A recently launched software package for clinical trials from Phase Forward is being used in a US government programme to see if it can pick up safety issues with medicines much earlier than current post-marketing-based surveillance systems.

Phase Forward’s Clinical Trial Signals Detection software – which won a ‘Best of Show’ award at the BioIT convention last month – builds on techniques used in the company’s range of post-marketing safety scanning software and applies them to the much smaller but deeper datasets seen in clinical trials.

And while a number of companies compete with Phase Forward in post-marketing safety signal detection software, it believes it is the first to market with a product designed for use in Phase II and III clinical testing.

The hope is that, using the software, adverse reactions with a drug – such as the elevated risk of heart attacks and stroke that led Merck & Co to withdraw its COX-2 inhibitor Vioxx (rofecoxib) from the marketplace in September 2004 – can be picked up long before it is launched onto the market.

An investigation of the safety of non-steroidal anti-inflammatory drugs, including Vioxx, that makes use of the CTSD software is being carried out by the Arizona Department of Health. This project is looking at the records of Medicare recipients to try to look for the onset of cardiac problems and stroke, and compare patients on COX-2 inhibitors and traditional NSAIDs as well as NSAID non-users.

Phase Forward’s Judy Hanover, who heads up product marketing for the company’s Lincoln Technologies subsidiary, told Pharma Times that CTSD is designed to be used in the Phase II/Phase III trials setting, when individuals are being exposed to drugs and safety and efficacy data is being generated.

Data mining

At this stage, trial sponsors are dealing with reports of adverse events, but generally do not have a lot of background information on the drug under test. But what they do have, according to Hanover, is what could be described as a ‘mini electronic medical record’, i.e. a package of information on each patient that can be mined for clues about a drug’s safety profile.

What Phase Forward sought to do with CTSD was take the dataset from trials, which could include lab test results, medical images and other clinical information, place it in a database and applies a couple of data mining analyses that allow you to build a picture of the statistical significance of observed events, even in a small dataset.

“CTSD looks at all the data from clinical trials, including adverse events, trends in lab values that don’t necessarily register as adverse events, including liver and kidney function analyses and QT interval data, for example, to help you get a safety profile for your drug a little sooner,” said Hanover.

In essence, the software serves as a kind of early warning system, picking up small changes in lab values that might be overlooked but lead to adverse reactions when the drug is used in a wider patient population after marketing.

One early benefit for clinical researchers is to suggest tests that may be incorporated into future trials to identify whether a safety signal is an issue or not, according to Hanover, and it could also be used to design studies that, for example, illustrate an advantageous safety profile compared to an established drug in the same class.

Moreover, as the US Food and Drug Administration (FDA) is using the CTSD system, the software can point out issues that may be raised by the regulators during the review process.

“It’s about knowing what you need to know about the safety of your drug, as early as you can, so that you can make decisions on it,” she said.

CTSD can be used as early as Phase II to beef up certain data signals, but comes into its own in the Phase III trial setting where you are generally dealing with a couple of thousand patients or more.

Unless an event occurs in at least 1 in 1,000 patients, you probably won’t pick it up in a clinical trial. This means that an extremely rare side effect such as rhabdoymolysis, the often-fatal muscular reaction that led to the withdrawal of Bayer’s Baycol/Lipobay (cerivastatin) in August 2001, would be impossible to detect in a clinical trial using current statistical techniques.

Using the ‘shrinker’ algorithms in CTSD, it should be possible to use lab data to identify trends that might pick up this potential from the much smaller population in a clinical trial, and gauge the significant of a lab value – for example a rise in blood levels of an enzyme – even if the patient does not actually have the adverse event itself.

The CTSD also allows data to be pooled among trials, increasing the size of the patient population that can be analysed in order to identify safety signals.

While CTSD is still a new product, it has been road-tested by looking at historical data on a number of drugs that were withdrawn from sale. “We definitely see safety signals when we apply the methodology, at least months and usually years sooner,” said Hanover.

Future directions

Most of the use of CTSD to date has been at Otsuka Pharmaceuticals, Phase Forward’s development partner for the software, and the FDA which has used the package to gauge the significance of lab values in clinical trials. But the product was only launched a little under three months ago as an additional feature of an existing product known as WebSDN, and there is an additional launch soon that will include some enhancements to the methodologies.

Specifically, the next launch adds some logistic regression techniques that allow you to separate the effects of concomitant medications.

“Tools like this that can avert safety failures are growing in importance,” according to Hanover, and big pharmaceutical companies are becoming more willing to make use of them. GlaxoSmithKline is one that has made a concerted effort to adopt this type of approach, and has recently become a CTSD customer, she noted.

The next step for this product, said Hanover, is to develop in-line safety analysis functionality for the software.

“Right now, this product takes finished trials and analyses the datasets. In the future, data will be pulled out of trials while they’re still ongoing, allowing the course of a trial to be changed,” she said.

Once this functionality is in place, the tool could potentially be used to fail drugs earlier when safety problems emerge, reducing the exposure of patients to a hazardous drug and also reducing investment in doomed projects.

Getting projects to ‘fail faster’ has been identified by the Tufts Center for the Study of Drug Development in the USA as a key objective in improving the decline in R&D productivity that has plagued the drug industry in recent years.

- A recently-published report from consultancy firm Frost & Sullivan indicates that the world market for eClinical software was around $202 million in 2004, with a compound annual growth rate of 16.4% that should lead the market to more than triple in size to $678 million by 2012. F&S notes that the USA is the largest market for these kinds of tools accounting for 55% of the total market.

One emerging trend in the sector is for eClinical trial vendors to partner with clinical research organisations and co-bid for large clinical trial accounts.