Treating patients rather than their functional neuroimages. Comment on Br J Anaesth 2018; 120: 299–307

Editor

Using robust neuroimaging methodology, Wanigasekera and colleagues¹ have identified an interesting suppressive effect of pregabalin on brain activity in individuals with neuropathic pain in the absence of significant changes in self-reported pain intensity. This highly respected research team has made many important contributions to the field. However, we disagree with their claim that functional magnetic resonance imaging (fMRI) to identify ‘objective evidence of pharmacodynamic effects’ is superior to self-reported analgesia in early proof-of-concept trials.

The authors' central premises are that pain report in randomised placebo-controlled trials has halted development of promising analgesics, and changes in brain activation after drug administration can identify the most promising candidates for further development. This assumes: (i) the drug is effective; (ii) slow progress in new analgesic development is attributable to problems with pain measurement; (iii) placebo and drug effects can be disentangled; and (iv) fMRI signal changes are more meaningful endpoints in early proof-of-concept trials.

• (i)In their study, changes in brain activation and self-reported pain did not appear to be associated. The authors draw a distinction between ‘behavioural’ analgesia and ‘objective evidence of potential analgesic benefits’ via neuroimaging to support their failure to show fMRI relationships to pain report. Rather than conclude the drug was ineffective, the authors suggest fMRI as a more sensitive measure. Instead, small self-report effect sizes might result from poor drug efficacy. It is also possible that the association between pain intensity and brain activation is non-linear, or that the detected changes in brain activity were associated with other aspects of pain experience (e.g. cognitive evaluation).² Had these results been obtained in the early years of validating fMRI, we would have concluded that fMRI is not a valid means for assessing pain mechanisms.
• (ii)The literature suggests that self-report measures are reliable³ and sensitive to treatment effects.⁴ Pain-related fMRI metrics have only fair-to-good reliability, and do not outperform self-report.5, 6 The authors initially attribute poor randomised placebo-controlled trial outcomes to ‘noisy’ self-report measures, but later suggest moderating factors (mood, expectation) accounted for the lack of drug effects. Variability in self-report is not necessarily error. Numerous factors predict this variability (e.g. affect, distraction).7, 8 Assuming fMRI measures truly reflect pain mechanisms, they should also reflect these factors. There are no empirical data to support self-report as a hindrance to drug development.
• (iii)Conventional wisdom suggests placebo effects confound ‘active’ drug effects.⁹ Placebo effects are integral to patient responses to all treatments.¹⁰ In clinical settings, the primary goal is to achieve maximal treatment response. Expectations are modulated by numerous factors, including provider endorsement, advertisements,¹¹ and alternative sensations/off-target effects. Separation of placebo and ‘active’ components of drug action does not reflect clinical reality, because they are inextricably intertwined. If the authors aim to control for psychoactive or sensory drug effects that potentially confound patterns of brain activity, active placebo controls with similar side-effects are required.
• (iv)Changes in functional brain activation during stimulation may not specifically index antinociceptive effects. Alternate explanations for these findings should be considered, including iatrogenic effects or differences in cognitive/emotional state.12, 13 Affective state influences self-reported pain and is reflected in functional brain measures.13, 14 If we set aside these concerns and believe changes in brain activity more sensitively index analgesia than self-report, the very problem the authors purport to solve may be exacerbated. If the drug advances to the next trial phase in the absence of self-report effects (but significant fMRI), power estimates would be derived from smaller self-report effects, resulting in enrolment of a substantially larger patient sample. Using fMRI as the primary endpoint in early trials will likely result in increased financial costs and resources to sufficiently power subsequent trials. We do not believe this approach will translate meaningfully to clinical practice when the ultimate goal is to produce significant analgesia for each patient.

We propose an alternate strategy for rapidly screening candidate analgesics based on patient self-report of pain, and feasibility and cost-effectiveness of subsequent larger trials. Small pilot/feasibility studies with an a priori threshold for analgesic efficacy should be conducted; drugs not exceeding this effect size threshold would not progress further and those that do continue to larger trials. At this early stage, neuroimaging could provide mechanistic information informing the drug's mechanism(s) of action. To advance, drugs must: (i) result in subjective symptom change meeting or exceeding a priori thresholds; (ii) meet financial/delivery feasibility criteria; and (iii) demonstrate a plausible mechanism. Critically, the maximum acceptable sample size for trials would be determined ahead of time, setting an upper limit on drug screening costs.

Neuroimaging scanning sessions and analysis are expensive. This approach allows sufficient power to understand mechanisms and grants subsequent phases additional resources to recruit larger samples over longer enrolment periods, use electronic pain diaries, or both.¹⁵ fMRI is a very informative tool for understanding pain sensation and analgesia. For drug trials, fMRI can help guide understanding of mechanisms associated with treatments. We, like many other investigators, hope that understanding pain mechanisms through neuroimaging will someday aid in the development of new treatments. However, it has not been sufficiently articulated how fMRI and other imaging technologies actually inform the development of new drugs. Regardless, we strongly believe that fMRI measures should not replace self-report as the primary endpoint in any study phase.

The authors' suggestions for fMRI as a primary endpoint in early proof-of-concept trials have problematic implications. We encourage drug development efforts to develop more potent drugs, rather than using indirect measures of pain to replace self-report. fMRI can serve as a supportive tool to elucidate mechanisms of these compounds on pain reduction. However, investigators should avoid misattributing a failure to detect analgesic drug effects as a failure of the patient to accurately and reliably report their pain. If we do so, we run the risk of creating compounds that target changes in fMRI signal instead of treating patients' reported pain experience.

Authors' contributions

Contributed to all aspects of manuscript preparation including initial conception, writing, proof-reading, and approval of the final version: all authors.

Declaration of interest

The authors declare that they have no conflicts of interest.

Funding

Center for Pain Research and Behavioral Health at the University of Florida.