines:
The projected life expectancy at age 50 years was on average 14.0 years (95% CI, 11.8–16.2) longer among female Americans with 5 low-risk factors compared with those with zero low-risk factors; for men, the difference was 12.2 years (95% CI, 10.1–14.2).
So before getting much further into the guts of the study, we already have some more information:
  • We are referring to life expectancy at age 50.  
    • This means that the gain for someone at age 30 vs. someone age 60 will be different.
  • 14 years and 12 years are average gains for females and males, respectively
    • This means some get more, and others get less. 
    • And also, we get ranges, indicated by CI, which means confidence interval, to show the distributions. 
    • Turns out a good portion of men and women at age 50 could experience the same gain!
  • The gain in life expectancy represents the difference between having 5 low-risk factors and having 0 low risk factors. So this is the most extreme result! 
    • An important nuance here is that this is a comparison in mortality between groups. It is not the observed difference based on movement between groups. (this would be insanely hard to figure out!)  In other words, even though the 3's have longer lives than 2's, you are not guaranteed to extend your life by moving up one category. On the other hand, you may get even more gains! 


According to the study, a low risk profile is someone who
  1.  Eats food that is healthy-ish.
  2.  Never smokes.
  3.  Does 30 minutes per day of moderate to vigorous activity
  4.  Does not drink too much alcohol
  5.  Falls within a certain BMI
Score 1 point if you meet a criteria, zero points if you do not.

You will get a number between 0 and 5, with 5 representing the lowest risk.

The study grouped a large number of people (we will talk about the study group later) into these 6 groupings.: 0-5.

In the paper they breakdown how much of the groups they studied fall into each score.
  • In one group, the zeros were about 7% of the total and the fives were about 2%.  
  • In another group the zeros were closer to 10% and the fives were 1.5%.   

Let's think about the headline claim with this bit of information If the population they study is a good match for the rest of us (more on this later):

  • Only a small percentage of us will get to be a five.  
  • And, most likely, we are not coming from a zero.  
  • So the bummer is that the gain we will see is probably lower than a 0 to 5 jump.y
The graph below shows the gains in life expectancy by age. Each line represents getting one more point in your low-risk score. The difference between two lines at a given age is how much more life expectancy you get by lowering your risk.

Should each risk factor be counted equally?  

The researches did do a variant of this question. But to help quantify this, they used a tool called the 'hazard ratio'.  Under this metric, if groups were the same, the ratio would be 1.  A number less than one means there is less hazard. 

The fives had a hazard ratio of 0.21 relative to the zeros. The hazard, in this case, is dying for any reason during the study period.  They studied other hazards as well, like cardiovascular disease and cancer.

The researchers group each risk factor into five groups (they must really enjoy the number five). And within a risk factor compared the rate of death of people within each subgroup.

A key point here is the idea of relative rates, each category has a different basis of reference. 
  1. Eating/Diet - Each improvement in eating score reduced the hazard ratio a little bit. 
  2. Smoking- Smoking cessation has by far the most significant impact relative to smoking at all. 
  3. Movement/Activity - The baseline is zero to 0.4 hours per week. Anything above that shows reduced hazard ratios. The impact is very similar to eating.
  4. Alcohol is interesting. The abstainers and rare drinkers showed a hazard ratio greater than 1 relative to moderate drinkers. And it's about the same magnitude as heavy drinkers.
  5. BMI: Too low has higher hazards, probably since some diseases lead to thinness. Too high,also has a higher hazard ratio. 
  6. Comparing Low-Risk Factors: Having a score of 1, (we don't know which one item it is though), takes the hazard to 0.79.  And ultimately, scoring 5 gets you to 0.26.

How Do We Get From Relative Risk to Years of Life?

This is the magic sauce.  The hazards of death go to into something called a life table, from which we take the average number of years lived. It's all math.

Lower risk of death means more people will make it to more years of life. 

The increase in life is mostly attributable to reduced risk in heart issues, a little bit from reduced cancer risk, and then everything else.

Other Considerations:

A lot of this stuff lines up with my Assumptions and Approaches post and Peter Attia's Studying Studies.
  1. The study population:  
    • A lot of the data collected was either from a group of nurses (age 50 and older) or from medical workers (age 30 and older). If you are a member of these populations, the implications of the study may very well apply to you.  
    • If you are outside the age range, have different work environments (if you even have a job), you may show a different sensitivity to lifestyle change
    • Also, the statistics from the study show the participants as around 97% white.
  2. The data:  The researchers did a few things to try to reduce noise. But there may still be some things that are hard to avoid. A lot of the information comes from self-reported assessments. And people lie, or at the very least, don't remember well.
  3. The methodologies: 
    1. When they went to apply their findings to the general population, they used a study called NHANES, which is a snapshot of food choice.  It doesn't  reflect long term dietary patterns. 
    2. Speaking of diet, they also used one specific measurement of food choice health.  There are probably other assessment options, but you have to choose something.
    3. They don't have any analysis about the different compositions of risk factors from 1-4. Is moderate drinking and not smoking better or worse than excising and not smoking?
  4. Flies in the soup.  There may be more stuff going on than what was observed or called out.  Maybe one indicator of lifespan is your honesty in a survey! Most of the study involves some form of consumption (alcohol, food, smoking). Sleep, stress, family-life, genetics, and many more things may also show up as having the potential to improve your longevity.

A New Headline

It is easy to be a critic and hard to be a researcher.

So after all this analysis, where does it leave us? 

If you smoke, stop.  This is well known, but reaffirmed here.

Otherwise, eating more balanced meals and adding movement to your daily activities seems to have benefits that could result in tacking on a few more years to your life. To boil this down to a front page headline:


The new headline may seem a bit watered down, but it is at least more honest in its impact and scope.

Have a better suggestion? What did you think of this analysis?