For "A" Students - Another Useful Tool for Weighing Risk/Benefit
The Cost of Prevention
When thinking about the potential benefit of competing treatment options, one has to consider both the effectiveness of therapy and its cost. An interesting way to think about this is to calculate the number of people you would need to treat in order to prevent one adverse outcome. If you also know the cost of treatment, it is easy to calculate the cost of preventing one adverse outcome with a given treatment or preventive strategy. If you were to calculate this for competing preventive strategies, you would have a convenient way of comparing their cost effectiveness.
You already have the tools to do this. Suppose you were interested in preventing cardiovascular disease in people who were classified as having an increased risk. Statins
are a class of drugs that have been demonstrated to be effective in lowering blood levels of cholesterol and significantly reducing the incidence of major cardiac events (heart attack, stroke, severe angina) in patients with elevated cholesterol levels. However, certain groups of people who do not have elevated cholesterol levels are still at increased risk of having a major cardiac event, including people who have elevations in an inflammatory marker called C-reactive protein. In Nov. 2008 the New England Journal of Medicine published the results of a study in which the investigators enrolled 17,802 subjects who had no history of heart disease. All subjects had elevated levels of C-reactive protein, but they all had normal cholesterol levels. Subjects were randomly assigned to receive either the statin Rosuvastatin (Crestor) 20 mg. per day or a placebo that looked identical to the active agent. The drugs were coded, and neither the investigators nor the subjects know who was receiving the active drug. Subjects were followed for an average duration of about two years.
Several points from the Methods section of the paper:
- "Follow-up visits were scheduled to occur at 13 weeks and then 6, 12, 18, 24, 30, 36, 42, 48, 54, and 60 months after randomization. Follow-up assessments included laboratory evaluations, pill counts, and structured interviews assessing outcomes and potential adverse events."
- "The primary outcome was the occurrence of a first major cardiovascular event, defined as nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, an arterial revascularization procedure, or confirmed death from cardiovascular causes."
- "All reported primary end points that occurred through March 30, 2008, were adjudicated on the basis of standardized criteria by an independent end-point committee unaware of the randomized treatment assignments. Only deaths classified as clearly due to cardiovascular or cerebrovascular causes by the end-point committee were included in the analysis of the primary end point."
Other important information: CanadianDrugs.com is has been selling 20 mg. Crestor for about $2.00 (US) per pill (i.e., $2.00 per day to treat). This was the lowest cost source that I was able to identify.
The following table summarizes the main findings of the study.
| Group |
Subjects |
Major Cardiac Events |
Person-years at risk |
|---|---|---|---|
|
Crestor (a statin) |
8,901 |
142 |
18,442 |
|
Placebo |
8,901 |
251 |
18,529 |
One could easily compute the rate ratio:
Rate Ratio = (142/18,442) / (251/18,529) = 0.57
The 95% confidence interval for the rate ratio is 0-46-0.70, and the p-value < 0.00001.
However, to evaluate cost versus benefit, it would be more useful to consider how much it costs to prevent a single major cardiovascular 'event.' How would you compute this from the data shown in the table and knowing that the cost of Crestor therapy is about $2.00 per day?
One can easily compute this from the rate difference:
Rate Difference = (142/18,442) - (251/18,529) = - 0.005847 per person-year
Since the result is a negative number, I can interpret this as a reduction in risk of about 58 major cardiovascular events among 10,000 treated persons over a year. In other words, if we treated 10,000 such subjects with statins for one year, we could expect to prevent 58 major CVD events.
"Number Needed to Treat"
Another way of thinking the rate difference is to consider how many people one would have to treat for a year in order to prevent a single CVD event. This is often referred to as the "number needed to treat" or NNT.
If 10,000 treated subjects prevented 58.47 events, then the number that would need to be treated to prevent one event is
NNT = 10,000 treated for a year / 58.47 = 171 treated for one year to prevent one event
Note that the NNT is simply the reciprocal of the rate difference for a year, and note also that NNT is conveniently calculated for you in EpiTools.XLS in the worksheet for cohort.-type studies.
Rate difference = 58.47 / 10,000 over a year
NNT = 10,000 over a year / 58.47 = 171
Finally, if one needs to treat 171 people for a year to prevent one major CVD event, then the cost of preventing one such event is:
171 x $2.00/day x 365 days = $124,830 per year to prevent one major event
And the cost of treating 10,000 such persons would be $7,300,000 per year.
With these calculations in mind, consider the management of a 50 year old who has normal cholesterol levels, but elevated C-reactive protein. This individual would, of course, have to be treated for many years. Would you support or recommend long term treatment of such individuals with Crestor? Why or why not?