Attributable risk vs ARR

[ChessMaster3000]

Advertisement - Members don't see this ad

I was hoping someone could clarify something for me.

In terms of calculations, are attributable risk and absolute risk reductions the exact same thing? Is the only difference that we use AR for cohort studies (and therefore relative risk) and ARR for case-controls? Likewise, is the number needed to treat/harm the same thing as well (NNT for case-control and NNH for cohort studies)?

thanks!

 

[ulikedaggers]

Yeah, calculation is exactly the same. AR is for exposed vs unexposed and AAR is for treated vs untreated.

And yeah, since NNT and NNH are functions of AAR and AR, respectively, they would also be the "same" thing.

It's pretty stupid that they have to be called different things.

 

[ChessMaster3000]

Yeah, calculation is exactly the same. AR is for exposed vs unexposed and AAR is for treated vs untreated.

And yeah, since NNT and NNH are functions of AAR and AR, respectively, they would also be the "same" thing.

It's pretty stupid that they have to be called different things.

Thanks a lot--I was just like, WTF is the difference? And then of course pissed because I thought I didnt understand it, when in fact I do.

 

[NeuroLAX]

I thought the same until I realized why they're differentiated. They might be calculated the same way, but they are saying two different things.

 

[ulikedaggers]

I thought the same until I realized why they're differentiated. They might be calculated the same way, but they are saying two different things.

Yea, but I look at it as they are saying the same thing about exposure in one case and treatment in another case. And, if you consider being treated as "being exposed" then AR and AAR are essentially the same in non-statistical terminology.

 

[NeuroLAX]

Yea, but I look at it as they are saying the same thing about exposure in one case and treatment in another case. And, if you consider being treated as "being exposed" then AR and AAR are essentially the same in non-statistical terminology.

Fair enough. Correct me if I'm wrong, but I think the easiest way to conceptualize them is the equation: (I exposed) - (I unexposed).

Label exposed and unexposed according to the scenario that you are given. Increased risk vs decreased risk will dictate which combination you need to use- AR and NNH vs ARR and NNT, respectively. That is to say, if you're dealing with an exposure that increases risk (e.g. thalidomide), then you need to be quantifying this relationship in terms of AR and number needed to hARm (NNH). If you're dealing with an exposure that decreases risk (e.g. a vaccine), then you need to be quantifying this relationship in terms of ARR and number needed to treat (NNT).

 

[ulikedaggers]

Fair enough. Correct me if I'm wrong, but I think the easiest way to conceptualize them is the equation: (I exposed) - (I unexposed).

Label exposed and unexposed according to the scenario that you are given. Increased risk vs decreased risk will dictate which combination you need to use- AR and NNH vs ARR and NNT, respectively. That is to say, if you're dealing with an exposure that increases risk (e.g. thalidomide), then you need to be quantifying this relationship in terms of AR and number needed to hARm (NNH). If you're dealing with an exposure that decreases risk (e.g. a vaccine), then you need to be quantifying this relationship in terms of ARR and number needed to treat (NNT).

Yup, that's exactly what I do (and what Kaplan teaches). I just try to do whatever I can to simplify epidemiology because everything always seems so similar, but test questions usually want to make sure you can distinguish things.