Pain measures and cut‐offs – ‘no worse than mild pain’ as a simple, universal outcome

Now admittedly the patient in question had undergone a life-saving operation in a brand new hospital staffed by some wonderful and talented people. However, a pain score of 6/10 is not mild, but borderline between moderate and severe, and the patient did need something for that. Was the nurse ignorant of what pain scores meant, or was it just that caring professionals typically underestimate patients' pain 1? Both these questions are deserving of research, but for this patient, the only point is that the system failed him, and for him the reasons behind the failure were of little interest. He had pain that was borderline severe, and it was not treated. This is not uncommon. A fairly recent survey of Italian hospital wards came to the hardly original conclusion that those wards in which less analgesic was prescribed had higher rates of patients experiencing severe pain than those where more analgesics were prescribed 2; Fig. 1 shows the clear inverse relationship between the presence of severe pain and the percentage of patients treated for their pain. There is a wealth of evidence that pain is poorly treated, and that significant proportions of patients suffer from moderate or severe pain, whether it is acute pain in hospital 3 or chronic pain in the community 4, 5. Barriers to progress are many and varied, but one particular and important barrier is a degree of confusion about pain scoring systems and what they mean. How does a simple categorical verbal rating system (no pain, mild, moderate, severe) relate to a 100-mm visual analogue scale (VAS) or an 11-point numerical rating scale (NRS)? Does it matter what the anchors are at each end? Where are the boundaries for moderate or severe pain? What is the minimal clinically significant difference? Does it matter whether the scale has been ‘validated’ in Welsh, or Farsi, or Urdu? All of which makes for terrific grist for the mill of grant applications, but does not make things easy for clinical practice, because it suggests that there is considerable uncertainty over pain measurement. Pain scales continue to proliferate, not least because they can be copyrighted and significant sums paid to use them in clinical trials, as do academic arguments about whether, on average, a few millimetres change on a 100-mm scale represents a useful outcome 6. One outcome of all this is additional uncertainty over which scale is ‘right’. All this academic angst is unimportant for everyday practice, where the simple question is whether pain is present at an unacceptable level or not. It may be time to consider a divorce between all these interesting academic complexities and the theoretically simple, but practically difficult requirements for successfully treating patients with pain. The philosophical principle of ontological parsimony, rather easier understood as keeping it simple, was suggested about 700 years ago by William of Ockham, and more recently restated as the KISS principle by Kelly Johnson 7. We (the authors) are increasingly of the opinion that in measuring pain there should be a simple principle – that only ‘no worse than mild pain’ is acceptable in clinical practice, and the important practical outcome of ‘no worse than mild pain’ is the only outcome of interest to be taken from clinical trials. We are relaxed about just how ‘no worse than mild pain’ is measured, but take the view that the borderline between mild and moderate pain is about 30 mm on the 100-mm VAS 8 in acute pain or 3/10 on a 0–10 NRS. Figure 2 shows data from almost 14 000 paired observations taken at rest or on activity over a 48-h postoperative period in surgical patients that would agree with 30 mm as a cut-off point (data from 9); other analyses give similar if slightly different cut-off points 10. There is some evidence that strict adherence to ‘no worse than mild pain’ principles can lead to virtual elimination of severe pain in hospital 11. The same cut-off point works well for chronic pain in the community 12, and pain measures work at least as well as more complicated scales in conditions such as inflammatory arthritis 13. Electronic searching alone is known to retrieve only a minority of observational studies relevant to a research question 14, 15. Experience shows that broader types of searching can capture many more studies 14, 16. We therefore conducted a search to find studies informing on the three main themes above. These searches comprised a series of different free text searches of PubMed (to November 2012), with follow-up on any potentially useful publication using the ‘related citation’ facility. For useful publications, we also checked on citations of that publication using Google Scholar™. In addition to electronic searches, retrieved articles were read for any other sources of data, as were general review articles and book chapters. When asked what they would consider treatment success, patients with chronic pain specify a large reduction in pain intensity, by 50% or more 17-19. Their ideal outcome is pain intensity of 3/10 or below on a 0–10 NRS, or its equivalent when pain is rated categorically, i.e. no worse than mild pain. They also want substantial reductions in fatigue, distress and the loss of quality of life that accompanies chronic pain. Chronic pain patients want mean decrements in excess of 50% on measures of interference on either the Brief Pain Inventory (BPI) or the Multidimensional Pain Inventory (MPI) 20. Patients in this latter study thought that an acceptable level of pain was around 3–4/10, no worse than mild pain. Much the same is true in migraine, where the outcome specified is that of complete pain relief 21. Patients would therefore agree that a clinically important difference in pain outcomes would be at least the 33% level suggested in breakthrough pain 22, the 30/100 mm pain reduction defined as adequate pain control in acute pain 23, or a more than 40 mm reduction in pain defined as much better in musculoskeletal pain 24. In fibromyalgia, pain severity reductions of about 40% were argued to be clinically important 25. For painful diabetic neuropathy and fibromyalgia, patients describing themselves as much or very much better typically had pain intensity reductions of 40% or more 26. These are far greater than the minimally important difference of a 6% reduction in pain suggested by rheumatoid arthritis patients 27. Pain outcomes in trials are usually described in terms of change in pain, rather than pain level at the end of a trial. As Fig. 3 demonstrates, patients can be improvement responders (50% pain reduction), state responders (have ‘no worse than mild pain’, below 30/100 mm on a 100-mm VAS), or both, or neither. There is potential complexity here, and arguments could be raised about whether pain improvement response or low pain state is better, or what ‘better’ means in this context. However, patients consistently say, when asked, that the response they want is either large reductions in pain intensity or being in a low pain state (no worse than mild pain), and ideally both, so these are patient reported outcome measures we need to take seriously. The evidence we have is that low pain state, no worse than mild pain, is consistently rated highly by patients in clinical trials when validated against global questions of response. In fibromyalgia, for example, being ‘very much improved’ at the end of trials lasting 8–14 weeks was associated with a pain score of 30/100 mm or below in 1858 completers (Fig. 4; data from 28). In the overall population of 2575 patients starting on treatment for fibromyalgia, 73% were non-responders by virtue of withdrawing from treatment or because of inadequate pain relief; only 27% were state responders only, improvement responders only, or both state and improvement responders (Fig. 5). There are indications that some of the more complex composite outcomes are no more informative than straightforward pain scores or patient global rating in chronic pain 29, although it is clearly the case that pain itself is not the only issue for patients with chronic pain 17, 30. In rheumatology, the patient acceptable symptom state (PASS) has been defined as the value beyond which patients consider themselves well. For osteoarthritis, the junction between satisfactory and unsatisfactory was about 32/100 mm 31, and similar results were obtained with numerical rating and function scales 32. In acute pain also, satisfaction is associated with pain scores generally below 30/100 mm, and low analgesic requirement using intravenous fentanyl via patient-controlled analgesia 9. In chronic pain, responders tend to keep being responders. Clinical experience is that responders typically show temporal stability in their response, as concluded by studies to specifically test consistency in chronic pain 33 and musculoskeletal pain 34. In fibromyalgia a long-term, enriched enrolment, randomised withdrawal study emphasised that individual response status was stable over six months 35. In chronic low back pain, initial responders to duloxetine had further significant improvement in pain and other outcomes for another 41 weeks 36. And in osteoarthritis, responders at two weeks overwhelmingly were responders at 12 weeks 37, and up to one year 38. While more evidence would be welcomed, the emerging picture is that ongoing response over the long term appears to occur in those who respond initially. Moreover, current evidence indicates that in osteoarthritis 39 and fibromyalgia 28 response, or lack of response, is largely determined within 2–4 weeks from the start of treatment. We are beholden to averages. As soon as we begin to learn about statistics, we are presented with a Gaussian distribution, and thereafter the implication is that distributions are always Gaussian, and the average is something applying – more or less – to most people. And that is why pain studies will almost always tell us the average pain score, or the average change in pain score, or the average postoperative analgesic requirement. The trouble is that while the world may be Gaussian for many things – height, for instance – it isn't Gaussian for others, like hair or eye colour. Nor is it Gaussian for pain or for response to analgesics. For example, while some people have moderate or severe pain soon after surgery, for many, pain is delayed by many hours, and about one in 20 may not get pain at all 40. Distributions other than Gaussian are found for many measures associated with pain. For example, analgesic consumption after surgery follows a highly skewed distribution in which mean, median, and mode are all very different from one another 41. However, it is in the response to analgesics in acute and chronic pain that we find the largest deviation from Gaussian distributions. Two classic examples are the response to analgesics in acute postoperative pain or osteoarthritis (Fig. 6). They have a clearly bimodal distribution with placebo or active drug, with some people getting a very good response, whereas others get virtually none 39, 42; few patients in either study could be described as average. The example shown in acute pain is for 120 mg etoricoxib, one of the most effective analgesics in this pain model, where good responses predominate. For most other analgesics at commonly used doses, the proportion with a good response is much less than this 43. For chronic pain also, the same bimodal distribution applies as for osteoarthritis 39, ankylosing spondylitis 44, chronic low back pain 45 and migraine 46, as well as in neuropathic pain 47. All these examples have used change in pain as their measure, with a reduction of at least 50%. While there are few analyses directly relating the level of pain change to a pain score of ‘no worse than mild pain’, it is probable that this covers most patients. There are uncertainties and grey areas, as the results of analyses by improvement responders and state responders in Fig. 5 show. Interpretation at the level of the individual may not always be easy; Fig. 7 shows 200 individual patient responses over 14 weeks in patients with fibromyalgia treated with pregabalin 450 mg per day (data from 28), some of the data from which were used to calculate the results in Fig. 5. Simple inspection demonstrates that large reduction in pain and low final pain score are highly associated. While more individual patient data analyses are needed to increase our understanding (the academic viewpoint), the broad message is that very large decrements in pain will result in a low pain state (the pragmatic take home message). In acute pain, inadequate pain management has substantial consequences for patients, with high pain scores associated with lower quality of life measures some four weeks after surgery 48, 49. Higher pain scores substantially impaired patients' sleep, sexual function and their ability to perform physical activities during the postoperative period 50, 51. It is also the case that there is a direct relationship between higher average postoperative pain levels and patient dissatisfaction with their postoperative experience 9. However, it remains unclear whether techniques to improve postoperative pain result in improved sleep, function and quality of life 52. We sought studies linking pain scores changes with quality of life changes in chronic pain treatment in a literature review; it found 13 studies with about 8000 patients treated with a variety of drugs for different chronic pain conditions (Table 1). Each individual study reported measurements of quality of life or function appropriate to the condition studied. The majority of the studies were individual patient analyses of randomised trials of appropriate duration for the condition being treated 53-64; one was a small trial lasting four weeks 65. Two long-duration prospective cohort studies (1387 patients) also provided some data 66, 67. Appropriate study duration is one important marker of clinical trial validity, as is the use of appropriate imputation methods 68, 69. Conditions studied included migraine, fibromyalgia, neuropathic pain, osteoarthritis, rheumatoid arthritis, chronic low back pain and ankylosing spondylitis. Treatments used included tumour necrosis factor (TNF) antagonists, tramadol/paracetamol, topical diclofenac, topiramate, pregabalin, duloxetine and placebo. Each study reported a positive association between good pain relief and measures such as quality of life, activities of daily living, function, work, enjoyment of life, global impression of benefit, depression, mood, sleep and fatigue (Table 1). A consistent finding was improvement in quality of life with successful treatment. Figure 8 shows the patterns of change in Short Form (36) Health Survey (SF-36) sections with responders experiencing at least 50% pain intensity reduction in fibromyalgia 59. The magnitude of the changes reported is far from trivial. For example, the largest individual patient analysis in fibromyalgia, with almost a third of the total number of patients providing evidence of an additional benefit of effective pain treatment, reports that pain intensity reduction of 50% or more is associated with reversion towards population norm values for sleep, fatigue, depression, with a trend for the SF-36 section scores to return towards the population normative values 59. Quality of life benefits were also significant, with EuroQol health status (EQ-5D) score increases over one year of 0.22 with TNF antagonists in rheumatoid arthritis 64, 0.35 for ≥ 50% pain intensity reduction in painful diabetic neuropathy 58 and a one-year quality-adjusted life year (QALY) gain of 0.11 for the same outcome in fibromyalgia (Fig. 9). In an analysis of tapentadol trials, patients who tolerated the treatment with tapentadol or oxycodone and completed the trial (and therefore likely to be those with good pain benefit) had EQ-5D average increments of 0.31 64. Similar benefits are likely to accrue from successful non-drug interventions in pain 70. A systematic review of QALYs for estimating effectiveness of healthcare reported the utility gains over six months to one year for various different interventions 71. Of the 31 examples reported where interventions actually worked, and where there were appropriate utility measures reported, most had one-year utilities of 0.1 or below. Some (total hip replacement, cochlear implants in children or profoundly deaf adults, TNF antagonists in rheumatoid arthritis) delivered large one-year gains of 0.2 or above, but only 9/31 (29%) had one-year gains above 0.1. The quality of life or utility gains found with successful treatment of chronic pain are clearly at least comparable, and often superior. It is important to note that while all the studies found reported at least one positive association between effective pain treatment and some aspect of quality of life or functioning, no study was found reporting research showing the absence of any link between pain relief and benefit. It is also important to note that these studies consistently report the lack of benefits in quality of life, functioning, work, sleep, or depression in the absence of good pain relief. The message is that reducing chronic pain to levels equivalent to ‘no worse than mild pain’ carries significant health and economic benefits to patients. It should normally be both more affordable and probably cost saving, compared with ineffective treatment. Improved sleep, reduced depression, better quality of life and greater ability to function and work come with good pain relief; without pain relief, there were no improvements in these outcomes. The strength of the evidence is such as to indicate that any patient-centred treatment programme that does not include the achievement of adequate pain relief as part of its goals is likely to fail to deliver on expected benefits. However, treating pain is clearly not easy; if it were, reports of moderate or severe acute pain in 44% of patients in hospital in the USA 72 and 91% with chronic pain living in the community in Australia 12 would not be so common. Investigations into the barriers to adequate treatment of acute pain suggest a variety of possible problems 73, but investigations into different levels of performance between centres come to no particular conclusion 74. Guidelines for treatment of chronic pain are less or more restrictive regarding the range of therapies allowed; for example, guidelines for the management of osteoarthritis pain 75 allow a broad range of therapies to be tried, whereas for neuropathic pain, available therapies are limited to just a few 76. It is unquestionable that there are difficulties in treating pain, regardless of its origin and mechanism. The growth in so-called evidence is not necessarily helping, as we in all about treatment of musculoskeletal pain, good evidence that its benefits over placebo are clinical whereas good evidence that a simple, effective and for in works well is largely while different to treatment are for particular or particular patients, the of treatment can be the of the evidence being It should be ‘no worse than mild pain’ as by the patient, not to professionals often get it most often by the pain the is not or the evidence is the of ‘no worse than mild pain’ will not have been The academic is to have trial data using ‘no worse than mild pain’ as an to clinical trial results to be useful for clinical in understanding and how not to be by are very and while they use outcomes of at least 50% pain intensity reduction rather than ‘no worse than mild pain’, the two are probably The practical is to improve treatment of acute and chronic pain. There are many to and many and professionals to use those for the patients in need of pain relief. It is not so much the that are but a clear understanding of the of treatment. This outcome of ‘no worse than mild pain’ is simple to and and is as it can be with one with the measures describing change would need at least two outcome worse than mild pain should be and should be a of analgesic or treatment We the many and from around the who have our and on work we have There was no for this came from the Pain