Failed back surgery syndrome (FBSS) is a therapy-refractory pain condition characterised by persistent low back pain after spinal surgery, associated with disability, low quality of life and high unemployment.1,2 Some 25 to 38% of patients develop FBSS.1,3–6 Given the high prevalence of chronic low back pain,7,8 the increase of spinal surgery9,10 and its high failure rate, the number of patients with FBSS is substantial and expected to rise. Considering this together with the high costs of spinal surgery11 and the associated burden of disease, it is important to identify presurgical predictors of poor surgical outcome. This would support decisions on the indication for surgery, alternative treatments and preventive measures.
We are currently unable to identify patients at risk of developing FBSS with an acceptable degree of confidence. Evidence on sociodemographic, psychological, clinical and surgery-related predictors is inconclusive.12,13 Therefore, investigating other contributing mechanisms is important. Prolonged or intense nociceptive input induces neuroplastic changes that lead to central nervous system hypersensitivity.14 This can be assessed using quantitative sensory tests (QST).15,16 Previous case–control studies found lower pain thresholds for QST in patients with chronic low back pain, compared with pain-free controls.17–19 This suggests that central hypersensitivity is involved in chronic low back pain. Theoretically, the surgical trauma may enhance these neuroplastic changes, thereby cancelling out the benefits of surgery and producing persistent pain.
We tested the hypothesis that central hypersensitivity as assessed by QST predicts FBSS making QST a potential tool for informing clinical decisions in patient selection for spinal surgery. Unlike any previous study, we included pain, disability, a composite endpoint of pain and disability,20–22 and also a comprehensive set of potential presurgical predictors.
The study protocol was approved by the Ethics Committee of the Canton of Bern, Switzerland (application number 176/11) and conducted in accordance with the Declaration of Helsinki.23 We obtained written informed consent from all participants. This article adheres to the applicable STROBE guideline.24 We recruited patients for this prospective cohort study at three tertiary care centres in Bern, Switzerland.
We included patients undergoing spinal surgery for chronic low back pain associated with degenerative changes of the lumbar spine. Patients with planned surgery for lumbosacral radiculopathy due to herniated discs, surgery for cancer or trauma were ineligible because clinical presentation, surgical approach and prognosis after surgery typically differ in these patients compared with those with low back pain associated with degenerative changes.11 Chronic low back pain was defined as lumbar back pain of at least three on a numerical rating scale (NRS) with 0 ‘no pain’ and 10 ‘worst pain imaginable’ on most days during the week and with a minimum duration of 3 months, with or without radiation to the leg. We excluded patients with bilateral pain below the knees because of possible interference with QST, patients with rheumatologic inflammatory diseases, neurological comorbidities potentially affecting the neurological function of the lower extremity to be tested, psychiatric comorbidities other than unipolar depressive disorder, previous instrumented spinal surgery (previous total disc replacement or spinal fusion with pedicle screws, cages or internal splints), planned surgery of more than three segments, and those with multiple somatic comorbidities. We also excluded patients who could not be contacted by phone or mail before surgery. We chose a pragmatic approach with broad eligibility criteria to best reflect the clinical setting and thus conclusions based on this study cohort are more likely to inform clinical decision-making.25
All patients underwent a detailed orthopaedic and neurological assessment and spinal MRI. Two assessors performed a repeat clinical examination at study entry to confirm study eligibility and other study-related procedures according to a previously applied, standardised, prospective protocol.17 We performed all study-related procedures at the Department of Anaesthesiology and Pain Medicine of the University Hospital of Bern.
Quantitative sensory tests
We performed QST, according to a previously applied prospective protocol17, in a quiet room to avoid distraction. For testing, patients lay in bed with a leg rest placed under the knees to obtain 30° semiflexion. All patients received identical instruction regarding the testing session and underwent a training session so that they were familiar with the procedure before data collection was initiated. This is common practice for testing protocols.17,26 We performed tests at the most painful area of the lower back (regional hypersensitivity) and the contralateral extremity (widespread hypersensitivity). In case of bilateral back pain, the testing extremity was selected randomly according to a computer-generated list. Except for the cold pressor test and the assessment of conditioned pain modulation (CPM), for which only one measurement was taken, we made two measurements and considered the mean value for data analysis. We randomly assigned the sequence of testing modalities according to a computer-generated list to avoid bias as a result of testing order.27 Supplemental Digital Text 1, http://links.lww.com/EJA/A202 includes details of the QST assessment.
Sociodemographic predictors included age, sex, education (higher/lower), working status (regular work including houseworkers/no regular work) and civil status (married/not married). We considered patients with high school education or a university degree as having higher education. We administered the Beck Depression Inventory version 2,28 the State-Trait-Anxiety-Inventory29 and the Pain Catastrophising Scale.30 The clinical assessment included BMI, smoking (yes/no), finger ground distance (>10 cm/≤10 cm), positive Lasègue sign (yes/no), previous noninstrumented back surgery (yes/no), pain radiating to the leg (yes/no), pain duration (>5 years/≤ 5 years), pain intensity, disability, intake of nonopioid and opioid analgesics (yes/no). For pain intensity, we used the maximum NRS during the 7 days preceeding the baseline assessment, and the Oswestry Disability Index for disability.31 For pain intensity, we chose these sociodemographic, psychological and clinical characteristics because of their potential prognostic value for FBSS according to previous studies.12,13
All MRI were examined by an independent radiologist at Balgrist University Hospital in Zurich, Switzerland, blinded to type of surgery, spinal levels operated and surgical outcome. He evaluated degenerative changes of the spine by characterising spinal stenosis, spondylolisthesis, endplate changes, scoliosis, facet joint degeneration, disc degeneration and fatty degeneration of paraspinal muscles. He first rated each segment between L1 and S1 according to these features. We then identified the worst segment for each feature in each patient and considered degenerative changes of the worst segment as a covariate for further statistical analysis. Supplemental Digital Text 2, http://links.lww.com/EJA/A202 describes in detail the classification criteria.
Surgery and postsurgical care
All procedures were performed by senior surgeons under standard general anaesthesia (combined volatile anaesthetics and intravenous opioids). The surgeons based their choice of spinal procedure and the number of segments on which to operate according to clinical reasoning and radiological findings.6,32 Postsurgical treatment was standardised for all patients and included intravenous patient-controlled analgesia (morphine as routine, fentanyl in case of renal insufficiency or intolerance to morphine), prescription of nonopioid analgesics postdischarge and stepwise rehabilitation. Rehabilitation consisted of stabilising muscle exercises in the supine position for trunk muscles. Patients were encouraged to walk as much as could be tolerated. Rehabilitation training began 2 months after surgery. Evidence linking the type of surgery (instrumented/noninstrumented surgery) and number of segments to be operated on (multisegmental/unisegmental)33,34 to FBSS is inconclusive. We therefore adjusted our analyses for these procedural characteristics. We recorded maximum pain intensity at the first day after surgery and on the last in-hospital day using the NRS, and considered the average value for analysis. Acute postsurgical pain is an important predictor for FBSS35 and also an intermediate outcome potentially lying on the causal pathway. We did not include it in the main model but performed a sensitivity analysis adjusting for this variable.
Definition of failed back surgery syndrome
We performed follow-up assessments 6 and 12 months after surgery to determine the presence of FBSS. We defined FBSS as less than 30% reduction in maximum pain intensity during the last 7 days compared with baseline and less than 30% reduction in disability from baseline.20–22 In addition, we used a composite endpoint and defined FBSS as persistence of either pain or disability to integrate both outcomes as recommended.20–22 The primary outcome was persistence of pain at 12 months. We considered all other outcomes as secondary outcomes.
We expected a frequency of FBSS of 30%.3–5 We estimated that a sample size of 155 patients would enable the detection of a dichotomised predictor that is approximately twice as frequent in patients with FBSS compared with patients without FBSS, if the frequency of the predictor was at least 25%. For continuous predictors, this sample size would detect a difference between patients with and without FBSS of 0.5 SDs. We considered a power of 80% and a two-sided alpha of 0.05. A sample size of 155 patients allowed the inclusion of approximately nine variables in multivariable models.36
To determine presurgical predictors of FBSS we used logistic regressions based on multiple imputations37 to calculate the odds ratios (ORs) for sociodemographic, psychological, clinical and radiological characteristics, as well as QST. As specified, we dichotomised education, working conditions, civil status, pain radiating to the leg and all radiological variables to facilitate clinical interpretation. We dichotomised finger-ground distance and pain duration at baseline because these variables were neither normally nor log-normally distributed. QST data with electrical or pressure stimulation were normally or log-normally distributed. Heat and cold pain detection thresholds and hand withdrawal time of the cold pressor test were truncated and were neither normally nor log-normally distributed. Therefore, we dichotomised these variables using the maximally attainable stimulus as a cut-off. To ensure comparability of regression coefficients for continuous and binary covariates, we expressed the effect for all continuous variables per 2 SD change on the normal or logarithmic scale.38 For continuous sociodemographic, psychological and clinical variables, the effect was expressed per 2 SD increase. For continuous QST, it was expressed per 2 SD decrease. ORs above one imply that pathological values of QST (i.e. lower thresholds after pressure, electrical and heat stimulation, higher thresholds after cold stimulation, shorter hand withdrawal time of the cold pressor test and impaired CPM) are associated with an increased risk of FBSS.
We imputed predictors and outcomes using chained equations with predictive mean matching for continuous variables and logistic regression for binary variables generating 15 multiply imputed datasets. We performed stepwise imputation, first imputing psychological, clinical, radiological and procedural characteristics, then values of QST, and finally all outcome variables.
In our main analysis, we calculated crude, partially and fully adjusted ORs of all predictors for persistence of pain at 12 months. We included the type of surgery and the number of segments operated on as procedural characteristics in partially adjusted analyses. We estimated fully adjusted ORs including sociodemographic, psychological, clinical and radiological characteristics that were associated with the primary outcome at P value 0.10 or less in partially adjusted analyses and forced procedural characteristics and sex into the models. We performed four sets of fully adjusted sensitivity analyses of the association between QST and the primary outcome. First, we included acute postsurgical pain for the reason mentioned above. Second, we restricted the analysis to patients with complete follow-up at both time points, using multiple imputation only for missing covariate data. Third, we performed a linear regression with pain intensity at 12 months as continuous outcome variable. Fourth, we conducted a subgroup analysis including only patients with no previous surgery to rule out the possibility that FBSS at baseline would influence the results. We conducted several secondary analyses exploring the fully adjusted associations of QST and FBSS according to different outcome definitions and follow-up time points, including the same set of covariates as in the main analysis. P values are two-sided and confidence intervals (CIs) refer to 95% boundaries. We performed all analyses with Stata (Version 12.1; StataCorp, College Station, Texas, USA).
We screened 958 patients with chronic low back pain undergoing spinal surgery between 2012 and 2015, found 392 eligible patients (41%), and tested 141 (Fig. 1). Time and resource constraints led us to close the study 14 patients (9%) short of the planned 155. Ninety-six (68.1%) patients had surgery that involved a single, 34 (24.1%) two and 11 (7.8%) three segments. Twenty-eight (19.9%) patients had previous noninstrumented back surgery. In 49 (34.8%), decompression without additional instrumental stabilisation was performed. We did not encounter any surgical complications. Treatment of patients developing FBSS was left to the discretion of the treating surgeon and was not monitored by the study team. Eleven (7.8%) patients underwent re-operation during the follow-up period as a result of incident FBSS. A total of 140 patients (99%) and 137 patients (95%) presented for the 6 and 12 months’ follow-up, respectively.
Failed back surgery syndrome according to definition of outcome and follow-up time point and completeness of data
Table 1 shows the frequency of FBSS depending on outcome definition and time of assessment. Forty-four patients (31.2%) developed FBSS, defined as persistent pain 12 months after imputing missing data. The frequency of FBSS, as defined by persistence of pain or by persistence of disability, was around 30% at both follow-ups. Some 40% of all patients presented with FBSS, if defined according to the composite endpoint. The frequency of FBSS remained robust per outcome if we based the calculation on data as observed, on patients with complete follow-up data or on data after multiple imputation. Supplemental Digital Table 1, http://links.lww.com/EJA/A202 shows details of data completeness and distribution of all predictors. Data on heat and cold pain detection thresholds, hand withdrawal time of the cold pressor test and CPM were missing due to logistical reasons and were thus considered as randomly missing. Data completeness of these QST ranged from 82 to 97%. We were able to evoke a nociceptive withdrawal reflex (NWR) in only 67% of patients, as painful stimulation became intolerable before a NWR could be elicited. We had already encountered this issue in a previous study.39 We cannot rule out that the inability to evoke a NWR is the result of normal central pain processing. Therefore, the assumption of data missing at random for multiple imputation was likely to be violated and we refrained from analysing NWR threshold.
Sociodemographic, psychological, clinical and radiological predictors of failed back surgery syndrome
Table 2 presents crude and partially adjusted associations of sociodemographic, psychological, clinical and radiological predictors for FBSS, defined as persistence of pain after 12 months. Sociodemographic and radiological predictors were similarly distributed in both patient groups. We found equal scores depression and anxiety in patients with and without FBSS. However, higher scores for catastrophising, increased BMI, greater finger-ground distance, positive Lasègue sign, higher baseline disability, intake of nonopioid analgesics and intake of opioid analgesics were associated with an increased risk of FBSS, with ORs ranging from 1.85 to 2.44 and P values from 0.02 to 0.10 in partially adjusted models. We included these predictors along with procedural characteristics and sex in all fully adjusted models. Procedural characteristics were not associated with FBSS. Instrumented surgery as compared with simple decompression showed an OR (95% CI) of 0.80 (0.40 to 1.64, P = 0.54) and multisegmental as compared with unisegmental surgery showed an OR of 0.94 (0.43 to 2.04, P = 0.87). After full adjustment, we did not find any significant association for sex (OR 0.59; 0.26 to 1.34, P = 0.21), catastrophising (OR 1.58; 0.66 to 3.79, P = 0.31), BMI (OR 1.73; 0.77 to 3.93, P = 0.19), finger-ground distance (OR 1.42; 0.55 to 3.66, P = 0.47), Lasègue sign (OR 1.85; 0.75 to 4.56, P = 0.18), baseline disability (OR 1.13; 0.46 to 2.77, P = 0.78), intake of nonopioid analgesics (OR 1.95; 0.83 to 4.54, P = 0.12) and intake of opioid analgesics (OR 1.35; 0.46 to 3.94, P = 0.58).
Quantitative sensory tests as predictors of failed back surgery syndrome
Table 3 and Fig. 2 show crude and partially adjusted associations as well as fully adjusted associations of QST with FBSS, defined as persistence of pain 12 months after surgery. All point estimates appeared randomly scattered around one and all 95% CI included one as a measure of no association. We found none of the QST to be associated with the primary outcome in any fully adjusted sensitivity analyses (Supplemental Digital Table 2, http://links.lww.com/EJA/A202). OR and corresponding 95% CI of all QST remained similar to the main analysis after including acute postsurgical pain (sensitivity analysis a) and after restricting the analysis to patients with complete follow-up at both time-points (sensitivity analysis b). Except for cold pain detection threshold at the leg, none of the P values showed a statistical trend or a significant association between QST and pain intensity after 12 months as continuous outcome (sensitivity analysis c). Table 4 shows fully adjusted secondary analyses of the associations between QST and FBSS, defined as persistence of disability and persistence of pain or disability after 12 months. Supplemental Digital Table 3, http://links.lww.com/EJA/A202 shows association of QST and FBSS after 6 months (secondary outcomes). Results were similar to the main analyses, with no statistically significant associations and point estimates randomly scattered around one.
In this prospective cohort study of 141 patients with chronic low back pain undergoing up to three level spinal surgery, none of the investigated 14 QST at baseline showed a statistically significant association with FBSS at 12 months. The negative conclusion of our study remained robust to three sets of sensitivity analyses and five sets of secondary analyses with only one statistically significant association across clinical outcome definitions, time points and analytical strategies. Therefore, the potential association of cold pain detection threshold at the leg analysed as continuous outcome with persistence of pain after 12 months (sensitivity analysis) is probably a chance finding in view of a total of 126 statistical tests performed.
We are aware of a single study linking QST with pain and disability after spinal surgery in 38 patients with lumbar disc herniation.40 QST values showed low or no correlation with pain and disability. The study was limited by its small sample size, no standard definition of FBSS, and a lack of a comprehensive set of sociodemographic, psychological, clinical and surgery-related predictors. Two previous cohorts included patients with other musculoskeletal pain syndromes.41,42 Petersen et al.41 examined the prognostic value of 7 QST for persistent pain after total knee replacement. Only pressure pain detection threshold at the leg was associated with pain at 12 months.41 Wylde et al.42 used only pressure pain detection threshold at the forearm in over 400 patients with total knee or hip replacement, and found a statistically significant association with persistent postsurgical pain after 12 months.
Strengths and limitations
To our knowledge, our study is the first to prospectively assess a comprehensive number of QST, testing different pain modalities as presurgical predictors for FBSS in a large sample. We did not limit our outcome assessment to pain intensity, but also included disease-specific disability and a composite endpoint of pain and disability. The higher incidence of the composite outcome defined as persistence of pain or disability after surgery compared with the single component outcomes reflects the lack of consistent association between pain and disability. Still, the definition of FBSS followed established concepts20–22 and the incidence of FBSS of 30% for the primary outcome was in agreement with previous studies.1,3–6 Other strengths include the long follow-up period with near complete follow-up at 6 and 12 months. A limitation was the difficulty in recruiting, partly due to the large number of ineligible patients and to 46% of eligible patients who refused to participate. This might have compromised the generalisability of our results. Time and resource constraints led us to close the study 14 patients (9%) short of the planned number of 155. This decreased statistical precision slightly. However, it is extremely unlikely that the negative conclusions are merely due to limited statistical precision because the results remained robust to several secondary and sensitivity analysis including exploring associations between QST the primary outcome as a continuous variable. A major strength of our study was the application of an extensive, multimodal QST protocol using 14 tests to assess different dimensions of nociception and pain processing. Due to the consistent evidence for central hypersensitivity in chronic low back pain, we did not include a control group of pain-free patients.17–19
We also included type of surgery and number of operated segments as covariates in all multivariable analyses, in view of the potential variation of clinical outcome associated with these factors.33,34 We did not see any relevant differences in effect estimates of the associations between QST and FBSS before and after adjusting for procedural characteristics. To account for missing data, we used multiple imputations.37
Implication for further research
Our negative findings do not necessarily mean that central hypersensitivity is not involved in FBSS. First, the current study assessed the association of QST with persistence of pain at distinct time-points after surgery, rather than with the clinical course over time. Different, as yet unknown, phenotypes of patients may experience distinct patterns of pain or disability over time. These time-depending patterns are commonly referred to as trajectories.43 Future research should investigate if patients belonging to different trajectories have different prognoses after surgery. Second, current QST may be limited in its detection of clinically relevant central pain processes. Future research should aim at identifying biomarkers of central hypersensitivity that are better linked to patient-relevant outcomes.
The study indicates that assessment of altered central pain processing using current QST is unlikely to identify patients at risk for FBSS and therefore unlikely to inform clinical decisions.
Acknowledgements relating to this article
Assistance with the study: we would like to thank Katrin Ziegler, MSc, Clinical Trials Unit, University of Bern, Switzerland, for her support in data-management.
Financial support and sponsorship: this work was supported by the Swiss National Science Foundation (grant no. 32003B_138361) and the Scientific Funds of the University Department of Anaesthesiology and Pain Therapy of the University of Bern.
Conflicts of interest: none.
1. Chan CW, Peng P. Failed back surgery syndrome. Pain Med
2. Manca A, Eldabe S, Buchser E, et al. Relationship between health-related quality of life, pain, and functional disability in neuropathic pain patients with failed back surgery syndrome. Value Health
3. Fritzell P, Hagg O, Wessberg P, et al. 2001 Volvo award winner in clinical studies: lumbar fusion versus nonsurgical treatment for chronic low back pain: a multicenter randomized controlled trial from the Swedish Lumbar Spine Study Group. Spine
2001; 26:2521–2532. discussion 2532–2534.
4. Brox JI, Sorensen R, Friis A, et al. Randomized clinical trial of lumbar instrumented fusion and cognitive intervention and exercises in patients with chronic low back pain and disc degeneration. Spine
5. Dimar JR 2nd, Glassman SD, Burkus JK, et al. Two-year fusion and clinical outcomes in 224 patients treated with a single-level instrumented posterolateral fusion with iliac crest bone graft. Spine J
6. Zaina F, Tomkins-Lane C, Carragee E, et al. Surgical versus nonsurgical treatment for lumbar spinal stenosis. Cochrane Database Syst Rev
7. Hoy D, March L, Brooks P, et al. The global burden of low back pain: estimates from the Global Burden of Disease 2010 study. Ann Rheum Dis
8. Freburger JK, Holmes GM, Agans RP, et al. The rising prevalence of chronic low back pain. Arch Intern Med
9. Deyo RA, Mirza SK, Martin BI, et al. Trends, major medical complications, and charges associated with surgery for lumbar spinal stenosis in older adults. JAMA
10. Deyo RA, Nachemson A, Mirza SK. Spinal-fusion surgery – the case for restraint. N Engl J Med
11. Deyo RA. Back surgery – who needs it? N Engl J Med
12. Celestin J, Edwards RR, Jamison RN. Pretreatment psychosocial variables as predictors of outcomes following lumbar surgery and spinal cord stimulation: a systematic review and literature synthesis. Pain Med
13. Mannion AF, Elfering A. Predictors of surgical outcome and their assessment. Eur Spine J
2006; 15 (Suppl 1):S93–S108.
14. Woolf CJ. Central sensitization: implications for the diagnosis and treatment of pain. Pain
2011; 152 (3 Suppl):S2–S15.
15. Curatolo M. Diagnosis of altered central pain processing. Spine
2011; 36 (25 Suppl):S200–S204.
16. Birklein F, Sommer C. Pain: ouantitative sensory testing – a tool for daily practice? Nat Rev Neurol
17. Neziri AY, Curatolo M, Limacher A, et al. Ranking of parameters of pain hypersensitivity according to their discriminative ability in chronic low back pain. Pain
18. Giesecke T, Gracely RH, Grant MA, et al. Evidence of augmented central pain processing in idiopathic chronic low back pain. Arthritis Rheum
19. Blumenstiel K, Gerhardt A, Rolke R, et al. Quantitative sensory testing profiles in chronic back pain are distinct from those in fibromyalgia. Clin J Pain
20. Farrar JT. Advances in clinical research methodology for pain clinical trials. Nat Med
21. Dworkin RH, Turk DC, Peirce-Sandner S, et al. Research design considerations for confirmatory chronic pain clinical trials: IMMPACT recommendations. Pain
22. Ballantyne JC, Sullivan MD. Intensity of chronic pain – the wrong metric? N Engl J Med
23. World Medical Association. WMA Declaration of Helsinki – ethical principles for medical research involving human subjects. Seoul: World Medical Association; 2008.
24. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet
25. Ford I, Norrie J. Pragmatic trials. N Engl J Med
26. Backonja MM, Attal N, Baron R, et al. Value of quantitative sensory testing in neurological and pain disorders: NeuPSIG consensus. Pain
27. Grone E, Crispin A, Fleckenstein J, et al. Test order of quantitative sensory testing facilitates mechanical hyperalgesia in healthy volunteers. J Pain
28. Morley S, de CWA, Black S. A confirmatory factor analysis of the Beck Depression Inventory in chronic pain. Pain
29. Laux L, Glanzmann P, Schaffner P, et al. State-Trait-Angstinventar (STAI). Göttingen: Hogrefe Verlag; 1981.
30. Sullivan MJL, Bishop SR, Pivik J. The Pain Catastrophizing Scale: development and validation. Psychol Assess
31. Fairbank JC, Pynsent PB. The Oswestry Disability Index. Spine
2000; 25:2940–2952. discussion 2952.
32. Resnick DK, Watters WC 3rd, Mummaneni PV, et al. Guideline update for the performance of fusion procedures for degenerative disease of the lumbar spine. Part 10: Lumbar fusion for stenosis without spondylolisthesis. J Neurosurg Spine
33. Park DK, An HS, Lurie JD, et al. Does multilevel lumbar stenosis lead to poorer outcomes?: a subanalysis of the Spine Patient Outcomes Research Trial (SPORT) lumbar stenosis study. Spine
34. Smorgick Y, Park DK, Baker KC, et al. Single- versus multilevel fusion for single-level degenerative spondylolisthesis and multilevel lumbar stenosis: four-year results of the spine patient outcomes research trial. Spine
35. VanDenKerkhof EG, Hopman WM, Goldstein DH, et al. Impact of perioperative pain intensity, pain qualities, and opioid use on chronic pain after surgery: a prospective cohort study. Reg Anesth Pain Med
36. Vittinghoff E, McCulloch CE. Relaxing the rule of ten events per variable in logistic and Cox regression. Am J Epidemiol
37. Spratt M, Carpenter J, Sterne JA, et al. Strategies for multiple imputation in longitudinal studies. Am J Epidemiol
38. Gelman A. Scaling regression inputs by dividing by two standard deviations. Stat Med
39. Curatolo M, Muller M, Ashraf A, et al. Pain hypersensitivity and spinal nociceptive hypersensitivity in chronic pain: prevalence and associated factors. Pain
40. Tschugg A, Loscher WN, Lener S, et al. The value of quantitative sensory testing in spine research. Neurosurg Rev
41. Petersen KK, Graven-Nielsen T, Simonsen O, et al. Preoperative pain mechanisms assessed by cuff algometry are associated with chronic postoperative pain relief after total knee replacement. Pain
42. Wylde V, Sayers A, Lenguerrand E, et al. Preoperative widespread pain sensitization and chronic pain after hip and knee replacement: a cohort analysis. Pain
43. Leffondre K, Abrahamowicz M, Regeasse A, et al. Statistical measures were proposed for identifying longitudinal patterns of change in quantitative health indicators. J Clin Epidemiol