P&G's nonexistent plateau - part I

Here is the abstract I presented at the International Bone and Mineral Society Congress in Montreal last week (in between a lost passport, vomiting and cancelled trains). It involves the smallest of the three implicated Procter and Gamble publications (backstory here). Two more to follow.

Relationship of fracture risk to change in bone resorption with risedronate in the HIP study - is there a plateau response?

Authors: A. Blumsohn (1), J. L. Hutton (2)
(1) Sheffield Teaching Hospitals, Sheffield, UNITED KINGDOM,
(2) Department of Statistics, University of Warwick, Coventry, UNITED KINGDOM.

Understanding the determinants of bisphosphonate induced change in fracture risk is a prerequisite to rational prescribing and therapeutic monitoring. A previous abstract (Blumsohn, Barton, Chines, Eastell. JBMR 2003;18 S2:S89), and draft publications failed to shed light on the true relationship between change in bone resorption (uNTX/Cr) and fracture risk in the HIP study. The study included 938 women (FN T score <-3, age 74 SD 3) who received Ca and either 5mg risedronate/day, 2.5mg or placebo for 3 years. Randomization and event codes were supplied to authors in 2006. Data did not provide evidence to support previous conclusions.

Previous reports on these data suggested risk of incident vertebral fracture (V#) was non-decreasing when NTX decreases beyond -30% (%?NTX < -30%), and the relationship was "non linear" with "little further improvement in fracture benefit below a decrease of 30 to 35%". It further suggested that another marker (%?PINP) was significantly predictive of V#.

We used several statistical models as well as visual inspection to evaluate a potential "plateau" effect at a putative threshold -30% or -40%. Cox and logistic regression models were used, with thresholds of -30% and -40%, and two transformations of NTX: %?NTX and ?log(NTX). The response was allowed to take different values above and below the threshold, for both linear and quadratic functions. Conclusions were essentially the same for all models, with or without inclusion of data on the unlicensed (2.5mg) dose.

Visual inspection showed no evidence of a plateau near the putative threshold. With 5mg risedronate most (9/11) incident V# occurred with change in NTX beyond the proposed -30% threshold (median %?NTX with V# was -49%). No patients with %?NTX < -61% sustained V# (-61% was also the approximate lower limit of data plots presented to authors by the sponsor). However 44% of patients on 5mg had %?NTX < -61%. Regression models showed no evidence for a plateau at either threshold, and significant evidence of no plateau (Cox P < p="0.010," p="0.013,">0.22).

In conclusion, this study provides no evidence to support a plateau relationship between NTX change and fracture risk with threshold near -30% in patients taking risedronate.

Author Disclosure: Study funded by Procter & Gamble Pharmaceuticals.
Reference: A. Blumsohn,A., Hutton,J.L. (2007) Relationship of fracture risk to change in bone resorption with risedronate in the HIP study - is there a plateau response? Bone 40(6 S2) S303-4.

Here are the poster and poster handout in pdf format. Both explain the relevant background to this abstract. This was also the presentation subjected to the Dr Purple treatment.

Much more to follow (including some serious science).

"The time has come," the Walrus said,
"To talk of many things:
Of shoes-and ships-and sealing wax-
Of cabbages-and kings-
And why the sea is boiling hot-
And whether pigs have wings."

Earlier|Later|Main Page


Comments on: P&G's nonexistent plateau - part I


Anonymous Anonymous said ... (December 10, 2009) : 

You need think about it. Despite the emails, the overwhelming evidence showing global warming is happening hasn't changed.
"The e-mails do nothing to undermine the very strong scientific consensus . . . that tells us the Earth is warming, that warming is largely a result of human activity," Jane Lubchenco, who heads the National Oceanic and Atmospheric Administration, told a House committee. She said that the e-mails don't cover data from NOAA and NASA, whose independent climate records show dramatic warming.


post a comment