When access to data is a real bone of contention

Should academics put their names to papers on drugs trials if they have not analysed or had only restricted sight of the data? Phil Baty investigates

November 25, 2005

Doubts about the conduct of research at Sheffield University's Bone Metabolism Research Unit first surfaced in 2002.

Like many medical researchers, the Sheffield group, led by Richard Eastell, received funding from a pharmaceutical company hoping to prove the worth of one of its medicines. The academics were working with Procter & Gamble Pharmaceuticals to evaluate the effectiveness of its osteoporosis drug Actonel (generic name risedronate).

In spring 2002, Professor Eastell was quizzed by an overseas academic about how he had reached his conclusions in a paper he presented at the annual meeting of the International Osteoporosis Foundation (IOF). Professor Eastell told his colleague that P&G had carried out all the analyses of the drug's trial data.

Professor Eastell explained the problem in an e-mail to P&G's statistician, Ian Barton, in May 2002. He wrote: "I was discussing our work with a US investigator (who will remain anonymous) who was really surprised when I told him that all the analyses for the IOF presentation were done by P&G employees.

"II think that to avoid criticism in the future it would be good if we could say that we had done the analyses independently," he suggested. The Sheffield academics should be able to say that they had been able to reach the same conclusions as P&G, he said.

But the company declined. Mike Manhart, director of clinical development at P&G, replied that it would "add an extra layer of external credibility" to the research if the academics were allowed to analyse the data. But it would also mean that "industry loses the opportunity to demonstrate its ability to be a true partner in scientific endeavours".

He said that P&G had invested "hundreds of millions of dollars" in the drug trial databases, so transferring them to an outside academic "is not something to be taken lightly".

A P&G spokesman told The Times Higher that it was standard industry practice to limit access to raw drug trial data (see below).

SCIENTIFIC IMPACT: THE THRESHOLD EFFECT

The Sheffield research unit's work on risedronate has changed the understanding of the treatment of osteoporosis - a bone disease that affects 3 million people in the UK - and also altered the perception of the effectiveness of the drug.

Risedronate and its main rival alendronate (proprietary name Fosamax), both reduce the risk of bone fractures by increasing bone density and reducing the "turnover" of bones - the rate at which they break down and regenerate themselves.

It is generally understood that Fosamax is more effective at doing this.

But research authored by Professor Eastell suggested that risedronate's inferior potency did not mean it was less effective at reducing the risk of fractures.

Based on studies of the first two of three large drug trials that resulted in risedronate being approved, named the VERT trials, the unit had suggested that both reduced turnover and increased density did not account for "all" of the reduced fracture risk in users of risedronate.

The work also suggested a threshold effect - a point beyond which any further reduction in bone turnover did not affect fracture risk.

The research, contested by other researchers, was reported in 2001 by Professor Eastell at the meeting of the American Society for Bone and Mineral Research and the thesis was further developed in the 2002 IOF presentation.

PUBLISHED PAPER

The research formed the basis of a journal article, published in 2003 in the Journal of Bone and Mineral Research , titled "Relationship of early changes in bone resorption to the reduction in fracture risk with risedronate".

It was stated in the journal, which warns authors that they must disclose "any limitation to the full access to all material", that "all authors had full access to the data and analyses".

Rosemary Hannon, a research associate in the unit listed along with Professor Eastell and two P&G staff as authors for the paper, told The Times Higher that she did not see all the data "but neither did I request access to all the data". Professor Eastell declined to comment on the May 2002 e-mail in which he says he did not carry out independent analysis of the VERT trial data.

NEW CONTRACT AND STUDY

In summer 2002, the unit signed a $250,000 (£146,000) contract to carry out a further study based on the third of the three risedronate trials, called the HIP study, with the aim of presenting a combined analysis of all three drug trials. The unit's senior lecturer, Aubrey Blumsohn, was to take the lead in this study.

In his May 2002 e-mail, Professor Eastell asked P&G's statistician Ian Barton to allow Dr Blumsohn independent access to the data. It said: "Could I suggest that Aubrey works with you to see how you did the analysis for the VERT trial and then, when we have the HIP data, we could have the analyses run by you and by Aubrey so that we can say that we got the same result with independent analyses?"

Mr Barton replied two weeks later. "I just feel that we're being very clear in what we are doing and don't need to ask an independent person to analyse the data just to make a few people happy," he said.

By late 2002, Dr Blumsohn had completed several thousands of measurements on blood and urine samples as part of his work. He handed P&G the results and says he asked for the company's raw data in return. He wanted access to what are known as the company's randomisation codes and event codes, which relate the measurements to actual patients, showing which patients had fractures and which were taking a placebo drug, for example. The contract he and Professor Eastell signed with P&G had specified that the "final written report" of the research would include "data, interpretation, opinions and recommendations based on the results of the project".

ACCESS TO DATA

By April 2003, Dr Blumsohn says he had still not been given access to the data. But P&G staff had carried out their own analyses and produced and submitted two research abstracts to be presented at the forthcoming American Society for Bone and Mineral Research - citing Dr Blumsohn as the lead author.

In July, Dr Blumsohn says he discovered that a third abstract had been written in his name - after receiving a note from P&G congratulating him that it had been accepted for a presentation at the annual meeting of the American College of Rheumatology (ACR) later in 2003.

The company also wanted to follow up the abstracts by publishing journal articles.

In late April, Mr Barton e-mailed Professor Eastell and Dr Blumsohn introducing "one of the external medical writers", Mary Royer, suggesting that she could help in the writing process.

But in June, Dr Blumsohn wrote to Mr Barton: "Before I go ahead with the big writing push, I think I would want to get a better feel for what the data is telling us... (it) would inspire us with confidence that the arguments are sound."

Mr Barton replied that this would be a "distraction" from getting the journal manuscripts "written and submitted before... our competitors pip us to the post".

But at the end of July Dr Blumsohn was allowed to visit P&G's offices in Egham, Surrey, to discuss the data. After this meeting, Mr Barton e-mailed further statistics to Dr Blumsohn saying he wanted to "provide you with assurance that we are not hiding anything".

But this access to the data, which had been granted to Dr Blumsohn, did not reassure him about the results.

In particular, he said that the way the scale on an important graph had been plotted meant that 40 per cent of the data set was not displayed. He wrote to Mr Barton on August 18. He said it was not "reasonable" to conclude that there was a threshold effect.

"It seems almost impossible to talk about such a threshold... indeed our data would perhaps be compatible with exactly the reverse of what we say," he wrote.

UNPUBLISHED PAPER

In late October, despite his concerns, Dr Blumsohn delivered the ACR oral presentation. "In retrospect I might have refused to present this work," Dr Blumsohn told The Times Higher . Instead, he used the oral presentation to address some of the problems he had with P&G's conclusions.

In November, a "publication brief" based on this presentation was produced by P&G as the basis for a future journal article. It included the graph amended to include the missing 40 per cent of data, but its conclusions were now more circumspect and it did not contain any reference to the threshold effect.

Dr Blumsohn was satisfied with the brief as far as it went, but said he was, and remains, unable to verify all the conclusions due to a continuing lack of data.

However, P&G said: "Dr Blumsohn had full access to the relevant data and could (and did) request many additional analyses to make sure he was fully comfortable with the approach and conclusions."

Further revisions to the publication brief were made, which Dr Blumsohn argued were unacceptable. He objected that P&G was trying to change the hypothesis of the work in retrospect, and discussions gradually broke down.

In May 2004, he made a formal complaint to Professor Eastell, who was also research dean for Sheffield's medical school at the time. The complaint was copied to Tony Weetman, the medical school dean.

"No self-respecting scientist could ever be expected to publish findings based on data to which they do not have free and full access," Dr Blumsohn wrote.

But in December 2004, in a letter to Dr Blumsohn, Professor Eastell said that, under guidelines from the US Pharmaceutical Research and Manufacturers of America (PhRMA), "there is not access to the data...".

He added: "The approach we have taken for this manuscript of working closely with the statisticians to identify the best approach to analyse the data is an example of best practice." He said that if Dr Blumsohn wanted to be a co-author of the paper he would have to sign an author's agreement stating he was in "full agreement" with the work. No subsequent paper has been published with or without Dr Blumsohn's name.

In a statement, Professor Eastell declined to comment on the issues, saying only that Dr Blumsohn should take his concerns "through the appropriate university channels".

He said that the detailed points put to him by The Times Higher were not correct or "create a misleading picture". He added: "I welcome the opportunity to address these concerns in the proper manner."

phil.baty@thes.co.uk

PROCTER & GAMBLE: 'WE ALLOW EXTERNAL RESEARCHERS SUFFICIENT ACCESS'

A spokesman for P&G said: "Dr Blumsohn was not an investigator in the clinical trials which led to the drug's approval. He participated in the design and execution of post hoc exploratory analyses that were agreed by both parties.

"The raw clinical data in question were generated by P&G at great effort and expense. They are from the pivotal clinical trials that supported the approval of risedronate for osteoporosis in many countries.

"P&G has a very legitimate proprietary interest in these clinical trial data. It is standard industry practice to limit access to the raw data by external researchers. Typically, analyses developed by or in collaboration with external researchers will be performed by company statisticians, and the results shared with the researcher. Occasionally, the researcher is given temporary and limited access to the data, to perform the analyses directly.

"Our policy is to allow external researchers sufficient access to the data and to perform those analyses necessary for them to be confident and comfortable with the conclusions they state in scientific communications.

"Dr Blumsohn had full access to the relevant data and could (and did) request many additional analyses to make sure he was fully comfortable with the approach and conclusions.

"P&G provided Dr Blumsohn with the raw clinical data relevant to the abstracts, posters and presentations published or made in his name.

"P&G use medical writers to provide drafting assistance when desired by the authors. In this case, the medical writer had a pre-agreed outline of the manuscript from which to draft, and the manuscript was thoroughly vetted and agreed with the authors prior to its submission."

BLUMSOHN: SUSPENSION DEADLOCK

In spring 2004, Aubrey Blumsohn first presented his managers at Sheffield University with formal concerns about the conduct of his research unit's study of Actonel with Procter & Gamble.

But 18 months later, the university has not yet initiated any investigation into the matter and Dr Blumsohn is suspended from duties, facing disciplinary action for discussing concerns with The Times Higher.

Sheffield said in a statement that after Dr Blumsohn's May 2004 complaint "the head of school advised Dr Blumsohn to raise his concerns through the proper channels in order for them to be investigated fully in accordance with the university's policy. No formal complaint was raised.

"Since then, there has been almost continual communication from the university to Dr Blumsohn urging him to bring forward his concerns and evidence through the proper channels, in order that they may be fully and thoroughly investigated."

It said that Dr Blumsohn had been offered a range of procedures "including the offer of an independent investigatory panel" but that it had never "received any substantive complaint".

Dr Blumsohn argues that raising his concerns through the university's internal complaints procedures is not the correct way to address the issue.

He said he had simply been asking Sheffield, in numerous letters, if it would support him, primarily with legal advice, if he were to challenge P&G directly.

The vice-chancellor had declined his request for a face-to-face meeting on the issue.

Dr Blumsohn also said that he had lost confidence in Sheffield's internal procedures after an internal investigation into a separate complaint.

You've reached your article limit.

Register to continue

Registration is free and only takes a moment. Once registered you can read a total of 3 articles each month, plus:

  • Sign up for the editor's highlights
  • Receive World University Rankings news first
  • Get job alerts, shortlist jobs and save job searches
  • Participate in reader discussions and post comments
Register

Have your say

Log in or register to post comments