Tuesday, July 5, 2011

Chantix should not be withdrawn.

Despite the rising rates of obesity, smoking is still the single leading cause of preventable death in the US.  Quitting smoking is very difficult because one needs to address both the behavioral and pharmacologic aspects of nicotine addiction  Though other agents are available, Chantix or varenicline is the most effective agent to assist in smoking cessation.  This has been proven in several large, randomized clinical controlled trials (RCT's). RCT's are the gold standard when it comes to scientific proof. (More on that in a minute).

Chantix is not without issues.  The main side effect is nausea, which about 30% of users will get.  It is usually mild and usually goes away, though a small percent of people will not tolerate this.  The more recent concern with Chantix was exacerbations of neuropsychiatric symptoms: depression, anxiety, and suicidal ideation.  These side effects were not seen in many of the initial studies, but in later on via reports by doctors and patients after the drug was on the market.  This method is called post-market surveillance. Post-market surveillance is critical in determining drug safety, because rare but serious side effects may not be seen when you only study thousands not millions of patients.  However, unlike RCT's , proving cause and effect can not be determined. In regards to psychiatric symptoms, in the original studies that the Pfizer submitted to the FDA, Chantix had few interactions and did not show any. However, because Pfizer compared Chantix to bupropion (the only other pill indicated for smoking cessation, but also used for depression), patients with mental illness were purposely excluded from the study. (See more about this in Where's the Good News about Chantix? and More FDA warnings should not be cause for worry.) In fact, since these warnings first appeared, further studies seem to indicate that just stopping smoking, not necessarily Chantix, can cause these problems.  Furthermore, warnings were not just added to Chantix but also to bupropion. Regardless or whether symptoms like depression or even suicidal thoughts is linked to Chantix or stopping smoking, doctors and patients should be aware of this concern for any patient quitting tobacco. 

Now we have a new concern regarding Chantix causing cardiovascular events.  The initial concern was raised by the FDA in their review of a study of 700 patients with known cardiovascular disease randomized to Chantix or placebo. Though Chantix was far more effective in helping these patients with known heart disease quit tobacco, there was a small number of increased cardiovascular events, more in the Chantix group than placebo. The total number of events was 28 in the Chantix group vs. 17 in the placebo.  The study was not designed to show whether or not this number was statistically significant (was really true), but the FDA added a warning to Chantix' label.

However, a new study raises questions about Chantix and people without heart disease.  This is being blasted all over the media.  The Wall Street Journal reports "Drug Tied to Heart Risks."  The New York Times reports "Study Links Smoking Drug to Cardiovascular Problems."  ABC News states "Chantix: Quit Smoking, But Risk Your Heart?" All of them have similar language to the ABC news site stating:

Study authors looked at 14 past studies of Chantix and found that overall, people on the drug had a 72 percent increased risk of being hospitalized with a heart attack or other serious heart problems when compared with those taking a placebo.

That seems pretty bad! Unless, of course, you look at the actual data. The new study is a meta-analysis of studies looking at patients on Chantix without cardiovascular disease.  The study is from the Canadian Medical Association Journal.  They looked at data from 14 RCT's 8216 participants. They found that Chantix was associated with a significantly increased risk of serious adverse cardiovascular events compared with placebo on 1.06% [52/4908] in varenicline group compared to  0.82% [27/3308] in placebo group.  In other words, the absolute difference between Chantix and placebo is 0.24% or 24/10,000. This sounds a lot less scary than 72% increase (relative increase) being reported in the media.  It is also very, very important to note that the technique used to derive these numbers is a statistical technique, a meta-analysis, which is not nearly as rigorous as an RCT.  Meta-analysis are designed to ask questions, not to answer them.  ( I have blogged previously about the pros and cons of meta-analysis). Furthermore, even patients without a history of cardiovascular disease who smoke, have a risk for cardiovascular disease, which is why they need to stop in the first place.

Bottom Line: One must always question 1) the results of a meta-analysis because it has many limitations and 2) any non-RCT, especially a meta-analysis, with a very small absolute risk (i.e. 0.24%) especially if the authors/journalists are trumpeting a large relative risk (i.e. 72%) and  also 3) take into account the context of the situation, i.e. the single best agent we have available for the leading preventable cause of death in the United States, might possibly have an associated very small increase in heart disease in smokers that will likely have a much greater risk of heart disease if they don't stop smoking. While I agree more research is needed, and warnings about a possibly increased cardiovascular risk are not inappropriate, pulling Chantix from the market would be a huge mistake.


Consejos de belleza para mujeres said...

Directly and indirectly, smoking poses a danger to the public. An indirect public health concern that cigarettes may pose is accidental fire. Cigarette smoking, mainly attacks the cardiovascular system, leading to a heart attack, respiratory diseases and lung substance

Rebecca said...

Dr. Mintz, I like your focus on the actual numbers, but, like most physicians, you seem confused about sources of precision in clinical trials. I'm not sure where you are getting your very definite ideas on the differences between "Meta analysis" and "randomized trials" per se -- but there' a bit off. The issue isn't whether "randomized clinical trials" are more believable than "meta analysis". The issue is the size of the effect you are trying to detect with the randomized evidence, and the number of people that have been randomized, either in one trial or in many. perhaps you should look up the Early Breast Cancer Trialists Collaborative group -- hundreds of investigators looking at the totality of the worldwide randomized evidence. The results of proper analysis of the totality of the worldwide randomized evidence are, in fact, much more reliable than the results of any individual trial. In fact, some of us would go so far as to say that any single trial is more appropriately regarded as a subgroup of the totality of the worldwide randomized evidence.

Dr. Matthew Mintz said...

Your points are very valid.
The point of the post was that an absolute risk of .24% from a meta-analysis is not cause for alarm. However, you are correct that a large, well done meta-analysis can provide reliable data and a less than large RCT is sometimes wrong.
In this particular study, the looked at any RCT with chantix that reported CV events as an adverse outcome. Though they looked at studies that reported "no CV events" as an outcome, it is very possible that in the 9 studies they excluded, a CV outcome was not reported because it didn't happen. Specially, none of these studies were designed to look for CV safety. Thus, even though combining the studies improves the power to detect a significant difference, if the studies were designed to specifically look at CV events, the meaning of the results comes into question. In addition the length of the studies ranged from 7 -52 weeks. Based on the mechanism of action of Chantix, I think that if a patient had a heart attack or stroke in the first few weeks of being on it, it probably wasn't related to the Chantix.
Well designed, well powered meta-analysis can be quite informative. Unfortunately, what I have seen recently is that someone who is trying to prove a point does a meta-analysis with certain selection criteria and statistical techniques to prove their point, which they end up doing. In addition, these types of studies tend to show a large relative risk, when the absolute risk is small. The Avandia meta-analysis is the perfect example. An analysis of the same studies using a different statistical technique found no harm with Avandia, and the absolute risk for for Avandia was 0, though the relative risk was 42%. This poorly done, meaningless meta-analysis drew a lot of attention, pulled a good drug off the market, and had made it almost impossible for any new diabetes or obesity drug to get approved.