MONTREAL—In the study of pain, the gold standard for assessing discomfort and suffering in human clinical trials is simply to ask participants how much pain they are feeling. Most experts agree that this metric is flawed, but they also acknowledge the lack of suitable biomarkers or other objective alternatives to replace self-reported measurements. Given these limitations, researchers are investigating better ways to make sense of people's pain ratings to improve trial design.

In an attempt to do so, Robert Palmer from Forest Laboratories in Jersey City, New Jersey and his colleagues reevaluated data from two phase 3, randomized, double-blind trials testing the drug milnacipran versus placebo. Milnacipran, a selective serotonin and norepinephrine reuptake inhibitor marketed by Forest Labs and Cypress Bioscience under the brand name Savella, was approved in the US last year for treating fibromyalgia, a chronic condition marked by extreme muscle and connective tissue pain.

They found that among the 2,000-plus participants only around 20% of those who self-reported smaller pain fluctuations before the study treatment responded to placebo, compared to around 35% of those with the largest swings in baseline pain. The difference among those who improved with milnacipran compared to placebo was also somewhat more pronounced in individuals with steady pain, suggesting that focusing on only subjects with stable baseline pain could increase the power of the clinical trial to find significant effects. “Your ability to discriminate between the placebo and the drug is stronger in the low-variability group,” says Palmer, who presented the findings at the World Congress of Pain here in August.

Even though focusing on only people with steady aches and pains could save companies a lot of time and money, scientists are not quite ready to exclude participants with variable discomfort from pain studies. “Our findings are provocative, but they're still preliminary,” says Richard Harris of the University of Michigan–Ann Arbor, who has also studied fluctuating pain and milnacipran on a smaller scale (Arthritis Rheum. 52, 3670–3674, 2005). Researchers now need to verify whether the effect is true in other data sets, with other metrics of pain and in other diseases that also rely on measures of self-reported pain severity, he adds. “Baseline pain variability could be a factor that increases the imprecision of clinical trial assessment,” Harris says.

Robert Dworkin, who studies the methodology of pain trials at the University of Rochester in New York, says the recent reports are part of a growing body of data supporting the need for evidence-based trial designs in the pain field. He notes that negative trials and longer trials have also been shown to lead to greater placebo responses. Considering that the vast majority of candidate painkillers have failed in late-stage clinical trials, Dworkin says that parsing people by their baseline pain levels could be one solution to improving the field's translational success.

“Maybe we're not doing the very best we can to design trials that test these analgesics in the best way possible,” says Dworkin, who last month received a $1 million contract through the US Food and Drug Administration's new Analgesic Clinical Trial Innovations, Opportunities, and Networks initiative to study the design, implementation and interpretation of pain trials. “If we can have adequately powered trials with fewer subjects that would be a huge plus.”