Thank you for the opportunity to join this interesting new event. I read with interest Jeff's comments related to a key question for those working with pain: "What is the reason for lack of translation of basic information into new treatment?" I have one general and two specific comments.
General: The lack of translational fruits is an issue for chronic pain. When we talk about acute pain there has been clear improvement in our handling of these patients, e.g acute postoperative management. The problem is the long-lasting, often persistent pain.
There are two additional points I would like to make although Jeff has alluded to them:
1. The temporal aspect. The processing of pain is dynamic and extremely plastic under physiological and certainly in pathological conditions with different changes occurring over seconds, minutes, days, and probably years . This is rarely (if ever) taken into account. There needs to be correspondence between the experiments observed in animals and in humans. For example in clinical trials there is recruitment of patients that may have had their pain for 3-6 months in one end to patients that have had pain for several decades in the other end. Given the plasticity of the nervous system one can easily envision that the processing of pain is altered even if the underlying etiology (e.g. post-herpetic neuralgia) is the same. Just recently has the point about reproducibility in human pain conditions been taken up (Geber et al., 2011), but much more is needed.
2. Wrong and irrelevant measures. Measures in animals and humans are usually unidimensional. Predictivity to drug X is based on non-comparable responses in rodents and in humans. For example In rodents with a CCI the measure can be a spinal withdrawal response to a von Frey filament applied to the hind paw. Although efforts are taken to look at different responses they mostly rely on spinal reflex behavior. It is unlikely that such responses can reflect the experience of pain in a patient who has lost his job and his family because of a plexus avulsion.
In humans the pain measure is also often unidimensional but in another sense. The pain measure is often based on as simple intensity recording of a pain experience or relief of such. In these cases clinical scientists try to squeeze the complex pain experience into a simplistic measure where the multidimensional aspect of pain is reduced to one dimension. What about other biomarkers?
References:
Geber C, Klein T, Azad S, Birklein F, Gierthmühlen J, Huge V, Lauchart M, Nitzsche D, Stengel M, Valet M, Baron R, Maier C, Tölle T, Treede RD. Test-retest and interobserver reliability of quantitative sensory testing according to the protocol of the German Research Network on Neuropathic Pain (DFNS): a multi-centre study. Pain. 2011 Mar;152(3):548-56. Epub 2011 Jan 14.
In his forum address, Jeff Mogil discusses several factors that may have contributed to a lack of translation from basic research to new pain treatments and asks for comments. As someone who spent a decade in the pharmaceutical industry in the area of early phase clinical trials of potentially new analgesic drugs, I would like to share my observations.
The pedigree of many compounds that have entered "proof-of-concept" clinical testing has been quite impressive. For example, compounds might have progressed to the clinic only if there was a high degree of selectivity for the target protein, the putative mechanism of action was defined and plausible for analgesia, pharmacokinetic (and sometimes functional) access to the CNS was confirmed, analgesic behavioral activity across multiple animal models and in multiple types of endpoints was demonstrated (often desired to be equal to or superior to "positive" controls known to be effective in at least one type of human pain condition). Yet the clinical failure rate has been extraordinarily high.
Do we get full value from animal models? The pharmaceutical industry has exploited animal models as a high throughput in vivo screen. The rush to progress compounds to the clinic overrides critically testing basic assumptions, such as causal evidence to support that the observed behavioral changes in the animal model directly relate to the putative mechanism of action of the compound under study. Avoidable, costly, premature or inappropriate clinical investments may be a symptom of the emphasis on speed and cost minimization at the preclinical stages (a false economy given the relative costs and timescales).
What constitutes a positive result in animal models? The declaration of an effect (statistical significance in mean change) may be insufficient for benchmarking an outcome of relevant or predictive value. We need to be more circumspect about which endpoints are important, the magnitude, completeness and duration of response normalization and the presence or absence of side effects.
What are effective or predictive doses? PK-PD modeling is based on ED50 (or EC50) estimates because this is relevant to comparative pharmacology methodology. But actual efficacy prediction and risk-benefit assessment may require exposure levels at ED95 or 3 to 5 x EC50. We just do not know if we are grossly underestimating target exposure.
Why are natural diseases in animals ignored? The very artificial nature of the models themselves may be a relevant issue. Natural diseases in animals provide access to abnormal nervous system tissues as well as potential pain models. A significant literature in pharmacological interventional trials in veterinary medicine demands more attention.
Who do we select for clinical trials? Negative clinical trial outcomes may have resulted from good studies in subjects with the wrong characteristics. A major difficulty in clinical trials for proof-of-efficacy is selecting the appropriate study population. Animal models have been poor at predicting what types of subjects to include or to exclude. Knowledge of a mechanism alone is insufficient. The results of the German Research Network on Neuropathic Pain indicate that most neurological abnormalities present across all types of neuropathic pain, suggesting that the common etiology-based approaches may be flawed. The choices made in selecting the study population and its characteristics may be crucial, but a meaningful basis to guide these choices a priori is lacking.
Is the effect size decreasing in clinical trials? Clinical trials are failing at the approval end as well. There are many possible reasons for this including a more "production line" approach to study site selection and to greater availability of approved therapies. However, the regulatory approval landscape has evolved significantly over the past decade, impacting the characteristics of acceptable clinical trial design and methodology of interpreting the results. Emphasis on analyses that account for trial non-completers and imputation methods that eliminate bias in the estimate of treatment effect and variance have a significant impact on the ability to design successful clinical trials, especially when the effect size is small. Changes in analysis methodology give the appearance of change in assay sensitivity.
Even a small improvement in the rate of translational success can make a huge difference to drug development efficiency. We need to be more circumspect about the animal models we use for decision making and how to interpret them.
Jeff's question is good and the comments posted so far are quite insightful and I agree with the points made; however there are structural problems in the health care system and the way research is reviewed and funded that create barriers to successful translation. The single biggest problem is that most of the time we don't know what's wrong with patients with chronic pain. Because of this we don't know whether any of animal models we use are valid or predictive. Why do we know so little about the patients? One reason is the vanishingly small group of physicians who both treat patients with chronic pain and understand the neurobiology of pain. The patients are complicated because of the unknown pain generating mechanism, the psychological responses to treatment failure and the possibility that certain medications can actually worsen pain. Unfortunately, there is no financial support for individuals who are willing to take the time to carefully and thoroughly examine patients. Without this it isn't possible to determine all the factors that contribute to their pain complaint. Other than serendipity, the careful examination of patients in the context of a sophisticated understanding of the neurobiology of pain is a fountain from which advances come. How do we create the incentive for physicians to spend more time learning neuroscience and to spend more time examining their patients and reading the relevant literature? I have no idea how to do this and I can't come up with a reasonable alternative.
Troels Jensen, Aarhus University Hospital