There is no greater burden of responsibility for scientists than that placed on those who conduct medical research on human subjects. On the rare occasions that this duty is inappropriately discharged, the results can be devastating. Even so, once the initial outcry dies down, little tends to change.

The diverse collection of institutional review boards (IRBs) that oversee such research in the United States barely qualifies as a 'system'. Despite repeated attempts by the Institute of Medicine and others to highlight their shortfalls, the quality and effectiveness of the boards remain patchy (see page 530).

As committees struggle with heavy caseloads, their ability to monitor ongoing trials is weakened. A large research hospital can process hundreds of applications per year, and gets little help from the federal government. The Office for Human Research Protections oversees thousands of local ethics committees and billions of dollars' worth of clinical research, and operates on an annual budget of just $7 million.

If the US government wanted to strengthen the way human clinical trials are overseen, adequate funding for the Office for Human Research Protections would be an obvious place to begin. Another improvement would be more widespread accreditation of the IRBs, to help ensure proper training and support for committee members.

But further reforms are also necessary. Some see centralizing the review of multicentre studies as a way of relieving the burden on local review boards. But the debate over whether to centralize, or even regionalize, the review of studies is complex. The United States is a large, diverse and, most of all, litigious country, and local boards help universities to fend off legal action by showing that they have taken responsibility for what goes on within their walls. Nonetheless, centralization deserves to be explored further.

One approach would set up national committees, perhaps run by the National Institutes of Health, that could establish clearer guidelines on the ethical quandaries commonly faced by local review boards. Questions over payments to volunteers, for example, or on what constitutes informed consent, need not all be answered on the current ad hoc basis.

Another challenge facing US authorities is the fact that not all research on human subjects is overseen by the federal government. The IRBs are not federal, but federally funded research on human subjects must be reviewed by them, as must any trial that becomes part of a submission to the Food and Drug Administration. Some states and institutions also require all research on human subjects to be vetted by an IRB. That leaves room for privately funded research to proceed without any requirement for ethical review. It isn't known how much research on human subjects occurs without review by government officials or IRBs. But such research may expose patients to unnecessary risks.

In 1999, the review-board system came under some scrutiny when 18-year-old Jesse Gelsinger died during a gene-therapy trial at the University of Pennsylvania in Philadelphia. In 2001, a healthy woman died after taking an unapproved asthma medication during a clinical trial at Johns Hopkins University. And looking farther afield, in London last year six men became seriously ill during a clinical trial of a monoclonal-antibody therapy.

Overall, it is not a bad record. But that is as much by luck as by design, and if it is to remain the case, real improvements need to be made to the IRB system. The impetus for such change should not have to rely on the bursts of interest that tend to follow mishaps during human trials.