Tuesday, November 11, 2008

Lies, Damn Lies and Statins

(By Guest Blogger, Sceptical Rogue)
Wow.

What a great morning to wake up to!

The Daily Mail is telling me all about the new statin drug that slashes the risk of heart attacks and strokes for EVERYONE by up to 44 per cent.

I can't believe it.

How happy I am.

Even BBC Breakfast is having a go, promising that 'rarely have we seen such clear and dramatic results'.

But before we all rush to the doctor demanding prescriptions of Crestor to crush up and put in our morning cereal, I thought I'd take a slightly closer look.

So, this story is about the The JUPITER Study which was published online in the New England Journal of Medicine on 9th Nov 2008. The study looks at patients with relatively low levels of total (HDL+LDL) and 'bad' (LDL-C) cholesterol but with elevated C-reactive protein (a marker of inflammation), and looks at the effect on several cardiovascular events (notably Heart Attack and Stroke) when treated with rosuvastatin 20mg or placebo.

Now let's start with the good stuff - this trial does tick the Thinking Is Dangerous box of acceptable trial design. The trial is Double-blind, Randomised, Placebo controlled, Multi-Centre and Peer review published. Disclosure of the randomisation is pretty good and thorough with very low differences in risk factors across the groups.

The headline grabbing stuff tells me that if I take rosuvastatin I can slash my risk of heart attack by 58% and stroke by 48% even if I have normal levels of cholesterol. Well what could be wrong with that? Surely, a miracle has occurred here today?

Well, no. Not really.

There were 31 Myocardial Infarctions (Heart Attacks) in the treatment group compared with 68 in the placebo group. Each group had 8901 patients in it. There were 33 strokes in the treatment group and 64 in the placebo group.

In real, proper, everyday terms, this means that if I have normal cholesterol levels, but higher c-reactive protein levels, over 2 years:-

- My risk of heart attack goes from a rather disappointingly low 0.76% to an indistinguishably lower 0.35%.

- My risk of stroke goes from a most decidedly non-headline-grabbing 0.72% to a yippee-I'm-gonna-live-forever 0.37%.

Putting this into context, it was the Scandinavian Simvastatin Survival Study (4S) in the 1990's which really propelled the statins into the biggest selling drug class of all time. This study, also Randomised, Multi-centre, double-blind, placebo controlled and peer review published, enrolled patients with high cholesterol levels, compared Simvastatin to placebo and looked again at cardiovascular outcomes.

In this study, the risk of having one or more major coronary event went from 30% in the placebo arm to 19% in the treatment arm! Now these numbers are impressive.

It is worth noting that the Relative Risk Reduction in both trials is roughly similar, but that the real world Absolute Risk Reduction of JUPITER pales in comparison to 4S. And yet, have a look at the media outlets with their impressive numbers and listen as they demand that NICE start funding high doses of expensive statin in primary prevention to make, well, little difference to most people really.

In the interests of fairness it should be pointed out that these studies looked at different patients with different risk factors. It should also be pointed out that 4S had a follow up of 5.4 years whereas JUPITER was stopped early after median follow up of 1.9.

However, my point still stands - The Media, and let's not forget Astra Zeneca themselves - (bless them) will go for the big numbers and the huge claims and to hell with context - but then 'New Study Reduced Risk of Heart Attack from 0.76% to 0.35% for only some people' won't set the media world on fire.

-------------------------------
Skeptical Rogue
===============================

BPSDB

6 comments:

  1. Brilliant! I just love statistics.

    ReplyDelete
  2. Harriet Hall is discussing this too over at SBM. Her verdict:

    "So rosuvastatin is effective and safe for primary prevention of heart disease in patients with normal cholesterol levels. Does that mean we should give it to everybody or put it in the drinking water? NO!..."

    There's a lively discussion in the comments, including someone from THINCS.

    ReplyDelete
  3. Thanks Claire,

    I've left a comment at SBM - I was listening the Skeptics Guide to the Universe podacst when your comment came through!

    T

    ReplyDelete
  4. Point well taken indeed: how come the authors do not mention ABSOLUTE risk reduction ?
    They are less than 1% !
    You do notice death reduction is significant (18000 patients makes it easier) but is it relevant (0.5%)?

    The early stoppage has no justification even pre-planned, since the placebo group patients were not harmed by treatment and it is hardly justified to decide they should all be statin treated based on just one study. It is a short time to detect potential more relevant benefits (beyond 1% in absolute risk) or late side effets (mental status deterioration and cancer require a little more time to appear).

    One first point makes me feel uncomfortable:
    The way the authors naively try and conceal the fact there is no difference in FATAL strokes or myocardial infarction in table 3 : this is ridiculous , anyone can make the maths. This is lousy data presentation spinning. Why should they do that?

    Another point:
    More diabetes cases, but the authors try and minimise this, stating
    "minimal difference in the median glycated hemoglobin value (5.9% and 5.8%, respectively;P = 0.001)." So when that does not suit you , a significant result is tagged "minimal" ; And indeed it is minimal. They say mean glucose was not difference. But this and the glycated hemoglobin level are irrelevant to the matter of diabets. Diabetes is a qualitative diagnosis, you are or you are not diabetic. The elevated glyc. haemoglobin or glucose levels from the diabetes diagnoses are very unlikely to impact on the average values for the group. I also call this a dishonest presentation.

    Hence I see in this paper where a lot of money (incentives) are at stake, some indications of a biased presentation.
    In essence: why are they trying to make it look better than it does, why this make up ?

    Now I acknowledge there are some significant results, but :

    1- When I calculate the absolute risks (that they don't indicate, why ? ) I see nothing impressive , less than 1% reductions if I'm not mistaken (absolute risk is real life, relative risk is well known and shamefully accepted in great journals as a way to enhance a result's appearance). Someone is trying to impress us. If they don't give the absolute risk then they are not satisfied with it, this is the logical conclusion.

    2- When I read there aren't more muscle complaints with the statin, I am amazed. Really ? Is this consistant with what is previously known? No. Then there may be something wrong. It is not a low dose of rosuvastatin.

    3 -When I compare this with the benefits of a healthy way of life (real Cretan diet or equivalent, not just olive oil, physical activity, not smoking) well I don't exactly remember the figures but (I don't have the figures to hand, so I hope I'm not wrong) if I remember well the order of magnitude of the benefits is much higher for no or little cost and extended benefits (bones, cancer , weight, glucose tolerance). This should have been mentionned in the discussion.

    4- My opinion on the early interim analysis is that it was unwarranted as stated above, and it feels lik a gambler who leaves the table early after a lucky hand. With more time, the results could have been more convincing with better improvements in absolute risk . With less time there is a lesser risk side effects such as cancer od cognitive decline may be detected.

    It is a pity and in my opinion, unethical, that this study was not carried out longer.
    This and the aforementionned reasons, and specifically the author's trying to make things look better than they actually are, this makes me feel someone is trying to manipulate me. This casts huge doubts on the study.
    And this adds to the fact that at least one authpor has a direct financial interest in selling the CRP test (he has a patent on ot) that discriminates those who "benefit" from treatment. Many authors have significant conflicts of interest. Look up the long list.

    ReplyDelete
  5. I see Ben Goldacre has used repeated this posting in his Guardian column this week.

    Not even some linky love :(

    *sniff*

    ReplyDelete

Share it