It is self evident that the recent explosive growth in our understanding of the molecular, cellular and system's mechanisms responsible for nociception and pain ...
Pain 77 (1998) 227–229
Editorial
Towards a mechanism-based classification of pain? Clifford J. Woolf*, Gary J. Bennett, Michael Doherty, Ronald Dubner, Bruce Kidd, Martin Koltzenburg, Richard Lipton, John D. Loeser, Richard Payne, Eric Torebjork Received 18 May 1998; accepted 1 June 1998
It is self evident that the recent explosive growth in our understanding of the molecular, cellular and system’s mechanisms responsible for nociception and pain has important implications for the clinical diagnosis and treatment of pain. A small group of independent basic scientists and clinicians met in New York in January 1998, for a wide ranging discussion on the possible need for and implications of a mechanism-based classification of pain. The group believed that acceptance of a mechanism-based classification could have profound implications: drugs may be developed which target distinct mechanisms, basic scientists may have new guidelines for experimental design, and clinicians may be eventually armed with more reliable and valid diagnostic tools for treatment and clinical investigation. Furthermore, a mechanism-based classification for clinical syndromes might generate testable hypotheses for selecting treatments which interact with specific mechanisms. We wish to initiate a wide debate on this important topic by highlighting what we consider to be some of the key issues. In general, taxonomies can be either natural or artificial. Examples of each, respectively, are the division of objects into animate or inanimate groups (which reflects order in nature) and a telephone book (which is merely a conventional way of listing peoples’ numbers and addresses). Natural taxonomies are based on theoretical ideas of how the world is organized. Artificial taxonomies provide convenient or practical methods for organizing the world. Consequently they do not easily facilitate the development of new ideas. A mechanism-based classification of pain requires a conceptual understanding of organization in nature, and would,
* Corresponding author. Neural Plasticity Research Group, Department of Anesthesia and Critical Care, Massachusetts General Hospital, 149 13th Street, Room 4903, Charlestown, MA 02129, USA Tel.: +1 617 7243622; fax: +1 617 7243632.
therefore, set a framework for scientific development. Current methods of classifying pain have, we believe, a number of major limitations. Pain syndromes are identified by parts of the body, duration, and causative agent. We believe that an anatomical-based classification of pain is limiting because the innervation of distinct anatomical regions is often analogous, bearing in mind differences of target organ innervated (e.g. skin vs. viscera), length of axon, myelination, etc. To the extent that universal mechanisms can be identified, anatomical differences should be disregarded in favor of mechanisms that apply to all parts of the body. The acute/chronic dichotomy is also not helpful. Acute and chronic do not readily differentiate mechanisms. The benign/malignant dichotomy too has no mechanistic basis for pain, although it will influence treatment strategies. Greater care needs also to be taken with the definition of terms such as allodynia and hyperalgesia. Both terms are a description of clinical symptoms and do not imply a mechanism. Allodynia (pain evoked by normally non-painful stimuli) is often used in the clinical context to refer to Ab-fiber mediated brush-evoked mechanical pain or an altered processing of sensory information in the CNS. However, there are several other distinct types of mechanical hypersensitivity that do not involve Ab fibers and probably no significant central reorganization, but which present as pain evoked by commonly non-painful stimuli. Reduction in threshold is not, therefore, useful by itself, for a mechanistic classification. From a practical perspective, clinicians use classification systems to predict treatment responses as well as prognosis and to search for risk factors and morbidities. Ultimately any classification system must be judged on its utility for clinical practice and research. The most powerful systems could be organized by mechanism, by disease or etiology. In the neuropathic disease category, at least, the disease classification system was considered by the group to fail to
0304-3959/98/$19.00 1998 International Association for the Study of Pain. Published by Elsevier Science B.V. PII S0304-3959 (98 )0 0099-2
228
C.J. Woolf et al. / Pain 77 (1998) 227–229
predict outcome and it was therefore believed to be time now to consider alternatives. Pain necessarily involves three different levels of classification – pain symptoms, pain mechanisms and pain syndromes. A syndrome is defined as taking into account what it known about aetiology, genetics, clinical symptoms, history and treatment response. Features necessary for a classification of symptoms and mechanisms of pain are quite different from the features one wants in a clinically useful classification of syndromes. A useful classification system needs to be reliable. Two people seeing the same patient should reach the same conclusions. There are two crucial features to developing a successful classification system: (1) truly operational criteria, and (2) the use of inclusion and exclusion criteria. The first feature is essential even if the level of knowledge of the mechanisms is good; the second feature holds because there is bound to be some degree of nosologic overlap. A classification system should also have validity, although this is difficult to achieve. Validity is an estimate of the degree to which the classification system corresponds to the underlying biology of the disorder being studied. Traditionally validity is defined with reference to some gold standard. The challenge in headache, psychiatry and pain classification is the absence of such a gold standard. The alternative approach is to use an iterative, fallible process of searching for and identifying symptom clusters, biological markers, history and treatment response. Another important feature for a classification system is its generalizability. The classification rules should apply to the milder, as well as the most severe forms of a disease or condition. The last feature is comprehensiveness. To be clinically useful there must be a place for every diagnosis in a classification system. With these features in mind it is important to include small-scale validation studies into the developing classification system from the beginning. The first challenge of the group was to attempt to identify broad categories of pain and the mechanisms that may operate to produce these different pains (Table 1) bearing in mind that a clinically meaningful system should also consider the role of affect, cognition, expectancy and environmental consequences in the genesis of this aspect of human behavior. The group explored what form a mechanism-based classification might take, but considered that this is a step that requires wider consultation and discussion to be really useful. Nevertheless a mechanism-based classification would enable the clinician to ask the following types of questions: What is the contribution of primary afferent sensitization to the pain? Is there evidence for ectopic discharges in the primary afferent neuron? Is there an involvement of the sympathetic nervous system? Is there evidence for CNS sensitization or CNS disinhibition? This is, we believe, a lot better than the description of ‘allodynia due to nerve injury’. Instead, it is a list of clinically useful mechanisms, albeit minimal, error-prone, and imperfect. Its utility is
dependent, however on being able to identify whether particular mechanisms are present in a particular patient or not, and whether the yes or no leads to different therapeutic paths. Once we identify mechanisms, this should help the evolution of more uniform descriptors for symptoms. A real challenge is to establish operational criteria for distinct mechanisms. Advances in this area will derive, therefore, from determining the symptoms that constitute a syndrome and finding mechanisms for these. For example, brush-evoked pain (a symptom), can be initiated by nerve injury and tissue injury, and there are several mechanisms involved in both. These mechanisms, in turn, will have certain commonalties: ongoing activity of nociceptors, and spinal cord and cortical changes. Although an anatomical-based classification of symptoms is undesirable, it is nevertheless clinically important to distinguish between different tissues, such as skin muscle, joint, and viscera. Many of the basic science approaches have concentrated on cutaneous rather than deep somatic pain. It is also important to differentiate between two of the major components of pain: stimulus independent pain (ongoing) and stimulus dependent hyperalgesia (e.g. allodynia). Pain does not occur independent of the brain and the contribution of psychological factors to pain perception needs to be part of any mechanism-based classification. Such an approach is likely to lead to specific pharmacological, surgical or physical therapy interventions for each identified mechanism involved in a particular syndrome. An analogous example is the way cardiologists provide cocktails of drugs for their patients which together combat the arrhythmia, contractility failure, and problems of circulation Table 1 Categories of pain and possible mechanisms Transient pain* Nociceptor specialization Tissue injury pain Primary afferent Sensitization Recruitment of silent nociceptors Alteration in phenotype Hyperinnervation CNS mediated Central sensitization recruitment, summation, amplification Nervous system injury pain Primary afferent Acquisition of spontaneous and stimulus-evoked activity by nociceptor axons and somata at loci other than peripheral terminals Phenotype change CNS mediated Central sensitization Deafferentation of 2nd order neurons Disinhibition Structural reorganization *Transient pain refers to the response to a noxious stimulus which does not produce long term sequelae, e.g. a pin prick.
C.J. Woolf et al. / Pain 77 (1998) 227–229
which together constitute heart failure. We may need to treat neuropathic pain by poly- rather than monotherapy, blocking ectopic activity with sodium channel blockers, central sensitization with NMDA receptor antagonists, augmenting inhibitory modulation with a2 agonists and sympathetic involvement with adrenergic antagonists, depending on which mechanism is operational in a syndrome, or better still if they can be identified, in an individual patient. Already NSAIDs and opiates are used synergistically to target two different mechanisms in cancer patients. A mechanism based classification of pain has implications for clinical drug development in terms both of defining targets and describing which patients are likely to benefit. The standard way of studying a population is to assemble the group on the basis of a classified disease-related symptom. A mechanism-based classification suggests carrying out a study on a population with the same symptom but not necessarily the same disease. The standard method will indicate that a particular drug is efficacious for postherpetic neuralgia, for example, but it may suggest nothing about diabetic neuropathy. Testing a drug across a variety of aetiologies will identify a drug’s generalized ability to treat that same symptom across different disease states. A compound that interferes with heat transduction of nociceptors may be of use, for example, whenever heat hyperalgesia is present, regardless of whether this occurs after tissue injury or nerve damage. Perhaps, in the future, when the analgesic armamentarium includes drugs that act on specific mechanisms, it might be possible to judge patient response to these therapies as diagnostic of underlying mechanisms.
229
In creating a mechanism-based classification care must be exercised in extrapolating from animal models to clinical mechanisms. Animal models are only approximations. Ultimately, it is advisable to align clinical and human models with animal models to be certain of dissecting a particular mechanism. Caution also needs to be exercised when grouping people based on symptomatology and then testing these people with a drug irrespective of their underlying disease. The underlying disease must still be treated by whatever means possible. There must be disease modifying therapy as well as pain reducing therapy. In conclusion, a mechanism-based classification need not be an alternative to existing classifications but could be seen to be a new development running parallel with the standard classification, acting as a research tool. Any such classification will necessarily change with increased understanding of mechanisms but even at this stage simply to encourage pain clinicians to explore what mechanisms are present in their patients and to try to use this information to optimize treatment, would be a useful benefit. We offer these ideas to the pain community in the hope that vigorous discussion of the issues may lead to the development of such a classification and to an exploration of its utility. A workshop on this topic will be held at the IUPS meeting in Vienna and we welcome input. This meeting was funded by an unrestricted grant from Glaxo Wellcome.
