Expect Analgesic Failure, But Seek Success
Posted by SB. Leavitt, MA, PhD Thursday, July 11, 2013
A better understanding of potentially high therapeutic failure rates in pain management may be a first step toward doing better with currently available treatments. Clinically, this means expecting analgesic failure, assessing pain, and considering options for stopping and switching therapies. This also requires casting aside a reliance on what works for “average” patients, and asking what works best, for whom, in what circumstances.
Most analgesic medications work well, but in only a relatively small percentage of people, according to Andrew Moore from Oxford University and colleagues writing in the British Medical Journal[Moore et al. 2013]. They propose a transformation in thinking about how analgesic efficacy and harm should be assessed, and suggest several practical implications of a better understanding and appreciation of therapeutic failure rates:
“No single drug will treat successfully more than a minority of patients with a painful condition.
Successful pain relief is also likely to improve sleep, depression, fatigue, quality of life, function, and ability to work.
Experience (and some evidence) suggests that failure with one drug does not necessarily mean failure with others, even within a class.
We do not know the best order in which to use drugs, in terms of efficacy, harm, or cost.
Success or failure can be determined within 2-4 weeks, and success, when achieved, tends to be long lasting.
Because success rates are low, a wide range of drugs is needed to do the best for most patients, especially in complex chronic conditions.”
Measuring Success
Individual patient responses to any therapy vary greatly, Moore et al. observe. Pain relief measurements delineating successful outcomes typically are not distributed along a normal bell-shaped curve, but are usually bimodal; that is, most patient responses are either very good (above 50% pain relief) or very poor (below 15%). Therefore, the frequency distribution curve is more “U-shaped”; rather than the classic bell curve in which most responses fall toward the center and correspond with the mean (average).
Due to the U-shaped response distribution, research outcomes based on averages are unhelpful and misleading since “average” pain relief is actually experienced by few, if any, patients. The mean score tells us nothing about how many patients will experience clinically useful pain relief; hence, Moore and colleagues suggest that research should be moving toward “responder analyses” — focusing and reporting on the proportion of patients achieving outcomes that patients themselves consider to be worthwhile.
In that regard, the authors observe that patients want large reductions in pain intensity (typically at least 50% relief and ideally no worse than mild pain), with amelioration of associated problems, such as sleep disturbance and depression, but without common adverse events interfering with treatment. Patients who get better (responders) typically do well, experiencing improvements in fatigue, depression, and sleep interference, plus better function and quality of life. Non-responders gain none of those benefits.
From a research perspective, as well as in daily practice, the authors suggest that all persons who discontinue treatment for any reason should be considered as non-responders. Furthermore, the scientific assessment of analgesia and the clinical practice of analgesic delivery could be simplified into 3 guiding principles: A) measure pain in individual patients, B) expect analgesic drugs to fail to provide a good response in most patients, and C) prepare for the next step if and when failure occurs.
Defining Analgesic Failure
In their article, Moore et al. examine some drug-specific success and failure rates for postoperative pain, migraine, and chronic musculoskeletal and neuropathic conditions, using data predominantly from good quality reviews and meta-analyses. In a table, they list outcomes for 44 studies evaluating placebo in comparison with NSAIDs and various other drugs (eg, acetaminophen, triptans, antidepressants, antiepileptics, and others). Only 2 studies of opioid monotherapy were included, both for chronic noncancer pain.
Overall, and with but a few exceptions, less than half of patients achieved at least a 50% reduction in pain intensity (responder definition), and failure rates were highest over the long term in patients with chronic pain conditions. Of the 44 studies, success rates were above 50% for only 4 drugs in acute postoperative pain (acetaminophen + ibuprofen; acetaminphen + oxycodone; etoricoxib; ibuprofen + codeine) and 1 drug for migraine (zolmitriptan). For all other drugs and in all other conditions, fewer than half of patients achieved at least a 50% reduction in pain intensity.
Analgesic failure rates generally ranged from 55% to ≥87%. Data for opioids in chronic noncancer pain were available only for tapentadol and oxycodone in a combined analysis of osteoarthritis and chronic low back pain trials; tapentadol (200-500mg) had a failure rate of 90% and oxycodone (40-100mg) had a failure rate of 100%. Those rates took into account therapeutic responders compared with placebo responders, and it should be noted that in absolute terms, on their own, 30% of tapentadol-group patients and 21% taking oxycodone did experience analgesic success.
As Moore et al. observe, “The magnitude of the failure to achieve good pain relief, especially over the longer term in chronic pain, is sobering.” The high failure rates reported in their paper are a consequence of using patient-centered definitions of benefit combining a significant level of pain relief (>50%) with tolerable adverse events (ie, allowing continuation in therapy), using high standards of evidence, and avoiding major imputation bias (ie, focusing on responders). These higher standards are backed by considerable evidence supporting their validity, but they do portray less favorable outcomes than are often reported in the research literature.
Moving Toward Pragmatic Approaches
The use of responder analyses changes judgments of benefit and risk. In cases of therapeutic failure, patients without benefit should be exposed to no risk because the drug is stopped when they drop out of treatment. The good news is that success is often achieved within the first 2 weeks or so of treatment or not at all, the authors note, and benefits tend to be enduring. Obviously, only successfully effective drugs should continue to be prescribed.
Of some importance regarding chronic pain, Moore et al. observe that typical clinical trials may inadvertently underestimate treatment efficacy if the data are closely examined. Fixed-dose regimens may exacerbate adverse events and discontinuations, resulting in higher failure rates. An alternate approach would be to allow patient-directed drug titration to achieve adequate pain relief with tolerable adverse events; at that point, only subjects with treatment success (responders) would be randomized blindly between continuing therapy and placebo.
Such trial designs would have lower failure rates and more directly mimic what occurs in clinical practice. Additionally, the authors continue, drug therapy is rarely the only treatment used for chronic pain; however, clinical trials designed for regulatory purposes consider only single, or unimodal interventions.
A most essential pragmatic implication of high failure rates is that populations with pain need access to a broad range of analgesics and/or other interventions to have a better chance of success. According to Moore et al., the problem is a dearth of research data to help in devising therapy starting, stopping, and switching rules. In other conditions, like depression, switching medications is often effective; randomized trials have shown that any antidepressant used initially may benefit fewer than half of patients, but the majority can benefit when failures are followed by switching to other medications for depression.
Practical Implications of Therapeutic Failures
Essential practice principles for pain management should include assessing pain, expecting and recognizing analgesic failure, and reacting to it by pursuing analgesic success rather than blindly accepting failure. In any condition, the order in which analgesics should be tried is predicated on efficacy and safety, and adjusted for individual patient characteristics and response, suggest Moore and colleagues.
The authors further observe that guidelines developers often restrict treatment recommendations to 1 or 2 drugs for any pain condition. The developers consider similar drugs to operate as a class, overlooking the fact that there can be important differences in pharmacokinetics or drug interactions across similar medications. Less restrictive guidance recommendations, centered on patient-practitioner interactions — taking into account clinical wisdom as well as available evidence — may do better, Moore et al. affirm.
The authors additionally suggest that regulatory authorities need to recognize that therapeutic failure is the norm and set standards of acceptance based on real-world expectations. For example, Moore et al. note that European regulators have refused to license any drug for fibromyalgia because of inadequate average effect sizes, ignoring the fact that these drugs work well (≥50% reduction in pain intensity) for treating this difficult condition in around 10% of patients. New drugs are unlikely to be much better, the authors suggest, so a change in regulatory attitudes is overdue, would be sensible, and will benefit patients.
Finally, Moore and colleagues acknowledge that chronic pain conditions are complex and associated with considerable comorbidity. Coupled with the nuances of neurobiological pain modulation, central nervous system transformations, and genetic influences, high failure rates with single pharmacologic interventions are unsurprising. “The new game in town is specificity of effect for specific targets, but with only a small percentage of patients benefiting,” they state. “We need to determine how best to use the interventions we have to provide better care for more people at lower cost.”
COMMENTARY:
Assertions about the importance of failure are somewhat unusual in scientific discourse; yet, Moore and colleagues believe that pain medicine has reached a degree of maturity where it can constructively confront, embrace, and learn from better understandings of therapeutic failings. That may or may not be the case; for example, some current arguments against opioid analgesics for chronic noncancer pain seem to demand that either the therapy works well and for all patients or it is unacceptable — there is no middle-ground or acceptance of failure.
The studies examined by Moore et al. in their paper, with the respective analgesic failure rates, are exemplary but not exhaustive of the possibilities — eg, dose/frequency variations and combinations of different agents — when it comes to effective pharmacotherapy for various pain conditions. Nor were the studies critiqued from quality-of-evidence perspectives; yet, there are important lessons to be learned.
Additionally, a soon to be published paper by Moore, as sole author [2013, see ref below], further explores some of the essential principles and is worth examining. Plus, readers with a deeper interest in evidence-based pain management should be following our series on “Making Sense of Pain Research” [see article listing here], which discusses the many factors affecting research quality.
Here is further comment on several of the critical points that emerge from the papers by Moore et al. [2013] and Moore [2013]:
Analgesic failure refers to the clinical reality that any medication (or other intervention) will not work for all patients all of the time. In most cases, as demonstrated by the research evidence provided by Moore et al., the very best analgesics provide ≥50% pain relief in roughly half of treated patients. This amounts to a number-needed-to-treat (NNT) of about 2 compared with placebo; that is, for every 2 patients treated with active drug rather than placebo 1 additional patient will benefit.
For almost any medication, an NNT=2 would be a large and clinically significant effect size; yet it does not begin to approach the 100% response rate that patients and many practitioners may desire or expect. Furthermore, Moore et al. found that NNTs for various analgesics can range widely up to NNT>100, depending on the pain condition; however, even in the worst of cases, a certain percentage of patients (albeit, possibly extremely small) can benefit.
The important message is that therapeutic failure is quite common and in significant proportions of patients, but this should not deter pursuing another analgesic within the same or different class of drug. Oftentimes, expectations need to be lowered to the level of clinical reality; eg, while long-term opioid therapy may not benefit all patients with chronic noncancer pain, it does help a certain proportion of patients and significantly so.
There is no such thing as an “average” patient, even though research trials tend to define outcomes based on average, or mean, scores on efficacy measures. In doing this, research can be misleading, since the greatest proportions of patients either experience successful outcomes (eg, more than 50% pain relief) or very poor results (eg, less than 15% pain relief). Interestingly, mean placebo responses tend to follow the same pattern, which can result in small absolute differences between active therapy and placebo overall (hence increasing NNT values to less significant levels [NNT is calculated by 1 divided by the absolute difference between groups on a measure]).
The most vital question is: what works, for whom, in what circumstances? Along with that, it must be remembered that pain relief is but one measure of therapeutic success that may be meaningful to patients.
In overcoming a slavish reliance on “averages,” Moore et al. appear to be advocating for “per-protocol” analyses in pain research trials that focus on “responders.” That is, subjects who drop out of a study for any reason are considered as therapeutic failures, even if they had achieved some degree of pain relief at the time of their discontinuation. This recognizes that benefits outweigh risks in treatment responders — they achieve desired outcomes with tolerable adverse effects, if any, over an extended period of time.
In contrast, much of the pain research uses “intention-to-treat, or ITT” approaches, whereby outcomes in all subjects are taken into account whether or not they complete the trial. In some cases, the last observation (eg, pain score) prior to discontinuation is carried forward (LOCF) in final analyses, as if it represents an overall therapeutic effect.
Moore and colleagues argue that this approach interjects bias into the analyses, and may be acceptable for statistically determining if an intervention has any analgesic effect, but not for determining clinical effectiveness for individual patients. In other words, if patients cannot continue with a therapy for some reason they will not realize further benefits; the fact that they might have achieved some benefits up to a point in time, but then had to drop out prematurely due to adverse effects, may rescind the value of the therapy.
As much as anything, Moore and Moore et al. are advocating for a pragmatic, patient-centered, practice-oriented approach to understanding therapeutic failure and success in pain management. This might be somewhat of a paradigm shift for how research in the pain field is conducted and how regulatory and other bodies reach decisions on drug approvals and labeling. In essence, an unbiased focus on therapeutic response requires that treatments should be stopped when they do not work to avoid undue exposures to risks; however, those patients who do respond sufficiently, no matter how small in numbers, can experience large benefits to offset against rare but potentially serious harms. A degree of failure should not defeat the pursuit of success when it comes to pain management.
REFERENCES:
> Moore A, Derry S, Eccleston C, Kalso E. Expect analgesic failure; pursue analgesic success. BMJ. 2013;346:f2690 [abstract here].
> Moore RA. What works for whom? Determining the efficacy and harm of treatments for pain. PAIN. 2013(Mar); online ahead of print [abstract here].
http://updates.pain-topics.org/2013/07/expect-analgesic-failure-but-seek.html
Posted by SB. Leavitt, MA, PhD Thursday, July 11, 2013
A better understanding of potentially high therapeutic failure rates in pain management may be a first step toward doing better with currently available treatments. Clinically, this means expecting analgesic failure, assessing pain, and considering options for stopping and switching therapies. This also requires casting aside a reliance on what works for “average” patients, and asking what works best, for whom, in what circumstances.
Most analgesic medications work well, but in only a relatively small percentage of people, according to Andrew Moore from Oxford University and colleagues writing in the British Medical Journal[Moore et al. 2013]. They propose a transformation in thinking about how analgesic efficacy and harm should be assessed, and suggest several practical implications of a better understanding and appreciation of therapeutic failure rates:
“No single drug will treat successfully more than a minority of patients with a painful condition.
Successful pain relief is also likely to improve sleep, depression, fatigue, quality of life, function, and ability to work.
Experience (and some evidence) suggests that failure with one drug does not necessarily mean failure with others, even within a class.
We do not know the best order in which to use drugs, in terms of efficacy, harm, or cost.
Success or failure can be determined within 2-4 weeks, and success, when achieved, tends to be long lasting.
Because success rates are low, a wide range of drugs is needed to do the best for most patients, especially in complex chronic conditions.”
Measuring Success
Individual patient responses to any therapy vary greatly, Moore et al. observe. Pain relief measurements delineating successful outcomes typically are not distributed along a normal bell-shaped curve, but are usually bimodal; that is, most patient responses are either very good (above 50% pain relief) or very poor (below 15%). Therefore, the frequency distribution curve is more “U-shaped”; rather than the classic bell curve in which most responses fall toward the center and correspond with the mean (average).
Due to the U-shaped response distribution, research outcomes based on averages are unhelpful and misleading since “average” pain relief is actually experienced by few, if any, patients. The mean score tells us nothing about how many patients will experience clinically useful pain relief; hence, Moore and colleagues suggest that research should be moving toward “responder analyses” — focusing and reporting on the proportion of patients achieving outcomes that patients themselves consider to be worthwhile.
In that regard, the authors observe that patients want large reductions in pain intensity (typically at least 50% relief and ideally no worse than mild pain), with amelioration of associated problems, such as sleep disturbance and depression, but without common adverse events interfering with treatment. Patients who get better (responders) typically do well, experiencing improvements in fatigue, depression, and sleep interference, plus better function and quality of life. Non-responders gain none of those benefits.
From a research perspective, as well as in daily practice, the authors suggest that all persons who discontinue treatment for any reason should be considered as non-responders. Furthermore, the scientific assessment of analgesia and the clinical practice of analgesic delivery could be simplified into 3 guiding principles: A) measure pain in individual patients, B) expect analgesic drugs to fail to provide a good response in most patients, and C) prepare for the next step if and when failure occurs.
Defining Analgesic Failure
In their article, Moore et al. examine some drug-specific success and failure rates for postoperative pain, migraine, and chronic musculoskeletal and neuropathic conditions, using data predominantly from good quality reviews and meta-analyses. In a table, they list outcomes for 44 studies evaluating placebo in comparison with NSAIDs and various other drugs (eg, acetaminophen, triptans, antidepressants, antiepileptics, and others). Only 2 studies of opioid monotherapy were included, both for chronic noncancer pain.
Overall, and with but a few exceptions, less than half of patients achieved at least a 50% reduction in pain intensity (responder definition), and failure rates were highest over the long term in patients with chronic pain conditions. Of the 44 studies, success rates were above 50% for only 4 drugs in acute postoperative pain (acetaminophen + ibuprofen; acetaminphen + oxycodone; etoricoxib; ibuprofen + codeine) and 1 drug for migraine (zolmitriptan). For all other drugs and in all other conditions, fewer than half of patients achieved at least a 50% reduction in pain intensity.
Analgesic failure rates generally ranged from 55% to ≥87%. Data for opioids in chronic noncancer pain were available only for tapentadol and oxycodone in a combined analysis of osteoarthritis and chronic low back pain trials; tapentadol (200-500mg) had a failure rate of 90% and oxycodone (40-100mg) had a failure rate of 100%. Those rates took into account therapeutic responders compared with placebo responders, and it should be noted that in absolute terms, on their own, 30% of tapentadol-group patients and 21% taking oxycodone did experience analgesic success.
As Moore et al. observe, “The magnitude of the failure to achieve good pain relief, especially over the longer term in chronic pain, is sobering.” The high failure rates reported in their paper are a consequence of using patient-centered definitions of benefit combining a significant level of pain relief (>50%) with tolerable adverse events (ie, allowing continuation in therapy), using high standards of evidence, and avoiding major imputation bias (ie, focusing on responders). These higher standards are backed by considerable evidence supporting their validity, but they do portray less favorable outcomes than are often reported in the research literature.
Moving Toward Pragmatic Approaches
The use of responder analyses changes judgments of benefit and risk. In cases of therapeutic failure, patients without benefit should be exposed to no risk because the drug is stopped when they drop out of treatment. The good news is that success is often achieved within the first 2 weeks or so of treatment or not at all, the authors note, and benefits tend to be enduring. Obviously, only successfully effective drugs should continue to be prescribed.
Of some importance regarding chronic pain, Moore et al. observe that typical clinical trials may inadvertently underestimate treatment efficacy if the data are closely examined. Fixed-dose regimens may exacerbate adverse events and discontinuations, resulting in higher failure rates. An alternate approach would be to allow patient-directed drug titration to achieve adequate pain relief with tolerable adverse events; at that point, only subjects with treatment success (responders) would be randomized blindly between continuing therapy and placebo.
Such trial designs would have lower failure rates and more directly mimic what occurs in clinical practice. Additionally, the authors continue, drug therapy is rarely the only treatment used for chronic pain; however, clinical trials designed for regulatory purposes consider only single, or unimodal interventions.
A most essential pragmatic implication of high failure rates is that populations with pain need access to a broad range of analgesics and/or other interventions to have a better chance of success. According to Moore et al., the problem is a dearth of research data to help in devising therapy starting, stopping, and switching rules. In other conditions, like depression, switching medications is often effective; randomized trials have shown that any antidepressant used initially may benefit fewer than half of patients, but the majority can benefit when failures are followed by switching to other medications for depression.
Practical Implications of Therapeutic Failures
Essential practice principles for pain management should include assessing pain, expecting and recognizing analgesic failure, and reacting to it by pursuing analgesic success rather than blindly accepting failure. In any condition, the order in which analgesics should be tried is predicated on efficacy and safety, and adjusted for individual patient characteristics and response, suggest Moore and colleagues.
The authors further observe that guidelines developers often restrict treatment recommendations to 1 or 2 drugs for any pain condition. The developers consider similar drugs to operate as a class, overlooking the fact that there can be important differences in pharmacokinetics or drug interactions across similar medications. Less restrictive guidance recommendations, centered on patient-practitioner interactions — taking into account clinical wisdom as well as available evidence — may do better, Moore et al. affirm.
The authors additionally suggest that regulatory authorities need to recognize that therapeutic failure is the norm and set standards of acceptance based on real-world expectations. For example, Moore et al. note that European regulators have refused to license any drug for fibromyalgia because of inadequate average effect sizes, ignoring the fact that these drugs work well (≥50% reduction in pain intensity) for treating this difficult condition in around 10% of patients. New drugs are unlikely to be much better, the authors suggest, so a change in regulatory attitudes is overdue, would be sensible, and will benefit patients.
Finally, Moore and colleagues acknowledge that chronic pain conditions are complex and associated with considerable comorbidity. Coupled with the nuances of neurobiological pain modulation, central nervous system transformations, and genetic influences, high failure rates with single pharmacologic interventions are unsurprising. “The new game in town is specificity of effect for specific targets, but with only a small percentage of patients benefiting,” they state. “We need to determine how best to use the interventions we have to provide better care for more people at lower cost.”
COMMENTARY:
Assertions about the importance of failure are somewhat unusual in scientific discourse; yet, Moore and colleagues believe that pain medicine has reached a degree of maturity where it can constructively confront, embrace, and learn from better understandings of therapeutic failings. That may or may not be the case; for example, some current arguments against opioid analgesics for chronic noncancer pain seem to demand that either the therapy works well and for all patients or it is unacceptable — there is no middle-ground or acceptance of failure.
The studies examined by Moore et al. in their paper, with the respective analgesic failure rates, are exemplary but not exhaustive of the possibilities — eg, dose/frequency variations and combinations of different agents — when it comes to effective pharmacotherapy for various pain conditions. Nor were the studies critiqued from quality-of-evidence perspectives; yet, there are important lessons to be learned.
Additionally, a soon to be published paper by Moore, as sole author [2013, see ref below], further explores some of the essential principles and is worth examining. Plus, readers with a deeper interest in evidence-based pain management should be following our series on “Making Sense of Pain Research” [see article listing here], which discusses the many factors affecting research quality.
Here is further comment on several of the critical points that emerge from the papers by Moore et al. [2013] and Moore [2013]:
Analgesic failure refers to the clinical reality that any medication (or other intervention) will not work for all patients all of the time. In most cases, as demonstrated by the research evidence provided by Moore et al., the very best analgesics provide ≥50% pain relief in roughly half of treated patients. This amounts to a number-needed-to-treat (NNT) of about 2 compared with placebo; that is, for every 2 patients treated with active drug rather than placebo 1 additional patient will benefit.
For almost any medication, an NNT=2 would be a large and clinically significant effect size; yet it does not begin to approach the 100% response rate that patients and many practitioners may desire or expect. Furthermore, Moore et al. found that NNTs for various analgesics can range widely up to NNT>100, depending on the pain condition; however, even in the worst of cases, a certain percentage of patients (albeit, possibly extremely small) can benefit.
The important message is that therapeutic failure is quite common and in significant proportions of patients, but this should not deter pursuing another analgesic within the same or different class of drug. Oftentimes, expectations need to be lowered to the level of clinical reality; eg, while long-term opioid therapy may not benefit all patients with chronic noncancer pain, it does help a certain proportion of patients and significantly so.
There is no such thing as an “average” patient, even though research trials tend to define outcomes based on average, or mean, scores on efficacy measures. In doing this, research can be misleading, since the greatest proportions of patients either experience successful outcomes (eg, more than 50% pain relief) or very poor results (eg, less than 15% pain relief). Interestingly, mean placebo responses tend to follow the same pattern, which can result in small absolute differences between active therapy and placebo overall (hence increasing NNT values to less significant levels [NNT is calculated by 1 divided by the absolute difference between groups on a measure]).
The most vital question is: what works, for whom, in what circumstances? Along with that, it must be remembered that pain relief is but one measure of therapeutic success that may be meaningful to patients.
In overcoming a slavish reliance on “averages,” Moore et al. appear to be advocating for “per-protocol” analyses in pain research trials that focus on “responders.” That is, subjects who drop out of a study for any reason are considered as therapeutic failures, even if they had achieved some degree of pain relief at the time of their discontinuation. This recognizes that benefits outweigh risks in treatment responders — they achieve desired outcomes with tolerable adverse effects, if any, over an extended period of time.
In contrast, much of the pain research uses “intention-to-treat, or ITT” approaches, whereby outcomes in all subjects are taken into account whether or not they complete the trial. In some cases, the last observation (eg, pain score) prior to discontinuation is carried forward (LOCF) in final analyses, as if it represents an overall therapeutic effect.
Moore and colleagues argue that this approach interjects bias into the analyses, and may be acceptable for statistically determining if an intervention has any analgesic effect, but not for determining clinical effectiveness for individual patients. In other words, if patients cannot continue with a therapy for some reason they will not realize further benefits; the fact that they might have achieved some benefits up to a point in time, but then had to drop out prematurely due to adverse effects, may rescind the value of the therapy.
As much as anything, Moore and Moore et al. are advocating for a pragmatic, patient-centered, practice-oriented approach to understanding therapeutic failure and success in pain management. This might be somewhat of a paradigm shift for how research in the pain field is conducted and how regulatory and other bodies reach decisions on drug approvals and labeling. In essence, an unbiased focus on therapeutic response requires that treatments should be stopped when they do not work to avoid undue exposures to risks; however, those patients who do respond sufficiently, no matter how small in numbers, can experience large benefits to offset against rare but potentially serious harms. A degree of failure should not defeat the pursuit of success when it comes to pain management.
REFERENCES:
> Moore A, Derry S, Eccleston C, Kalso E. Expect analgesic failure; pursue analgesic success. BMJ. 2013;346:f2690 [abstract here].
> Moore RA. What works for whom? Determining the efficacy and harm of treatments for pain. PAIN. 2013(Mar); online ahead of print [abstract here].
http://updates.pain-topics.org/2013/07/expect-analgesic-failure-but-seek.html
