Deborah Cohen: Roche responds

Deborah CohenIt’s worrying  that the quest to evaluate the evidence underpinning major public health decisions feels like a game of chess—how do you try to anticipate the next move of your opponent and how might they bide their time to stay in the game.

And so the next instalment of what can only be dubbed the “Tamiflu saga” has unfolded. Roche has finally responded to questions that the BMJ put to them last year during an investigation which was conducted alongside a Cochrane review of the data on Tamiflu in healthy adults and published last week.

This ultimately concluded that Tamiflu was effective in cutting down the duration of influenza symptoms by 21 hours. But other issues were also highlighted. This included concerns over the omission of side effects in the published literature; questions over the mode of action of the drug; the effects on antibodies; and issues around a consent form that the FDA said may risk coercion in prophylaxis trials.

Roche stand by their drug—they said so on Twitter—but wanted to wait until they saw what was written in its entirety before they responded to questions. 

Back in 2009, the Cochrane Collaboration and the BMJ together with Channel 4 News, found that there were problems with the published literature on Tamiflu. Claims that it reduces secondary complications and hospitalisations made in a Roche funded review could not be substantiated—8 out of 10 studies included in the review were unpublished; trials had been ghostwritten; lead authors were unable to provide raw data; and the largest ever treatment trial only appeared as an abstract.

Roche made some data available. The Cochrane group say it’s not the right data. Roche say they’ve provided the Cochrane group with access to 3200 pages that correspond to the trials in the Roche review. The Cochrane group say that a single full study report can be over 8,000 pages and those documents are not enough. Cochrane point out that the documents Roche gave them did not include specific details—the absence of definitions and microbiology testing for secondary complications, for example.

As the Toronto Star suggests: “the public is left to wonder who to believe: the drug company or independent researchers with partial information. Neither is a particularly good option.” A stalemate might have been reached had it not been for the release of documents by the European Medicines Agency and publicly available information from the FDA and the Japanese regulatory authorities. 

This formed the basis of the latest Cochrane review—and the results are revealing. Roche maintain that they have “made full clinical study data available to health authorities around the globe, according to the various extensive regulations, for their review as part of the licensing process. It is the role of global health authorities to review detailed information on medicines when assessing benefit/risk.” So that should have been the end of the game–but it wasn’t.

Making “available” is different to providing. How many health authorities have been provided with all Roche’s trial data? EMA certainly didn’t possess it—they didn’t ask for it. Roche have confirmed that WHO didn’t have the full dataset prior to compelling governments to buy it. And if the FDA had it, they didn’t independently analyse it all—including the unpublished biggest treatment trial which was conducted in North America. Can Roche say who—other than the company themselves—has all the data in its entirety or even knows the number of trials done?

Tamiflu – Response for the BMJ Questions

1. In the BMJ of 8 Dec 2009, you said that “full study reports” would be made available to researchers. Full study reports include multiple “modules” (usually 4-5). Why will you not release the full study reports to the Cochrane group?
 
Roche provided the Cochrane group with access to 3200 pages of very detailed information to address their questions. All completed Roche sponsored clinical studies on the safety and efficacy of Tamiflu are available as peer-reviewed publications or in summary form on www.roche-trials.com

More detailed clinical trial reports are available for use by investigators on a password-protected site, enabling researchers to verify the findings of these studies and publications relating to them.

Roche has made full clinical study data available to health authorities around the globe, according to the various extensive regulations, for their review as part of the licensing process. It is the role of global health authorities to review detailed information on medicines when assessing benefit/risk.

2. Why did you add dehydrocholic acid to the placebo when it was not in the Tamiflu capsules? Regulatory documents show that dehydrocholic acid was used in the placebo rather than an inert substance. A known side effect of this is diarrhoea and nausea. Why did you decide to use this in the placebo?
 
Matching the placebo as accurately as possible to the active is a standard approach taken in drug development clinical trials to maintain study blinding. The Tamiflu active ingredient, oseltamivir, is very bitter and so in order to preserve study blinding there was a need to mimic the bitterness of the active product in the placebo. To do this a very small amount of a known bitter agent – dehydrocholic acid USP – was added to the placebo.

For example, a placebo 75mg capsule contained 6.7mg dehydrocholic acid. This amount is a small fraction of the doses used for medical purposes of 250-500mg three times daily and so can be considered very unlikely to have any pharmacological effect. Health authorities were well aware of this approach in our submissions.

Note: We also noticed a question on the use of dibasic calcium phosphate dihydrate in the placebo. This is an odourless tasteless substance commonly used as an excipient in tablets and capsules. In this case it was used as a filler to replace oseltamivir and assist in ensuring the placebo is indistinguishable from the active treatment.

3. Did you give Professor Lipsitch and Professor Hernan* from Harvard University any trials relating to Tamiflu that you didn’t give the FDA? Did you give them the trial protocols for all trials? If not, why not? Their conclusions about the drug’s effectiveness on complications are different. If not, how do you explain the different conclusions?

[*Professor Lipsitch and Professor Hernan are Harvard academics who Roche gave data to analyse.]

Within the scope of a signed agreement, the information we initially gave to Profs Lipsitch and Hernan all related to the 10 studies (WV15670, WV15671, WV15730, WV15707, M76001, WV15812, WV15872, WV15819, WV15876, & WV15978) cited in Kaiser et al2. All of these study reports were shared with the FDA and EMA.

Initially we provided electronic data sets, then access to protocols and clinical study reports. Additional information from other trials (Roche and Non Roche studies on Tamiflu) was supplied upon request at a later stage of their independent analysis.

The FDA reviewed the data on influenza complications submitted by Roche and unlike other agencies did not accept a clinical diagnosis of pneumonia, bronchitis, or sinusitis, and requested evidence of, for example, pneumonia by chest X ray demonstrating absence or presence of pneumonia. These data were not collected in our trials. Thus data on a reduction of complications are absent in the US Tamiflu label, but present in the EU and other labels.

Professors Lipsitch and Hernan were able to re-confirm the result reported by Kaiser et al and using additional data from other published trials, extend their analysis to conclude that oseltamivir does reduce the risk of lower respiratory tract complications requiring antibiotic treatment.

4. Did WHO ask for all your unpublished data? By this I am referring to modules 1 to 5 of the full study reports. And did you give them the data? If not, why not?
 
The WHO did not specifically ask for this information. We do supply information that the WHO request as appropriate to address any specific questions they have.

5. Why did you not give modules 2-5 to the Cochrane group? GSK have offered individual patient data.
 
Roche provided the Cochrane group with access to 3200 pages of very detailed information to address their questions. All completed Roche sponsored clinical studies on the safety and efficacy of Tamiflu are available as peer-reviewed publications or in summary form on www.roche-trials.com. More detailed clinical trial reports are available for use by investigators on a password-protected site, enabling researchers to verify the findings of these studies and publications relating to them.

Roche has made full clinical study data available to health authorities around the globe, according to the various extensive regulations, for their review as part of the licensing process. It is the role of global health authorities to review detailed information on medicines when assessing benefit/risk.

6. How does Tamiflu work?
 
Please see the approved prescribing information – e.g. EU Summary of Product Characteristics, Section 5.1. [3]

In addition, a comprehensive review on the mode of action of neuraminidase inhibitors has been published by Moscona [5] 

7. Can you please explain some inconsistencies between the published papers and their equivalent clinical study reports (CSRs) obtained from the European Medicines Agency through FOI?
– Nicholson (2000) did not mention any adverse events and Treanor (2000) stated that “there were no drug-related adverse events,” but in the CSR there were 10 serious adverse events, three of which were classified as being related to Tamiflu
– Hayden (1999) describes headache as having “occurred in similar study proportions of subjects in the three study groups.” However, data from the Japanese regulators show a dose response effect.
 
It is incorrect to state that adverse events were not mentioned in these publications.
Clinical study reports are much more detailed than any publication of the same study. The total picture of drug adverse events is assessed by the sponsor and Health Authorities during the application process for marketing approval.

Tamiflu prescribing information documents (eg US Package Insert, EU Summary of Product Characteristics) include information on headache and overall reflect the totality of the safety data from the clinical development programme and the regular updates supplied in the intervening years from on market experience.

For example, we would also like to clarify the discrepancies you cite between reporting of SAEs in the publication by Treanor et al, and the associated Core Study Report WV15671 we provided to you (this clarification was previously sent to Professor Del Mar in August 2010). The publication by Treanor et al indicates that there were no drug-related serious adverse events. In total there were 10 SAEs (reported in 9 patients) listed in the Core Report for study WV15671 (table 34), including reports received with both placebo and active medication.

Six of the SAEs were described during the “on treatment” period, of which 3 SAEs (reported in 2 patients) were listed as being “possibly” related to oseltamivir.

Of the two cases, one was a reported pregnancy. The reporting of pregnancy itself during a clinical trial does not imply any side effect or abnormality. It reflects the standard operating procedure for the company to collect safety information in association with the use of any of our investigational drugs during pregnancy. The report clarifies that, at the time of writing the report, no abnormalities for pregnancy and fetal outcome were reported.

The second case was a report of pseudomembranous colitis, which resolved without sequelae. After reporting abdominal pain the patient was started on ampicillin /sulfabactam. In addition the patient underwent appendectomy (histology: normal). A Clostridium difficile (pseudomembranous) colitis was diagnosed and determined by the reporting physician to be definitely related to ampicillin/sulfabactam. The antibiotic therapy is a clear alternative explanation for this event; the event is labelled as a rare event for ampicillin. In addition, the investigator reported the event to be possibly related to oseltamivir. The company medical assessment was that the Clostridium difficile was unlikely to be related to the study drug, and likely to be related to ampicillin/sulfabactam. The abdominal pain was also listed as possibly related to oseltamivir by the investigator. It is important to note that the causality information, both as listed by the physician and as assessed by the company, were provided to the Regulatory Authorities. Abdominal pain is included in the Prescribing Information as one of the most frequent adverse drug reactions for oseltamivir. Pseudomembranous colitis (US) and hemorrhagic colitis (EU) are also listed.

8. There appears to be evidence that at least some of the Tamiflu trials were not properly randomized. In the treatment trials, patients were randomized to placebo and Tamiflu arms, and then tested for influenza. But the results of the studies are based not on an analysis of the “Intention to Treat” (ITT) population (which was randomized) but on the “Intention to Treat Infected” (ITTI) population. Would this not bias the results? The reason I ask is while the placebo and treatment arms of the ITT populations are of similar size in all studies, the treatment arm of the ITTI population is significantly smaller than the placebo arm in several of the most oft-cited studies. Thus, some infected patients seem to have been selectively removed from the treatment arm. The more biased studies also turn out to be the ones demonstrating a greater effect of Tamiflu. What might account for this bias, do you think?
 
In our clinical development programme to cover the global regulatory requirements for a marketing authorization application, the population used for the primary analysis population (ITTI) was agreed to by the health authorities. In addition, if you look at the ITT population, the difference comparing active to placebo is still statistically significant.

No patients have been selectively removed and we do not agree that there is bias in the results. In all of the study protocols, the primary analysis population was defined up front as ITTI with the definition of inclusion into this population. Bias would have been introduced if this population was decided upon after the study results became available. In the study reports and publications outcomes are presented for the ITTI as well as the ITT populations, thus enabling full transparency in assessing the treatment effect.

The trial publications focused on the primary analysis population agreed with the health authorities (ITTI). In addition, the development of all antivirals is focused on the population with the proven influenza disease and this is best reflected in the ITTI population.

9. Professor Hayden was rebuked by the FDA for a consent form in a prophylaxis trial: WV15673/WV15697. This was for including the statement that participants “will receive $300.00 for participating in and completing the study. No payment will be made to you if you withdraw from the study for personal reasons…” They said this was improper conduct. Could you confirm if this was a standard consent form or one that was adapted by Professor Hayden? How many trials employed this consent form?
 
The informed consent form was prepared by Professor Hayden’s group and reviewed by the IRB and the sponsor. The standard at Professor’s Hayden institution consistently has been to fully compensate participants who withdraw or are withdrawn for medical reasons. The questions from the FDA were fully resolved to their satisfaction.

The informed consent form for the study had a basic design from the sponsor but that was adapted at the site to comply with local IRB or ethics committee needs.

10. Also in the same studies, Japanese regulatory documents show that there was a total of 584 (314 in the Tamiflu group and 270 in the placebo group) nervous system related adverse events and 37 (24 and 13 respectively) psychiatric adverse events during the on-treatment phase. However—other than headache— this was not included in the published papers. Could you explain why not?
 
The full study reports were made available to health authorities globally in the assessment of the marketing applications. The various Tamiflu prescribing information documents (eg US Package Insert, EU Summary of Product Characteristics) reflect the totality of the safety data from the clinical development and the regular updates supplied in the intervening years from on market experience. This includes information on psychiatric and nervous system disorders.

A clinical study report is very detailed with hundreds of pages and a publication is much shorter, so precludes the same detailed discussion.

ADDITIONAL COMMENTS

Ghost writing comment
Roche maintains the highest ethical standard in the writing of clinical papers for publication, the interaction with external clinical investigators and authors, and has always clearly disclosed its own financial and expert involvement in studies and publications. During the period of time in question (1999 – 2002) it was common practice for scientific medical writers to provide writing support for publications with the authors having full access to data and full and final review of the publications. Since the introduction in 2003 of the Good Publication Practice guidelines for Pharmaceutical Companies (GPP), Roche has complied with the practice to acknowledge the involvement of professional medical writers.

Humoral antibody response to infection
The Cochrane review asserts that the product information on Tamiflu is incorrect where it states that treatment with Tamiflu did not impair normal humoral antibody response to infection. Development of specific antibodies can signal evidence of viral infection and allow an individual to develop protection against future infection. In ten pooled Tamiflu adult treatment studies a similar post treatment increase in HA antibody titer (a signal that an immune response has taken place) occurred in both the placebo and Tamiflu treatment groups, when baseline and antibody levels at day 21 were compared. A comparable antibody response has also been noted in an infection based animal model (Boltz et al JID 197:1315-23, 2008). Therefore, the statement in the Tamiflu product information is correct.

Central action Cochrane hypothesis
It has been demonstrated that oseltamivir and its active metabolite [6-7] do not enter the central nervous system in any significant quantity, and of the very small quantities that do, neither oseltamivir nor its active metabolite bind to any appreciable extent to known receptors in the CNS. Thus the hypothesis of a central action is not supported.

Tamiflu Indications in the US and EU
In the Cochrane review there were comments on the actual indications approved for Tamiflu in the US. For clarity, the approved Tamiflu indications from the relevant prescribing information in the US and EU covers treatment and prophylaxis, as follows:

US Prescribing information [4]
INDICATIONS AND USAGE
TAMIFLU is an influenza neuraminidase inhibitor indicated for:
Treatment of influenza in patients 1 year and older who have been symptomatic for no more than 2 days.
Prophylaxis of influenza in patients 1 year and older.
Important Limitations of Use:
Efficacy not established in patients who begin therapy after 48 hours of symptoms.
Not a substitute for annual influenza vaccination.
No evidence of efficacy for illness from agents other than influenza viruses types A and B.
Consider available information on influenza drug susceptibility patterns and treatment effects when deciding whether to use.
EU Prescribing Information [3]
Treatment of influenza
In patients one year of age and older who present with symptoms typical of influenza, when influenza  virus is circulating in the community. Efficacy has been demonstrated when treatment is initiated within two days of first onset of symptoms. This indication is based on clinical studies of naturally occurring influenza in which the predominant infection was influenza A (see section 5.1).
Tamiflu is indicated for the treatment of infants below 12 months of age during a pandemic influenza outbreak (see section 5.2).
Prevention of influenza
– Post-exposure prevention in individuals one year of age or older following contact with a clinically diagnosed influenza case when influenza virus is circulating in the community.
– The appropriate use of Tamiflu for prevention of influenza should be determined on a case by case basis by the circumstances and the population requiring protection. In exceptional situations (e.g. in case of a mismatch between the circulating and vaccine virus strains, and a pandemic situation) seasonal prevention could be considered in individuals one year of age or older.
– Tamiflu is indicated for post-exposure prevention of influenza in infants below 12 months of age during a pandemic influenza outbreak (see section 5.2).

Study ML76001 – Adult treatment trial
This study was not in the original licensing applications as it was not completed before the cut-off date for the database for the submission and was still recruiting. It was subsequently submitted to Regulatory authorities in accordance with the requirements. In addition, the study report is on the password protected site for investigators. Concerning the other studies cited as ongoing but not in the original NDA, these were in particular patient populations and agreed with the FDA that the final reports would be post approval commitments. Once fully completed they were submitted to the FDA later.

Health authority license submissions
Roche meets the requirements of the respective health authorities for the content and level of detail required for new drug applications /marketing authorization applications. These are well defined in global guidelines and any further detail is available on the request of the health authority.

References
1. http://www.medicinescomplete.com/mc/martindale/current/3716-v.htm
2. Kaiser L et al. Archives of Internal Medicine 2003;163:1667−1672
3. http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_-_Product_Information/human/000402/WC500033106.pdf
4. http://www.accessdata.fda.gov/scripts/cder/drugsatfda/index.cfm?fuseaction=Search.DrugDetails
5. Moscona A, N Engl J Med 2005;353:1363-73.
6. Hoffmann G, Funk C, Fowler S, et al. Nonclinical pharmacokinetics of oseltamivir and oseltamivir carboxylate in the central nervous system. Antimicrob Agents Chemother 2009 Nov;53(11):4753-61. 9/9
7. Jhee SS, Yen M, Ereshefsky L, et al. Low penetration of oseltamivir and its carboxylate into cerebrospinal fluid in healthy Japanese and Caucasian volunteers. Antimicrob Agents Chemother 2008 Oct;52(10):3687-93.

Deborah Cohen is the BMJ investigations editor.