Wednesday, December 7, 2011

Free Drugs For Heart Attack Patients: The Analysis Behind the Analysis of the MI-FREEE Trial

Assuming drugs are not free, should all patients in an insurance plan that covers medications get the same coverage at the same price?

While that may seem to be a no-brainer, there's plenty of research (for example) that demonstrates that out-of-pocket costs can reduce persons' willingness to take their pills as prescribed.  While that may be the price of doing business, why not give persons who really need a particular life-saving medication a price break?  While that may seem unfair, suppose everyone in the risk pool benefits from lower health care costs?

Enter the Post-Myocardial Infarction Free Rx Event and Economic Evaluation (MI FREEE) trial

Too long to read at one sitting you say?  The Disease Management Care Blog at your service!

The study involved Aetna beneficiaries who had just been discharged from a hospital following a heart attack.  While the trial was randomized and prospective, the randomization occurred at the level of the insurance plan. Various employer, union, local government or other association groups (and their patients) were randomized to one of two arms:

1) an intervention group where patients had no out of pocket cost sharing or co-pays for brands or generics in four classes of drugs that have been shown (go to page e227) to reduce the risk of death after a heart attack: 1) statins, 2) beta blockers, 3) ACE inhibitors and 4) ARBs, or

2) the usual co-pay for statins, beta blockers, ACEs and ARBs.

2845 persons were placed in the "no-cost" arm of the study and 3010 were in the "usual cost" arm of the study.  The mean age was 53 years and 75% were men.  About 34% and 27% of both groups had diabetes and heart failure, respectively.  Over time, the percent of persons that were fully compliant with their medications became different: 31% in the usual cost vs. 41% in the no-cost group.  The median duration of follow-up was 394 days.

And what happened?  When the number of first-time fatal and nonfatal cardiovascular events were grouped in and counted with heart surgeries (that included angioplasties, stenting or open bypass), there was no statistical difference between the two study arms: 18.8 events per 100 person-years "usual cost" vs. 17.6 per 100 person-years in the "no-cost" group.  There was no difference in the cardiovascular death rate either: 2.0 vs. 1.7 deaths per 100 person years.

However, there was some good news.  The combined endpoint similarity was largely driven by a high equal number of heart procedures in both groups (which seemed to involve more than 10% of the entire study cohort).  If the surgery patients are backed out, there were fewer first time fatal or non-fatal vascular events and strokes in the "no-cost" group. 

There is even more good news. Some patients had more than one event (a patient could have a heart attack, a stroke and then open heart surgery, for example). When the total number of events was added up, there was a difference and it was statistically significant: 329 per 100 person-years in the "no-cost" group vs. 406 per 100 person years in the "usual cost" (p=.03).

Did the insurers save any money?  Yes and no.  Average total spending in the "no-cost" group tallied up to $18,254, while the "usual cost" group averaged $20,238.  While that's a $2000 difference per patient, it failed to achieve statistical significance. However, when the DMCB multiplies those savings by the number of persons in the no-cost treatment arm, it calculates $5,690,000 in total savings. That sure sounds financially significant. 

DMCB criticisms:

1.  Heart procedures - which may be prone to factors other than clinical need - may have diluted the results in this trial, especially if a significant proportion of them were not evidence-based.

2.  The secondary prevention benefit of drugs for heart attack patients extends beyond 394 days.  If this study had gone longer, the difference may have expanded over time and achieved statistical significance.

3.  While a cost difference of $2000 did not reach statistical significance, that may have been because insurance claims follow a non-Gaussian distribution, making their analysis very tricky.  In addition, the DMCB thinks that a real world savings potentially exceeding $5 million is very noteworthy.

DMCB questions:

1. It'd be nice to know if patients with a higher burden of disease (for example heart failure or diabetes) and therefore more vulnerable benefited more than persons with less disease burden.  If so, would it make sense to limit the "no-cost" option to heart attack patients at the highest level of risk? 

2. There is no information on the level of out-of-pocket costs in the usual-pay group. As co-pays increase, medication adherence goes down. We don't know if Aetna's pharmacy benefit is typical of the rest of the market.

DMCB insights:

1. The possibility of $5 million (or more if compliance can be increased beyond 41%) in savings may make paying patients to take their pills seem reasonable.  Silly you say?  Think again.

2. Since $5 million in cost reductions that can result from just a 10% swing in medication compliance, readers should gain a better appreciation on the stakes behind disease management.  If nurses can talk patients into taking their pills, the downstream savings can be potentially huge.  And why stop there, why not combine no out-of-pocket costs with disease management?


While the authors dutifully report that the primary outcome of the study (the number of first non-fatal or fatal heart attacks or some type of heart procedure) was the same in both groups, the DMCB remains impressed that free drugs for heart attack patients may be worth it and that Oncle Karl may have been right and that the University of Michigan VBID folks are on to something.  That's because the total number of events achieved statistical significance and there were some specifics that may have blunted the study's ability to get at the other outcomes.

Image from Wikipedia

No comments: