Follow Us On
Contact Us
info@ampainsoc.org847.375.4715
American Pain Society
4700 W Lake Ave
Glenview, IL 60025
James Witter, MD PhD, Lee S. Simon, MD, and Raymond Dionne, DDS PhD
People differ in many ways, including their response to pain. While pain is a common experience, it also is an individualized experience.
Chronic pain patients have unmet medical needs. This article explores the rationale and advantages of an approach that focuses on the individual, rather than a group or mean, level of pain to explore its cause and treatment. We will call into question the way inferences have traditionally been made in analgesia with the “means” approach, and how difficult it can be to apply these results to general populations. In fact, a means approach may miss critical elements of treatment response that can ideally be captured in a responder approach. Properly constructed and validated responder indices hold the potential to better capture clinically important outcomes in clinical trials than traditional approaches that evaluate mean results of groups of patients.
Approaching aspects of pain from a patient perspective may be a valid, timely approach. Sample size estimates for trials comparing means are calculated from standard formulae requiring definition of Type I and Type II error rates, standard deviation (SD), and the delta to be observed in the means of treatment groups of interest. For clinical outcomes that engender a single endpoint such as a blood pressure measurement, this approach has many advantages. For clinical outcomes that require consideration of more than one endpoint, most of which may be less objective than a measure of blood pressure, a composite can better encompass and accommodate these important variables. Composite endpoints can then be grouped into an index capable of defining, at the individual responder level, whether a particular patient did or did not achieve the endpoints in this composite index during the study. New clinical trials can subsequently be powered based on calculations from earlier trials that take into account the proportion of patients who “responded” to the outcome domain measures included in the responder index.
This approach offers the advantage of grouping clinically important outcomes into a metric that defines individual response as the primary outcome measure, not individual contribution to a group outcome. Consequently, decisions may be made on the basis of individual response as part of the group, rather than inferring the person’s response to the variable of interest as part of a large group mean or average response. In the latter situation, there is no way to understand whether two or more clinical outcomes of interest occurred in the same person. In the responder approach, however, there is no doubt because this has been prospectively designed into the data interpretation. Therefore, for a clinical outcome in which individual response is key to proper result interpretation, the responder approach offers the flexibility, both clinically and statistically, to best capture a patient’s experience with a drug. Results similar to these can only be addressed with a means approach after substantial statistical efforts such as multivariate analysis (with their problems of imputation artifacts and bias) have been employed.
An example of the benefits of individual response, in this case a subset of the original population, is associated with the drug trastuzumab (Herceptin), which is used to treat breast cancer. The response rate in the general patient population is a very modest 6%, but it is nearly 100% effective in a subgroup of 15% to 20% of patients who express the HER-2 receptor. Given the otherwise poor prognosis for women with metastatic breast cancer, the ability to target a highly effective treatment to those with a biomarker for responsiveness makes the mean results of the group essentially meaningless compared to the individual responses of those in the subgroup. This phenomenon, in which mean responses may mask important findings or differences, should be considered in the development and assessment of analgesic drugs for which genetic composition, past experience, gender, ethnicity, expectations, disease processes, and temporal pattern of gene expression vary widely, yet are key components that determine individual response.
It is unlikely a magical silver analgesic bullet would be uniformly effective under pain’s heterogeneous conditions. Assessment of individual response relative to a generally accepted standard for pain relief, compared to the responses of individuals in appropriate control groups, may reveal subgroups for whom a treatment is particularly effective or results in unusual sparing of, or suffering from, analgesic-related toxicity. Outcomes might include identification of drugs with therapeutic value for chronic pain that is resistant to typical therapies, identification of drug classes with promise for pain in particular conditions, and identification of molecular-genetic markers for predicting analgesic drug actions in individual patients.
The biological basis that contributes to individual response to pain and analgesia at various points is perhaps illustrated by the recent demonstration that the melanaocortin-1 receptor gene mediates analgesia to an investigational kappa- opioid receptor agonist (U50,488) in female mice, but not males (Mogil et al., 2003). In humans, women with two variant melanocortin-1 receptor alleles display significantly greater analgesia from the kappa-opioid pentazocine than groups without these alleles, suggesting the melanocortin-1 receptor gene results in a distinct neurochemical genotype accounting for differing analgesic responses to a kappa-receptor specific drug.
Interindividual differences in analgesic response to morphine also are well-recognized; a recent report (Aubrun et al., 2003) demonstrated a nonlinear relationship between morphine dose and the initial visual analog scale (VAS) score. The range in the mean dose required to relieve pain (13.2 mg) in 3,045 patients was 1–48 mg, representing a range of 0.02–0.83 mg/kg (B. Riou, personal communication, September 1, 2003). The approximate 40-fold range in response to a single agent given in a controlled fashion to permit quantification of individual differences in opioid requirements demonstrates the wide interindividual variation in analgesic responsiveness inherent in any patient sample. While it is assumed that randomization minimizes differences across subjects, administration of a fixed dose to patients on the sensitive end of the population would be relatively higher than the dose needed to produce adequate analgesia and would likely result in CNS depression. Conversely, administration of the mean dose for this large sample, 13.2 mg, to morphine-insensitive patients would result in inadequate analgesia and the need for continued titration to achieve adequate pain relief. Evaluation of an investigational analgesic with potency similar to morphine at a fixed dose will likely produce similar results: inadequate analgesia in some patients, clinically useful analgesia in others, and excessive adverse effects in others. The assessment of the overall efficacy and safety of an investigational drug using a mean effect in a sample responding similarly to subjects in the report by Aubrun et al. might fail to detect efficacy or predict an undesirable side effect profile, thereby masking the response of individual patients experiencing adequate analgesia with an acceptable side effect profile.
Age differences in postoperative pain and analgesic responses have been documented (Gagliese and Katz, 2003), but appear measurement-dependent. Using the McGill Pain Questionnaire, older men had significantly lower pain scores following radical prostatectomy than younger men, but did not demonstrate any differences on the VAS scale. Both scales similarly detected change in pain from Day 1 to Day 2 following surgery, indicating both scales are sensitive, but the level of pain appeared to be better captured by verbal descriptor scales of pain qualities than nonverbal measures of intensity. Age differences that might explain this difference include varying cognitive abilities or linguistic skills across age groups or age-related changes in the neurobiology of pain mechanisms. Use of mean responses across a group of subjects might not detect a meaningful change in pain intensity in response to an analgesic drug if subjects vary sufficiently in age to make the standard VAS less sensitive to response in elderly subjects than in younger subjects.
While the development of plasticity in the nervous system in response to pain is becoming generally accepted, identifying its unique manifestation in humans is still problematic. Examination of patients with fibromyalgia, which may be considered a state of central hyperexcitability of the nociceptive system, reveals lower pain thresholds and greatly reduced tolerance to cold pain (Desmeules et al., 2003). Averaging the responses of subjects in a clinical trial, some of whom may have altered pain responses due to central sensitization, might fail to discriminate patients who benefit from centrally acting analgesics, such as drugs acting at NMDA receptors. Use of individual responses might identify a subgroup benefiting from an intervention that is active for a pain state such as central sensitization, but not particularly useful for other types of pain such as acute inflammatory pain.
Recent brain imaging studies provide evidence that subjective ratings of pain magnitude are closely related to objectively measured neural activity in a number of cortical regions involved in pain processing (Coghill, 2003). Interindividual differences in subjective reports of pain magnitude were closely related to the degree of activation in these regions, suggesting interindividual differences in subjective pain ratings reflect actual differences in the pain experience (Coghill and Eisenach, 2003). In another study, preoperative ratings of experimental pain were significantly correlated with postoperative pain ratings following elective cesarian section, providing evidence that individual pain sensitivity can be related to postoperative pain responses following surgical injury (Granot et al., 2003). While this study lacked controls to assess interindividual differences in the use of pain scales such as the VAS, these results suggest the possibility that simple preoperative tests can predict individual differences in the pain experience after surgery (Coghill and Eisenach, 2003).
Evidence that neural circuitry underlying expectation influences pain perception also is emerging from functional brain imaging studies; this variability across individuals may be related to differences in the degree of uncertainty associated with expectation (Ploghaus et al., 2003). Experimental manipulation of subjects presented with a painfully hot as well as innocuous warm stimulation and color cues revealed expectation of the painful heat-activated sites that could be distinguished from those activated by the painful heat (Ploghaus et al., 1999). Uncertainty about the impending nature of an unexpected event produces anxiety, resulting in anxiety-induced hyperalgesia associated with activation of the hippocampus, anterior cingulate cortex, and insula. The influence of cognitive factors such as expectation clearly varies across subjects based on past experience and current emotional state, confounding measurements across subjects rather than assessment of individual responses.
The principle behind pharmcogenomics is the belief that the responses of a patient to a drug regimen (the phenotype) can be related to some facet of that patient’s genetic composition (the genotype). If demonstrated in a sufficient number of individuals, it might be possible to derive a causal relationship between genetics and therapeutic response. The example of Herceptin, described earlier, demonstrates the potential for pharmacogenomic success, but, to date, few other success stories have emerged (Willis and Lesney, 2003). Another promise of pharmacogenomics is the ability to identify potential adverse reactions that occur in a subset of individuals who react poorly to a treatment; that is, sometimes related to the specific individualized metabolism of the parent drug. Given that most drugs are withdrawn from the market for patient safety, it may be more cost effective to identify individuals likely to experience toxicity due to a drug class or a particular drug or dose.
The relationship between a single nucleotide polymorphism (SNP) and pain is illustrated by neuropathic pain in individuals with Anderson-Fabry disease. This rare disorder is caused by an inherited deficiency of the enzyme alpha-galactosidase A, resulting in an accumulation of globotriosyceramide in many tissues including neuronal cells. The disease presents in childhood and progresses to a peripheral neuropathy, along with other manifestations of the disease. These patients are largely nonresponsive to conventional analgesics, sodium channel blockers, and other agents used to relieve neuropathic pain (MacDermott, 2001). Recent reports suggest enzyme replacement therapy with galactosidase A improves outcome and provides a basis for gene therapy to correct the underlying genetic defect. Susceptibility loci for the development of complex regional pain syndrome (CRPS) have also been identified by comparison of DNA from a group of CRPS patients to a group of control samples from healthy blood donors (Van de Beek, et al., 2003). A subset of these SNPs were elevated in 12 of the subjects who developed CRPS following trauma, suggesting an interaction between trauma and genetic factors conferring susceptibility (Van de Beek, et al., 2003).
The possible role of genetic factors in pain sensitivity was assessed in sibling pairs using experimental pain stimuli in comparison to polymorphisms for the mu and delta opioid receptor subtypes (Kim et al, 2003a). Moderate, but significant, intraclass correlation coefficients were demonstrated from sibling pairs for pain sensitivity similar to the degree of correlation seen for inherited traits such as body mass and longevity. Gender differences in experimental pain perception also were evident, suggesting the need to analyze males and females separately for genetic influences on pain sensitivity. A functional polymorphism in a delta opioid receptor gene resulted in differences in pain sensitivity that were dependent on both gender and the characteristics of the applied stimuli. These data, although preliminary and in need of replication in larger samples, demonstrate pain is familial, and suggest individual response to pain is heritable and influenced by polymorphisms in genes for molecules involved in pain perception. Other studies suggest a role for COMT polymorphisms on pain perception in humans (Zubieta et al, 2003), but this effect may vary by stimulus (Kim et al., 2003b).
If one’s genetic profile influences their molecular pathways to pain and response to an analgesic in terms of efficacy or safety, these important factors revolve around the unique individual. Combinations of individual outcomes may be grouped together into a responder index that may be both scientifically and clinically meaningful. For example, an index could combine knowledge of a person’s genetic composition (their genotype) and actual response into a single outcome (the pain phenotype) that could theoretically function as the “bridge” that allows for translation of basic scientific and objective information into clinical results that adequately describe a patient’s individual clinical experience with the analgesic.
Properly constructed individual outcomes hold the potential to better capture basic and clinical outcomes in clinical trials—more so than approaching the same problem from a group or means perspective. Ongoing studies to fully assess and validate the role of the individual responder approach in analgesia will determine if this approach will facilitate the construct of a human response grid to pain and analgesics that would ultimately improve clinical care in pain, especially chronic pain.
Aubrun, F., Langeron, O., Quesnel, C., Coriat, P., Riou, B. (2003). Relationship between measurement of pain using visual analog score and morphine requirements during postoperative intravenous morphine titration. Anesthesiology, 98, 1415–1421.
Coghill, R.C., Eisenach, J. (2003). Individual differences in pain sensitivity: Implications for treatment decisions. Anesthesiology, 98, 1312–1314.
Coghill, R.C., MeHaffie, J.G., Yen, Y.F. (2003). Neural correlates of interindividual differences in the subjective experience of pain. Proceedings of the National Academy of Sciences, 100, 8538–8542.
Desmeules, J.A., Cedraschi, C., Rapiti, E., Baumgartner, E., Finckh, A., Cohen, P., Dayer, P. et al. (2003). Neurophysiologic evidence for central sensitization in patients with fibromyalgia. Arthritis and Rheumatism, 48, 1420–1429.
Gagliese, L., Katz, J. (2003). Age differences in postoperative pain are scale dependent: A comparison of measures of pain intensity and quality in younger and older surgical patients. Pain, 103, 11–20.
Granot, M., Lowenstein, L., Yarnitsky, D., Tamir, A., Zimmer, E.Z. (2003). Postcesarean section pain prediction by preoperative experimental pain assessment. Anesthesiology, 98, 1422–1426.
Kim, H.S., Neubert, J.K., Iadarola, M.J., Miguel, A.S., Xu, K., Goldman, D., Dionne, R.A. (2003). Genetic influence on pain sensitivity: Evidence of heritability associated single nucleotide polymorphisms in opioid receptor genes. Proceedings of the World Congress on Pain. (pp. 513–520). Seattle, WA: IASP Press.
Kim, H., Rowan, J., Brahim, J., Goldman, D., Herscovitch, P., Dionne, R.A. (2003b). The association between catechol-O-methyltransferase gene (COMT) variation and acute post-surgical pain in humans [Abstract]. Annual Meeting of Society for Neuroscience, New Orleans.
MacDermot, J., MacDermot, K.D. (2001). Neuropathic pain in Anderson-Fabry disease: Pathology and therapeutic options. European Journal of Pharmacology, 429, 121–125.
Mogil, J.S., Wilson, S.G., Chesler, E.J., Rankin, A.L., Nemmani, K.V.S., Lariviere, W.R., Groce, M.K., et al. (2003). The melanocortin-1 receptor gene mediates female-specific mechanisms of analgesia in mice and humans. Proceedings of the National Academy of Sciences, 100, 4867–4872.
Ploghaus, A., Becerra, L., Borras, C., Borsook, D. (2003). Neural circuitry underlying pain modulation: Expectation, hypnosis, placebo. Trends in Cognitive Sciences, 7, 197–200.
Ploghaus, A., Tracey, I., Gati, J.S., Clare, S., Menon, R.S., Matthews, P.M., Nicholas, J., et al. (1999). Disassociating pain from its anticipation in the human brain. Science 284(5422), 1979–1981.
Van de Beek, W.J. T., Roep, B.O., van der Slik, A.R., Giphart, M.J., Van Hilten, B.J. (2003). Susceptibility loci for complex regional pain syndrome. Pain, 103, 93–97.
Willis, R.C., Lesney, M.S. (2003). Pharmacogenomics: Looking toward the payoff. Modern Drug Discovery, July, 30–34.
Zubieta, J.K., Heitzeg, M.M., Smith, Y.R., Bueller, J.A., Xu, K., Yanjun, X., Koeppe, R.A., et al. (2003). COMT val(158)met genotype affects mu-opioid neurotransmitter responses to a pain stressor. Science 299(5610), 1240–1243.
The views expressed in this article are those of the authors. No official support by the U.S Food and Drug Administration nor the National Institutes of Health is provided or should be inferred.
James Witter, MD PhD, is clinical team leader in the division of analgesics, anti-inflammatory, and ophthalmic drug products, of the Center for Drug Evaluation and Research (CDER) at the FDA.
Lee S. Simon, MD, is division director, analgesic, antiinflammatory, and opthalmologic drug products, ODEV/CDER/FDA, and associate professor of medicine, Harvard Medical School and Beth Israel Deaconess Medical Center.
Raymond Dionne, DDS PhD, is a senior investigator and chief of the pain and neurosensory mechanisms branch at the National Institute of Dental and Craniofacial Research, NIH.
