The storm continues to rage. Congress has issued an angry report on Avandia in which neither GlaxoSmithKline nor the FDA was spared. The results of these clinical trials were not surprising. Numerous clinicians have been discussing cardiac events for all of the thiazolidinediones (TZD) as a drug class for at least a decade. Rosiglitizone (Avandia) was issued a black-box warning in 2007 about the possibility of cardiac events although it included a statement that these data were still inconclusive. As far as I can tell there seems to be two camps. On one side are the clinicians who are focused on the data which show that rosiglitazone causes more heart failure events than other blood glucose lowering therapies and on the other side are the clinicians who are focused on the similarity in numbers of deaths and of hospitalizations among the groups.
Having looked at the data myself I firmly believe that indeed rosiglitazone does have a greater cardiac risk. However, I have also been struck by the numbers for more general measures of hospitalizations and deaths. Here is my reasoning as to why this is of interest. All drugs have risks and benefits which may not be readily apparent. The only way we can assess these unknown risks and benefits is to look at a generalized measure such as death or hospitalization by any cause and to carefully stratify the patient population so as not to be comparing apples and oranges.
Dr Juurlink and colleagues, in the Canadian retrospective study I discussed in the last post, looked at exactly this sort of measure. Just to remind you, Dr. Juurlink and colleagues looked at several thousand people of age 66 or older who had been prescribed either Actos (pioglitazone) or Avandia (rosiglitazone) and followed their medical records for a number of years. The graph of the data maps the proportion of the Actos treated population that had died over 3 years from the start of Actos treatment and compares this to the proportion of the Avandia treated population that had died over the 3 years since the start of that treatment. The two lines do diverge and it does look like the Avandia group had more deaths. Here are the numbers that I gleaned from making measurements from the graph provided (figure 4 from the report). The Actos treated group showed a consistent almost linear increase in the number of deaths from day 1 to year 3 with a total of about 7.2% of the group dying from any cause. The Avandia group, after about 6 months, began to show some increased deaths and this increase bounced around somewhat. In other words, the line was more “squiggly”. At its greatest separation from the Actos group there were about 0.9% greater number of deaths. This happened just before year 2 and again around year 3. At its closest approach the two populations showed a difference in deaths of around 0.15% of the population. To my eye, the two lines were not diverging but seemed to be tracking each other in a roughly parallel fashion. I say this because the distance between the two lines was about the same just before year 2 and again at year 3. The authors have reported “One additional composite outcome would be predicted to occur annually for every 93 patients treated with rosiglitazone rather than pioglitazone.” This would continue to be true if the lines do not dip again and the two populations retain that 0.9% difference. If the distance between the two lines dips to 0.15% again we would be looking at one additional composite outcome (death) for every 1500 patients. In my read of the data, this is the range of the risk between Actos and Avandia: 1 in 93 or 1 in around 1500.
In the RECORD trial (I’m looking at the data published in the Lancet now) the graph from all cause death is virtually the opposite of that seen in the Canadian retrospective trial. Now keep in mind that this was a different trial. It compared 3 groups: rosiglitazone plus a sulfonylurea, rosiglitazone plus metformin, or the combination of a sulfonylurea plus metformin. The two rosiglitazone groups were combined. Also the patients were younger. At any rate the all cause death graph looked at 6 years of data. By year 6 about 7% of the sulfonylurea/metformin group had died while the combined rosiglitazone groups had a death rate of about 6%. Again, the 2 lines were not diverging so much as rising together – sometimes getting closer together and sometimes getting farther apart.
How should we interpret these numbers? We know that heart disease is one of the major causes of death today. One would think that if a medication doubled the rate of cardiac events we should see a rather large difference in the all deaths measurement. We do not. This may be due to the complexities of life and death drowning out the specific numbers of cardiac deaths in a sea of other events. Alternatively, if we have faith in the power of our statistical analysis, perhaps these numbers are saying something else about rosiglitazone.
Just to play devil’s advocate here, what if rosiglitazone provided some additional protection to patients in addition to its role in dropping blood sugar levels? The truth of the matter is that we already are aware of at least one additional function for this class of drugs. They provide a strong protection from lung cancer.
The data for lung cancer protection consists of both analyses of human clinical trials as well as animal models. For example, a study headed by Dr. Rangaswamy Govindarajan retrospectively examined the records of US veterans with diabetes from 1997 to 2003, amounting to 87,687 individuals. The results were published in the Journal of Clinical Oncology in 2007. They found that use of either of the current TZD drugs on the market; Actos or Avandia, resulted in a 33% drop in lung cancer rates. They also looked at colorectal and prostate cancer rates and found them to be only slightly changed and this was below the level of statistical significance. When race was considered, African American males did especially well. As stated in the study, “After adjustment for age, BMI, HgbA1C, insulin use, and the use of other oral antidiabetics, the TZD-associated risk reduction for lung cancers was 26% among white and 62% among African American patients.”
We do not yet understand exactly how TZD drugs protect us from lung cancer. Indeed that is the point. There is still a lot about the TDZ drugs that we do not understand. Thus there is ample room for the possibility of additional protective functions for these drugs and we do not yet know whether Actos or Avandia will be better at those protective functions.
Perhaps the best way to end this particular discussion is to invoke the concept of balance. Therapeutics of all sorts cause changes in our physiology which acts in both beneficial ways and in ways that increase our risk. We can mitigate some of that risk through changes in behavior. In the case of the TZD drugs, for example, mild exercise has been shown to be efficacious in reducing cardiac risk. We can even extend this balancing act to include societal issues. If a doctor knew that their patient well enough to know that they exercised regularly, then a TZD prescription would be less of a concern. Ultimately, personalized medicine will help us to tailor therapies to minimize risk. When we understand that complex dance between nature and nurture that creates that which we are, we will know who is at greater risk of heart attack and who is at greater risk for cancer. Until that time we can only place our bets and take our chances.