Selective publication of clinical data on Roche’s breast cancer treatment Herceptin (trastuzumab) may have significantly overstated the drug’s effectiveness when used sequentially as an adjunct to chemotherapy, claim experts from, and associated with, New Zealand’s Pharmaceutical Management Agency (PHARMAC).

PHARMAC is the agency that decides which drugs should be funded by the government in New Zealand. Not insignificantly, the High Court of New Zealand last month asked PHARMAC to revisit its decision in July 2006 to turn down an application from Roche for funding of 12 months’ sequential treatment with Herceptin.

PHARMAC subsequently approved funding for nine weeks’ concurrent treatment with Herceptin in patients with HER2-positive early breast cancer in April 2007. The agency’s rejection of the 12-month sequential regimen was challenged by a group of Herceptin patients, resulting in a judicial review of the July 2006 decision.

Although the suggestion that there may have been publication bias in the clinical evidence for Herceptin is not in itself new – PHARMAC, for example, raised the issue last June in a justification for its funding of nine-week Herceptin in the Journal of the New Zealand Medical Association – the issue has been stoked up by a comment piece in the latest edition of The Lancet.

PHARMAC’s Scott Metcalfe and colleagues warn that selective publication of a key clinical trial with Herceptin means “patients are being given an important treatment sequence that may be much less effective than currently thought”. Sequential treatment with chemotherapy and Herceptin “is licensed, is standard practice, and is the publicly funded regimen in many countries, such as most of Europe (UK included)”, they note.

Head to head

Out of six relevant trials with Herceptin, just one randomised study by the North Central Cancer Treatment Group, NCCTG-N9831, has looked at sequential and concurrent treatment head-to-head, together with a control or usual care group, write Metcalfe et al. Despite its important implications for the optimal use of Herceptin, though, this trial has only been partly published, they add.

Specifically, data from 985 women given 12 months of trastuzumab following completion of treatment with paclitaxel “are in effect missing, despite publication of data from the 12-month concurrent and control groups of the same trial nearly three years ago”.

Interim results from all three arms of NCCTG-N9831 were, however, presented orally in 2005 at the annual meeting of the American Society of Clinical Oncology (ASCO), the authors observe. These data showed that, after 1.5 years of median follow-up, sequential trastuzumab produced a comparatively small 13% reduction in disease events compared with usual care – and with “a reasonable chance of being no better than the control group (hazard ratio (HR) 0.87, 95% confidence interval (CI) 0.67-1.13)”.

Concurrent trastuzumab, on the other hand, was significantly more effective than sequential therapy in the NCCTG trial, reducing disease events by one third (HR 0.64, 95% CI 0.46-0.91).

Combining the NCCTG sequential data from the ASCO conference slides with updated results from other trials of sequential trastuzumab (HERA and PACS-04) “shows a treatment effect one third less than initially estimated”, Metcalfe et al say. More events will have accrued since 2005, which will further affect pooled estimates, they point out.

“We understand that, although the updated combined analysis for the concurrent group in the NCCTG study and NSABP-B31 [another study of concurrent treatment with Herceptin] will soon be published, the NCCTG study’s sequential group data will be presented and published only when the number of events is believed sufficient to ensure analysis has the appropriate statistical power,” the authors comment.

Instead, they argue, the same criteria for analysing efficacy and adverse effects should apply to both the sequential and concurrent arms of the NCCTG study. “For the NCCTG and NSABP studies, the proper approach would be to publish all efficacy data from both trials separately, as has been done for cardiotoxicity. Instead, what has been published and presented is post-hoc pooled analyses limited to concurrent treatment groups of trials that were of appreciably different design.”

The NCCTG three-group study had enough power to detect a statistically significant difference between the concurrent and sequential treatments, Metcalfe et al point out, and “even when trials are underpowered, their data should still be published to inform meta-analysis”.

Interestingly, they add, the positive results from the NCCTG study’s concurrent group “were released despite reaching less than half of the pre-specified event count required for first interim analysis … Furthermore, toxicity data from all three NCCTG groups have been considered mature enough to publish: the sequential group included.”

Duty of care

There is “a duty of care to trial participants, sponsors, regulators, and the public to promptly publish outcomes in all exposure groups”, the authors contend. “This topic is notably underplayed in current statements on good clinical practice for reporting, and also in official directives on the conduct of trials. Is it now time the efficacy data from the entire NCCTG study, with updated events accrued since 2005, see the light of day?”

Roche’s stance is that it is up to the study’s investigators to publish the relevant data. The NCCTG is a national clinical research group sponsored by the US National Cancer Institute, although Roche’s affiliate Genentech did provide some support for the study, along with the National Institutes of Health and the Breast Cancer Research Foundation.

UK reporting on the Lancet article in the Guardian and Telegraph newspapers neglected to mention the judicial review of PHARMAC’s decision to deny funding for 12 months’ sequential treatment with Herceptin in New Zealand. In keeping with the High Court ruling, that decision has now gone out for consultation until 9 June 2008.