I’m sure you will have seen the furore recently over Avastin. To summarise, the National Institute for Health and Clinical Excellence (NICE, and yes it irritates me that they left the H out. If they’d rearranged it they could have at least had NICHE) have ruled out funding the drug on the NHS, on the grounds of cost effectiveness. The media jumped all over this and NICE have been subjected to another PR whipping as a result.
Cancer is, of course, an emotive subject and of endless public interest. So stories like this get traction pretty easily. Generally, the media reported it by saying that Avastin is shown (incidentally, the article that links to made me punch my desk) to prolong life in colorectal cancer, but the meanies at NICE were keeping the purse closed. No sweeties for you.
To “prove” the point that NICE were just a bunch of ironically acronymed bastards, many reported on the case of one Barbara Moss. She paid the £21000 cost of Avastin herself and is, happily, still alive and indeed currently in remission. This is excellent news for her and her family, and I sincerely hope she continues in her relative good health. However, despite what she has stated and what has been widely reproduced in the press, she actually isn’t “living proof” that Avastin works. From an evidential point of view, she is a single data point.
The claim that was made in the newspapers, and probably worst of all by the Daily Express (which is why the link above made me punish my desk), was that Avastin grants sufferers an extra 2 years of life. And, of course, Barbara Moss did the rounds to make the case. Unfortunately this is bollocks. The irritating thing is that it is demonstrably false using the manufacturer’s (Roche) publication of the phase III trial for the drug.
So, what does this study actually report? Is Avastin (the trade name, by the way, for bevacizumab) a life-extender? Well, simply put, the effectiveness is actually fairly negligible, in a study of 1401 patients. Not one. 1401. Keep that in mind when you read about Barbara Moss.
There are two main things reported in the trial. One is what’s called the progression-free survival (PFS). As you might imagine, PFS is essentially the period in which the cancer doesn’t necessarily regress, but it doesn’t really spread or grow. For someone suffering from cancer, PFS is important to quality of life, as the more a tumour grows and metastasises the more unwell a patient will feel. The other thing that was measured was the median life expectancy (MLE). This is a measure of how long, on average, patients live. Interestingly this is the aspect the media fixated on and totally cocked up.
The trial showed that the median PFS for the placebo group was 8 months. For the Avastin group it was 9.4 months. The median life expectancy in the placebo group was 19.9 months. In the Avastin group it was 21.3. So. The first thing to mention is that these differences are pretty small for both measures. You’re certainly NOT getting an extra 2 years from Avastin, as The Express stupidly claimed. No, the difference is 1.4 months. But are these differences even real or are they just chance variation?
*If you really really really don’t want to read a very brief summary of the stats behind the paper, then skip the next 4 paragraphs, but I’ve tried to be gentle!*
As I mentioned in my post on the H1N1 thing, numbers come with various statistical disclaimers. Don’t worry, I won’t go mad on this but it is important! The important numbers here are: Hazard Ratio (HR), the confidence interval (CI) and the p-value.
I won’t go into massive detail on these, as that would be a sure-fire way to switch off your brain, but here are the highlights (for this case anyway): HR will measure whether something bad is more likely to happen if you take the drug. CI gives you a range of values that you can expect your real value to occur in (in this case a 97.5%CI). This essentially allows you to judge the precision of your result. The p-value basically tells you how sure you can be your result is real.
I’ll put the values in footnotes so you can see why I make my arguments here, but I promise not to go nuts with numbers! In both measures the HR indicates that there is a marginal improvement in the Avastin group. This means that people in the Avastin group were less likely to have a shorter PFS, and less likely to have a lower MLE. The CI for the PFS shows that this is probably a fair enough measure, and the p-value shows that it is a statistically significant result.
On the other hand, the rest of the stats for MLE are… shit. The CI shows that the HR could actually be taken out of the positive-effect zone. Admittedly only by a small way but it renders this measure meaningless. And the p-value? This is not a statistically significant result!
*Right, for those who skipped those 4 paragraphs, welcome back*
Basically what Roche’s own study shows is that Avastin has no measurable effect on life expectancy and a small effect on the PFS. This is not a ground-breaking drug. It is not very effective. Although cost-effectiveness sounds like a cold measure, it’s actually not. £21000 could pay for a nurse for a year. It could pay for a lot of effective treatments. Why waste it on a drug which is shown to have virtually no impact on the suffering or outcome of patients? Barbara Moss is indeed lucky that her disease has behaved as it has. But there is nothing to suggest that it was the Avastin.
With any group of patients suffering from any particular form of cancer you will get a spread of disease progression and survival times. Some patients will die after a short period. A few will live much longer than others. Most will occupy a middle ground. Just like any standard distribution.
So the moral of this story? Actually NICE aren’t penny pinching monsters, in my opinion. They have to justify the cost of a drug with measurable benefit to the patient population. Otherwise they may as well piss the NHS’ limited funds down the drain. They don’t want to deny people life-saving drugs. And Avastin is not one of those drugs.
Of course… the pharmaceutical companies behind the drugs do have a vested interest in people buying the drugs. And all this palava in the press would be music to their ears. Nothing like a bit of free advertising, especially if it seems to come from public demand. But I’m sure this has never crossed their minds, and nor would they deliberately try and manipulate people to dance to their tune.
Footnote: Those stats, if you care! PFS in placebo group = 8.0 months PFS in treatment group = 9.4 months HR = 0.83 97.5% CI = 0.72-0.95 p = .0023 MLE in placebo group = 19.9 months MLE in treatment group = 21.3 months HR = 0.89 97.5% CI = 0.76-1.03 (!) p = .077 (!)