5 ms·

> Exercising (moderate physical activity) for 2-2.5 hours per week leads to a 20% reduction in all-cause mortality (not to mention its many other benefits). Exe

by zeteo 10y ago

> Exercising (moderate physical activity) for 2-2.5 hours per week leads to a 20% reduction in all-cause mortality (not to mention its many other benefits). Exercising for an additional 5.5 hours leads to another 20% reduction. With a very important caveat in the report: > Because all of the studies in the evidence base were observational epidemiologic studies with no randomized controlled trials, the data cannot prove causality of effect. People who don't feel that great when they exercise (despite everyone urging them to exercise) are probably not in the best health. The pervasive advice to exercise more will by itself confound the studies, because healthier people will tend to self-select into the "exercise" group.

QSIITurbo 10y ago
Yeah, it has its weaknesses and also from my comment below: "So I read the study for fun and the non-adjusted rates are 3.84% death rate (cohort study) in no exercise vs. 2.35% in 1-30 minutes and 2.08% in more. And that's not adjusted for pre-existing conditions if I understood correctly. So meh. Even though 20% and 20% sounds really great!" I didn't find anything there for longer periods of time but odds ratio is not a great indicator if your base chance of dying is already really low. Or maybe spend the whole week exercising to squeeze that last 0.1% out of your chance of dying.
- ramblerman 10y ago
  3.84% to 2.08% is a 45% reduction in the amount of people that died. I don't understand your point, there is no obfuscation here, that's just how percentages work.
  - QSIITurbo 10y ago
    Um, no. 3.84% to 2.08% being an approx 40% reduction is exactly what is being avoided currently in research since its reduction of less than 2 percentage units. Imagine a medication that has a "20% and 40% reduction in mortality": You'd have to have more than 150 people treated to the max to save one person vs other cases (more than 75 to save vs no medication). Not 5 as your percentage implies. Not very efficient way to spend money if you ask me.
    - ramblerman 10y ago
      (3.84 - 2.08) / 3.84 = 45% less people dead It's not 'my' percentage, nor subjective. I mean you could argue the study is biased or the sample size too small, but your understanding of percentages seems flawed.
      - QSIITurbo 10y ago
        I recommend you take basic university statistics. Especially study efficacy vs effectiveness so you might have an idea how you're being mislead here.
joshgel 10y ago
I agree with your comments entirely. I'll also add, however, that even though this is cohort data 1) there is A LOT of data, almost all of it in agreement for hundreds and hundreds of studies, over millions of person years and 2) there is pretty strong evidence of a dose-response relationship. Both of those two points help (though don't entirely eliminate) the weaknesses inherent in cohort data.
- QSIITurbo 10y ago
  It doesn't change the fact that if you take approx 150 from each of the activity groups, the cohort study will tell: Approximately 5 from the "slobs" will die. Approximately 4 from the "moderate exercise group" will die. Approximately 3 from the "forrest gump group" will die. On a personal level, to me these groups are more or less the same in terms of mortality, but somebody marketed the 20% really well.
  - zeamaize 10y ago
    Well if you truly don't understand math, please transfer 20% of your wealth to me. I mean, from your perspective it's the same amount of wealth, right?
    - tudorw 10y ago
      that's not the point, the posters point is that the way the math was used added dramatic effect, it did not aid in understanding the meaning of the findings.
      - QSIITurbo 10y ago
        Exactly. 20% or 40% sounds really big. In reality, you're maybe lowering your chance to die (during some time) by less than 2% by exercising almost constantly. Now, if you're not enjoying exercise, or feel exhausted after exercising, you actually reduce the amount of "quality life" by exercising. If exercise was a drug, they'd never allow it on the market.
sqeaky 10y ago
There are are plenty of well understood mechanisms for how exercise can improve health. There are way may plausible but not yet well understood ways it can help. Clearly any error from the self selection you described does not counter an demonstrable mechanism.
- kbutler 10y ago
  What if the study compared bedridden 100-year-olds as non-exercisers to the college track team? The track team was 20% less likely to die in the subsequent 10 years... Self-selection can completely override other well-understood mechanisms...
  - brutos 10y ago
    These studies are usually done as cohort studies. Meaning the study group shares a defining trait (such as the same date of birth). Obviously this is not as great as a double-blind study. But you can't let one group do real exercise and the other one placebo exercise. So the cohort study is probably the best possible study design in this case. The people doing these studies try very hard to minimize biases (sometimes failing, but ruling out every possible biases is hard).
woliveirajr 10y ago
This. And even your willing to exercise changes during live. It was easier and fun to exercise when was a teenager and so on. Become lot less when gained almost 40 pounds in a month due to medication. Become even less interesting when would get hurt for exercising while overweight. Become a no-go when got a muscle strain for simply going downstairs because a physiotherapist pushed a bit too much and was under muscular fatigue. I'd prefer to take a bath and have some benefits that fell all that pain again and immobilize my leg and loose what I've earned in the past 4 months. For those who already like to exercise (or are willing to despite saying "I don't like, just like the benefits", or any kind of wording of this idea), well, this study won't make a difference. For me, yes, it makes.
acchow 10y ago
This is exactly why I can't trust prescriptive conclusions from correlation studies. Didn't Google publish some method on resolving causation in correlation? Something about transfer of noise?

by zeteo 10y ago

QSIITurbo 10y ago

Yeah, it has its weaknesses and also from my comment below: "So I read the study for fun and the non-adjusted rates are 3.84% death rate (cohort study) in no exercise vs. 2.35% in 1-30 minutes and 2.08% in more. And that's not adjusted for pre-existing conditions if I understood correctly. So meh. Even though 20% and 20% sounds really great!" I didn't find anything there for longer periods of time but odds ratio is not a great indicator if your base chance of dying is already really low. Or maybe spend the whole week exercising to squeeze that last 0.1% out of your chance of dying.

ramblerman 10y ago
3.84% to 2.08% is a 45% reduction in the amount of people that died. I don't understand your point, there is no obfuscation here, that's just how percentages work.
- QSIITurbo 10y ago
  Um, no. 3.84% to 2.08% being an approx 40% reduction is exactly what is being avoided currently in research since its reduction of less than 2 percentage units. Imagine a medication that has a "20% and 40% reduction in mortality": You'd have to have more than 150 people treated to the max to save one person vs other cases (more than 75 to save vs no medication). Not 5 as your percentage implies. Not very efficient way to spend money if you ask me.
  - ramblerman 10y ago
    (3.84 - 2.08) / 3.84 = 45% less people dead It's not 'my' percentage, nor subjective. I mean you could argue the study is biased or the sample size too small, but your understanding of percentages seems flawed.
    - QSIITurbo 10y ago
      I recommend you take basic university statistics. Especially study efficacy vs effectiveness so you might have an idea how you're being mislead here.

joshgel 10y ago

I agree with your comments entirely. I'll also add, however, that even though this is cohort data 1) there is A LOT of data, almost all of it in agreement for hundreds and hundreds of studies, over millions of person years and 2) there is pretty strong evidence of a dose-response relationship. Both of those two points help (though don't entirely eliminate) the weaknesses inherent in cohort data.

QSIITurbo 10y ago
It doesn't change the fact that if you take approx 150 from each of the activity groups, the cohort study will tell: Approximately 5 from the "slobs" will die. Approximately 4 from the "moderate exercise group" will die. Approximately 3 from the "forrest gump group" will die. On a personal level, to me these groups are more or less the same in terms of mortality, but somebody marketed the 20% really well.
- zeamaize 10y ago
  Well if you truly don't understand math, please transfer 20% of your wealth to me. I mean, from your perspective it's the same amount of wealth, right?
  - tudorw 10y ago
    that's not the point, the posters point is that the way the math was used added dramatic effect, it did not aid in understanding the meaning of the findings.
    - QSIITurbo 10y ago
      Exactly. 20% or 40% sounds really big. In reality, you're maybe lowering your chance to die (during some time) by less than 2% by exercising almost constantly. Now, if you're not enjoying exercise, or feel exhausted after exercising, you actually reduce the amount of "quality life" by exercising. If exercise was a drug, they'd never allow it on the market.

sqeaky 10y ago

There are are plenty of well understood mechanisms for how exercise can improve health. There are way may plausible but not yet well understood ways it can help. Clearly any error from the self selection you described does not counter an demonstrable mechanism.

kbutler 10y ago
What if the study compared bedridden 100-year-olds as non-exercisers to the college track team? The track team was 20% less likely to die in the subsequent 10 years... Self-selection can completely override other well-understood mechanisms...
- brutos 10y ago
  These studies are usually done as cohort studies. Meaning the study group shares a defining trait (such as the same date of birth). Obviously this is not as great as a double-blind study. But you can't let one group do real exercise and the other one placebo exercise. So the cohort study is probably the best possible study design in this case. The people doing these studies try very hard to minimize biases (sometimes failing, but ruling out every possible biases is hard).

woliveirajr 10y ago

This. And even your willing to exercise changes during live. It was easier and fun to exercise when was a teenager and so on. Become lot less when gained almost 40 pounds in a month due to medication. Become even less interesting when would get hurt for exercising while overweight. Become a no-go when got a muscle strain for simply going downstairs because a physiotherapist pushed a bit too much and was under muscular fatigue. I'd prefer to take a bath and have some benefits that fell all that pain again and immobilize my leg and loose what I've earned in the past 4 months. For those who already like to exercise (or are willing to despite saying "I don't like, just like the benefits", or any kind of wording of this idea), well, this study won't make a difference. For me, yes, it makes.

acchow 10y ago

This is exactly why I can't trust prescriptive conclusions from correlation studies. Didn't Google publish some method on resolving causation in correlation? Something about transfer of noise?