Communicating risk/benefit effectively

I’ve written before about how AP/NP advocates misperceive the risks involved in various parenting practices. Aside from the pro-“natural” prejudice they display, I think part of the problem is that the information in medical research – and especially in abstracts, which is what the dons of AP Internet U usually read – is presented in a way that can mislead even professionals, much less the layperson, about the risk or benefit of a given medical intervention or treatment. You see it quite a lot in the materials drug reps present to doctors, for example – “drug X caused a 50% reduction in the incidence of disease Z!”.

The problem here is that that “50% reduction” doesn’t take into account whether the risk reduction is absolute or relative. If drug X reduces the incidence of disease Y from 51% of the population to 1%, that’s an absolute risk reduction (ARR) of 50%; if it reduces the incidence from 2% to 1%, that’s a relative risk reduction (RRR) of 50%. Obviously, the former is a far more impressive drug than the latter. Most drugs and interventions (including “natural” interventions like breastfeeding) are of the latter type, however. Which is why communication of the risks and involved to the public should, ideally, give some indication of the actual health impact of the intervention, the better for the individual and his/her healthcare practitioner to assess the real risks and benefits.

Unfortunately, most medical studies report their results in relative terms such as RRR’s or odds ratios (the latter being an oft- misunderstood concept). Is there perhaps a better way to accurately convey health risks and benefits?

Absolute Risk Reduction (ARR): This is probably the most straightforward and most understandable method of conveying risk. If, say, 50% of formula-fed infants develop an ear infection in their first year, as opposed to 40% of infants who breastfed for 4 months*, the ARR would be 50-40=10%.

The main problem with using the ARR in this way is that unlike drug vs. placebo or other randomized controlled trials (RCTs), you can’t randomize two groups which practice certain parenting methods, as these are usually self-selected (e.g., babies to breast- or bottle-feeding) nor control for various confounding factors which affect the risk profile of the two groups. This means you can only convey results not adjusted for various confounding factors (factors which can be adjusted in odds and risk ratio calculations).

NNTs and NNEs: The NNT (number needed to treat) expresses the number of patients who need to be treated (usually for a certain amount of time) in order to prevent one bad outcome. The smaller the NNT, the more effective the treatment. the NNT is the reciprocal of the ARR (1/ARR). For example, for the example above (regarding breast feeding and ear infections), NNT=1/0.1 = 10. You can also (by a more complicated method) calculate NNTs from odds ratios if you have other information.

A sampling of NNTs for various common (and less common) treatments.

A more user-friendly article discussing the significance of NNTs

Strictly speaking, NNTs, like ARRs, can only be properly caclulated when dealing with randomized controlled trials. Otherwise, the basic differences between the groups can distort the magnitude of the NNT – either way, depending upon the different characteristics of the treatment and the control groups. However, a related concept, the NNE (number needed to be exposed) can be used in studies where the two groups are not randomized and there are confounding variables to be addressed. Also called “adjusted NNT” in some studies (though this term has been used in other ways as well).

Population impact number of eliminating a risk factor over a time period (PIN-ER-t): A fancy way of saying “How many cases of disease would be avoided if everyone did things the healthier way?”. Here’s a discussion of this concept as regards communicating the benefits of breastfeeding to the British public. As the authors themselves point out, this assumes complete causality between breastfeeding and the given health outcome, which may not be true – it’s not certain that even in the unlikely event 100% of women breastfeed, we’ll actually be able to avoid the number of diseaqse cases predicted by the PIN-ER-t. There may be other confounders not accounted for in the research.

Of the 3 methods, I think the most practical and accurate way of conveying risk is the NNT/E. What about you?

—————————————————————————————-
* Percentages do not represent actual study results, just figures made up for the purpose of demostrating the mathematical concepts.

add to del.icio.us : Add to Blinkslist : add to furl : Digg it : add to ma.gnolia : Stumble It! : add to simpy : seed the vine : : : TailRank : post to facebook

Advertisements

One Response

  1. I think I need to brush up on statistics!!!!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: