Skip to content
Advertisements

Trial registration TIPPS

July 30, 2014

Heparin prophylaxis in pregnancy

Doesn’t work

The TIPPS trial, published online in The Lancet this week (click here), shows that among high risk women, heparin thrombo-prophylaxis does not reduce adverse pregnancy outcomes, and increases rates of  bleeding. The researchers did a great job, and their findings will protect many women from these toxic drugs.   In passing they also reminded us of the importance of trial registration.

Did unregistered trials get it wrong?

heparin SR lancet

Figure 3 (above) shows the results of all seven heparin thrombo-prophylaxis trials to date.  The top four, Rey, Gris 2010, Gris 2011, and Mello, were unregistered and all showed an implausibly large beneficial effect.  The bottom three, Martinelli, De Vries, and Roger, (Roger = TIPPS)  were registered in advance so the researchers could not stop early, decide not to publish, or alter their primary outcome, without someone  noticing. None showed a benefit.

The TIPPS authors  half suggest that the truth lies somewhere between the two groups, namely the “combined (random effects)” relative risk line at the bottom of the figure. I doubt it!  More likely the top four trials got their false positive results by repeatedly peeking at their data, selective choice of endpoints or selective publication. The bottom three show the truth.

TIPPS researchers also silently altered sample size

It seems churlish after all that, to question the TIPPS trial methods, but “needs must”.  In the published paper the authors claim to have “aimed for a sample size of 284”.  Recruitment ran from 28 Feb 2000 to 14 Sept 2012 and the trial was first registered with Clinicaltrials.gov (click here) on 29 May 2009. Between then and 23 April 2013 the Clinicaltrials.gov archives (click here – variable 27) recorded anticipated recruitment of 385 – that’s 101 more than they claimed in the paper! This was altered to 284 on 23 April 2013, and changed again to the actual recruitment number of 292 on 16 May 2013. None of this is mentioned in the published paper. Tut tut!

The TIPPS results were negative, so perhaps this doesn’t matter. Presumably the authors stopped early because they ran out of money, rather than because they had peeked at the data and discovered a nominally significant result.  But imagine if TIPPS had been positive.  Readers would have assumed that they had cheated.

Jim Thornton

Advertisements
3 Comments leave one →
  1. Marc Rodger permalink
    August 8, 2014 8:23 pm

    Daer Jim,

    The original TIPPS protocol in 2000 called for the 284 sample size to obtain 80% power. When we applied for Canadian Govt funding they insisted on 90% power (385 patients). When it was clear we would not be able to reach 90% power after exhaustive eforts to boost recruitment rate we reverted to the original (and traditional) 80% power.

    Space limitations in journal don’t permit much of this back and forth to be published but we certainly made this known to the journal and reviewers.

  2. August 14, 2014 3:30 pm

    I understand Marc. You’re forgiven. It’s a lovely trial BTW. I’m sad that we weren’t able to help here in Nottingham.

  3. March 17, 2017 6:35 pm

    It has recently been pointed out to me that Rey et al 2009 was registered ISRCTN78732833. However registration was applied for on 18/04/2008. The trial recruited between August 1 2000 and June 20 2007. i.e. the registration was 10 months after the trial ended. Readers cannot exclude the possibility that the claimed primary composite outcome of “severe pre-eclampsia, newborn weight ≤ 5th percentile or major abruptio placentae” was data driven.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: