Machine learning versus logistic regression for prognostic modelling in individuals with non-specific neck pain.

PURPOSE: Prognostic models play an important clinical role in the clinical management of neck pain disorders. No study has compared the performance of modern machine learning (ML) techniques, against more traditional regression techniques, when developing prognostic models in individuals with neck pain. METHODS: A total of 3001 participants suffering from neck pain were included into a clinical registry database. Three dichotomous outcomes of a clinically meaningful improvement in neck pain, arm pain, and disability at 3 months follow-up were used. There were 26 predictors included, five numeric and 21 categorical. Seven modelling techniques were used (logistic regression, least absolute shrinkage and selection operator [LASSO], gradient boosting [Xgboost], K nearest neighbours [KNN], support vector machine [SVM], random forest [RF], and artificial neural networks [ANN]). The primary measure of model performance was the area under the receiver operator curve (AUC) of the validation set. RESULTS: The ML algorithm with the greatest AUC for predicting arm pain (AUC = 0.765), neck pain (AUC = 0.726), and disability (AUC = 0.703) was Xgboost. The improvement in classification AUC from stepwise logistic regression to the best performing machine learning algorithms was 0.081, 0.103, and 0.077 for predicting arm pain, neck pain, and disability, respectively. CONCLUSION: The improvement in prediction performance between ML and logistic regression methods in the present study, could be due to the potential greater nonlinearity between baseline predictors and clinical outcome. The benefit of machine learning in prognostic modelling may be dependent on factors like sample size, variable type, and disease investigated.

Metastatic Versus Osteoporotic Vertebral Fractures on MRI: A Blinded, Multicenter, and Multispecialty Observer Agreement Evaluation.

BACKGROUND: MRI is assumed to be valid for distinguishing metastatic vertebral fractures (MVFs) from osteoporotic vertebral fractures (OVFs). This study assessed (1) concordance between the image-based diagnosis of MVF versus OVF and the reference (biopsy or follow-up of >6 months), (2) interobserver and intraobserver agreement on key imaging findings and the diagnosis of MVF versus OVF, and (3) whether disclosing a patient's history of cancer leads to variations in diagnosis, concordance, or agreement. PATIENTS AND METHODS: This retrospective cohort study included clinical data and imaging from 203 patients with confirmed MVF or OVF provided to 25 clinicians (neurosurgeons, radiologists, orthopedic surgeons, and radiation oncologists). From January 2018 through October 2018, the clinicians interpreted images in conditions as close as possible to routine practice. Each specialist assessed data twice, with a minimum 6-week interval, blinded to assessments made by other clinicians and to their own previous assessments. The kappa statistic was used to assess interobserver and intraobserver agreement on key imaging findings, diagnosis (MVF vs OVF), and concordance with the reference. Subgroup analyses were based on clinicians' specialty, years of experience, and complexity of the hospital where they worked. RESULTS: For diagnosis of MVF versus OVF, interobserver agreement was fair, whereas intraobserver agreement was substantial. Only the latter improved to almost perfect when a patient's history of cancer was disclosed. Interobserver agreement for key imaging findings was fair or moderate, whereas intraobserver agreement on key imaging findings was moderate or substantial. Concordance between the diagnosis of MVF versus OVF and the reference was moderate. Results were similar regardless of clinicians' specialty, experience, and hospital category. CONCLUSIONS: When MRI is used to distinguish MVF versus OVF, interobserver agreement and concordance with the reference were moderate. These results cast doubt on the reliability of basing such a diagnosis on MRI in routine practice.

Exercise treatment effect modifiers in persistent low back pain: an individual participant data meta-analysis of 3514 participants from 27 randomised controlled trials.

BACKGROUND: Low back pain is one of the leading causes of disability worldwide. Exercise therapy is widely recommended to treat persistent non-specific low back pain. While evidence suggests exercise is, on average, moderately effective, there remains uncertainty about which individuals might benefit the most from exercise. METHODS: In parallel with a Cochrane review update, we requested individual participant data (IPD) from high-quality randomised clinical trials of adults with our two primary outcomes of interest, pain and functional limitations, and calculated global recovery. We compiled a master data set including baseline participant characteristics, exercise and comparison characteristics, and outcomes at short-term, moderate-term and long-term follow-up. We conducted descriptive analyses and one-stage IPD meta-analysis using multilevel mixed-effects regression of the overall treatment effect and prespecified potential treatment effect modifiers. RESULTS: We received IPD for 27 trials (3514 participants). For studies included in this analysis, compared with no treatment/usual care, exercise therapy on average reduced pain (mean effect/100 (95% CI) -10.7 (-14.1 to -7.4)), a result compatible with a clinically important 20% smallest worthwhile effect. Exercise therapy reduced functional limitations with a clinically important 23% improvement (mean effect/100 (95% CI) -10.2 (-13.2 to -7.3)) at short-term follow-up. Not having heavy physical demands at work and medication use for low back pain were potential treatment effect modifiers-these were associated with superior exercise outcomes relative to non-exercise comparisons. Lower body mass index was also associated with better outcomes in exercise compared with no treatment/usual care. This study was limited by inconsistent availability and measurement of participant characteristics. CONCLUSIONS: This study provides potentially useful information to help treat patients and design future studies of exercise interventions that are better matched to specific subgroups. PROTOCOL PUBLICATION: https://doi.org/10.1186/2046-4053-1-64.

Efficacy, Effectiveness, Safety, and Cost-effectiveness of Epidural Adhesiolysis for Treating Failed Back Surgery Syndrome. A Systematic Review.

BACKGROUND: Failed back surgery syndrome (FBSS) has a profound impact on patients' quality of life and represents a major clinical challenge and a significant economic burden for society. Adhesiolysis is used as a treatment to eliminate perineural/epidural adhesions in patients with chronic pain attributed to FBSS. OBJECTIVE: To evaluate the efficacy, effectiveness, safety, and cost-effectiveness of epidural adhesiolysis compared with other procedures for treating FBSS. METHOD: A systematic review was conducted. The electronic databases Medline/PreMedline, EMBASE, Cochrane Library Plus, Centre for Reviews and Dissemination databases, SCOPUS, Science Citation Index, and PEDRO were consulted through April 2017. Predefined criteria were used to determine inclusion of the studies and to assess their methodological quality. RESULTS: Ten reports were included. No randomized controlled trials (RCTs) on efficacy or cost-effectiveness were found. Three reports (corresponding to two RCTs, N = 212) suggested that adhesiolysis was effective, especially for pain and disability. However, both studies presented serious methodological flaws. In addition to RCTs, seven observational studies with high risk of bias reported data on effectiveness and safety. Fifty-eight adverse events were reported among 130 patients undergoing endoscopic adhesiolysis, and 19 among the 110 undergoing percutaneous adhesiolysis. CONCLUSIONS: The evidence on the efficacy and cost-effectiveness of adhesiolysis for treating FBSS is nonexistent, whereas evidence on its effectiveness and safety is insufficient. Incorporating data from observational studies did not improve the quality of the evidence on effectiveness.

Predicting the evolution of neck pain episodes in routine clinical practice.

BACKGROUND: The objective of this study was to develop models for predicting the evolution of a neck pain (NP) episode. METHODS: Three thousand two hundred twenty-five acute and chronic patients seeking care for NP, were recruited consecutively in 47 health care centers. Data on 37 variables were gathered, including gender, age, employment status, duration of pain, intensity of NP and pain referred down to the arm (AP), disability, history of neck surgery, diagnostic procedures undertaken, imaging findings, clinical diagnosis, and treatments used. Three separate multivariable logistic regression models were developed for predicting a clinically relevant improvement in NP, AP and disability at 3 months. RESULTS: Three thousand one (93.5%%) patients attended follow-up. For all the models calibration was good. The area under the ROC curve was ≥0.717 for pain and 0.664 for disability. Factors associated with a better prognosis were: a) For all the outcomes: pain being acute (vs. chronic) and having received neuro-reflexotherapy. b) For NP: nonspecific pain (vs. pain caused by disc herniation or spinal stenosis), no signs of disc degeneration on imaging, staying at work, and being female. c) For AP: nonspecific NP and no signs of disc degeneration on imaging. d) For disability: staying at work and no signs of facet joint degeneration on imaging. CONCLUSIONS: A prospective registry can be used for developing valid predictive models to quantify the odds that a given patient with NP will experience a clinically relevant improvement.

Prevalence and factors associated with a higher risk of neck and back pain among permanent wheelchair users: a cross-sectional study.

STUDY DESIGN: Cross-sectional study. OBJECTIVES: To determine the prevalence of, and factors associated with, spinal pain among wheelchair users. SETTING: Four Spanish hospitals specialized in providing care for wheelchair users. METHODS: Persons who had used a wheelchair for a median (IRQ) of 10 (5;19) years, 27% of them due to reasons other than spinal cord injury, were recruited consecutively (n = 750). Data on 43 demographic, psychosocial, ergonomic, and clinical variables were collected, and analyzed. Main outcome measures were: point prevalence of neck (NP), thoracic (TP), low back pain (LBP), and pain at any spinal level (PASL); and factors associated with them. RESULTS: Point prevalence was 56% for NP, 54% for TP, 45% for LBP, and 76% for PSAL. PASL was associated with a lower quality of life (OR (95% CI) 0.91 (0.86; 0.97)). Multivariable regression models showed that the main factors associated with significant pain (≥1.5 VAS points) were: (a) For NP: cervical spinal injury and wheelchair seat cushion thickness, (b) For TP: thoracic spinal injury and sagittal index, (c) For LBP: thoracic or lumbar spinal injury, with some sensitivity remaining, (d) For PASL: being female, living alone, and using a non-power wheelchair. Discrimination (AUC) of these models ranged between 0.638 and 0.818. p-values in the Hosmer-Lemeshow test ranged between 0.420 and 0.701. CONCLUSIONS: Prevalence of spinal pain among wheelchair users is high. It is associated with a lower quality of life. Future studies should assess whether using a power wheelchair affects PASL, and if the thickness of seat cushion affects NP. SPONSORSHIP: Spanish Back Pain Research Network.

Agreement in Metastatic Spinal Cord Compression.

BACKGROUND: Metastatic epidural spinal cord compression (ESCC) is a devastating medical emergency. The purpose of this study was to determine the reliability of the 6-point ESCC scoring system and the identification of the spinal level presenting ESCC. METHODS: Clinical data and imaging from 90 patients with biopsy-proven spinal metastases were provided to 83 specialists from 44 hospitals. The spinal levels presenting metastases and the ESCC scores for each case were calculated twice by each clinician, with a minimum of 6 weeks' interval. Clinicians were blinded to assessments made by other specialists and their own previous assessment. Fleiss kappa (κ) statistic was used to assess intraobserver and interobserver agreement. Subgroup analyses were performed according to clinicians' specialty (medical oncology, neurosurgery, radiology, orthopedic surgery, and radiation oncology), years of experience, and type of hospital. RESULTS: Intraobserver and interobserver agreement on the location of ESCC was substantial (κ>0.61). Intraobserver agreement on the ESCC score was "excellent" (κ=0.82), whereas interobserver agreement was substantial (κ=0.64). Overall agreement with the tumor board classification was substantial (κ=0.71). Results were similar across specialties, years of experience and hospital category. CONCLUSIONS: The ESCC score can help improve communication among clinicians involved in oncology care.

Core outcome domains for clinical trials in non-specific low back pain.

PURPOSE: Inconsistent reporting of outcomes in clinical trials of patients with non-specific low back pain (NSLBP) hinders comparison of findings and the reliability of systematic reviews. A core outcome set (COS) can address this issue as it defines a minimum set of outcomes that should be reported in all clinical trials. In 1998, Deyo et al. recommended a standardized set of outcomes for LBP clinical research. The aim of this study was to update these recommendations by determining which outcome domains should be included in a COS for clinical trials in NSLBP. METHODS: An International Steering Committee established the methodology to develop this COS. The OMERACT Filter 2.0 framework was used to draw a list of potential core domains that were presented in a Delphi study. Researchers, care providers and patients were invited to participate in three Delphi rounds and were asked to judge which domains were core. A priori criteria for consensus were established before each round and were analysed together with arguments provided by panellists on importance, overlap, aggregation and/or addition of potential core domains. The Steering Committee discussed the final results and made final decisions. RESULTS: A set of 280 experts was invited to participate in the Delphi; response rates in the three rounds were 52, 50 and 45%. Of 41 potential core domains presented in the first round, 13 had sufficient support to be presented for rating in the third round. Overall consensus was reached for the inclusion of three domains in this COS: 'physical functioning', 'pain intensity' and 'health-related quality of life'. Consensus on 'physical functioning' and 'pain intensity' was consistent across all stakeholders, 'health-related quality of life' was not supported by the patients, and all the other domains were not supported by two or more groups of stakeholders. Weighting all possible argumentations, the Steering Committee decided to include in the COS the three domains that reached overall consensus and the domain 'number of deaths'. CONCLUSIONS: The following outcome domains were included in this updated COS: 'physical functioning', 'pain intensity', 'health-related quality of life' and 'number of deaths'. The next step for the development of this COS will be to determine which measurement instruments best measure these domains.

Disc degeneration and chronic low back pain: an association which becomes nonsignificant when endplate changes and disc contour are taken into account.

INTRODUCTION: The objective of this study was to assess the association between severe disc degeneration (DD) and low back pain (LBP). METHODS: A case-control study was conducted with 304 subjects, aged 35-50, recruited in routine clinical practice across six hospitals; 240 cases (chronic LBP patients with a median pain duration of 46 months) and 64 controls (asymptomatic subjects without any lifetime history of significant LBP). The following variables were assessed once, using previously validated methods: gender, age, body mass index (BMI), lifetime smoking exposure, degree of physical activity, severity of LBP, disability, and findings on magnetic resonance (MRI) (disc degeneration, Modic changes (MC), disc protrusion/hernia, annular tears, spinal stenosis, and spondylolisthesis). Radiologists who interpreted MRI were blinded to the subjects' characteristics. A multivariate logistic regression model assessed the association between severe DD and chronic LBP, adjusting for gender, age, BMI, physical activity, MC, disc protrusion/hernia, and spinal stenosis. RESULTS: Severe DD at ≥1 level was found in 46.9 % of the controls and 65.8 % of the cases. Crude odds ratio (95 % CI), for suffering chronic LBP when having severe DD, was 2.06 (1.05; 4.06). After adjusting for "MC" and "disc protrusion/hernia," it was 1.81 (0.81; 4.05). CONCLUSIONS: The association between severe DD and LBP ceases to be significant when adjusted for MC and disc protrusion/hernia. These results do not support that DD as a major cause of chronic LBP.

Post-implementation surveillance of a non-pharmacological health technology within a national health service.

OBJECTIVES: The aim of this study was to describe 8-year results from post-implementation surveillance of neuroreflexotherapy (NRT), a health technology proven effective for treating neck and back pain. METHODS: Post-implementation surveillance included all patients undergoing NRT across five regions within the Spanish National Health Service (SNHS). Validated methods were used to assess pain, disability, adverse events, use of health resources, and patient satisfaction. Logistic regression models were developed to identify the variables associated with the risk of a pain episode requiring more than one NRT intervention. The number of relapses among discharged patients during the 8-year period was calculated. RESULTS: Between January 1, 2004, and June 30, 2012, 9,023 patients (median age: 53 years), presenting 11,384 subacute (25.2 percent) and chronic (74.8 percent), neck or back pain episodes, were discharged after receiving NRT. Spinal pain improved in 89 percent of cases, 83 percent abandoned drugs, and 0.02 percent required spine surgery. The only adverse event was skin discomfort (8.0 percent of patients). Number of patient complaints was 0, and answers to a standardized questionnaire reflected a high degree of satisfaction (response rate: 76.7 percent). Of the pain episodes, 18.9 percent required more than one NRT intervention; logistic regression models identified the variables associated with this. Over the 8-year period, the proportion of discharged patients referred for treatment due to relapse at the same level for neck, thoracic, and low back pain, was 16.4 percent, 6.5 percent, and 14.5 percent respectively. CONCLUSIONS: Post-marketing surveillance for a non-pharmacological technology is feasible within the SNHS. These results support generalizing NRT across the entire SNHS under the current validated application conditions.

The use of risk sharing tools for post adoption surveillance of a non pharmacological technology in routine practice: results after one year.

BACKGROUND: To report results obtained by combining risk sharing tools with post-adoption surveillance mechanisms in order to control quality of care and implement a value-based reimbursement scheme for Neuro-reflexotherapy (NRT), a non-pharmacological treatment proven effective for neck pain (NP), thoracic pain (TP) and low back pain (LBP). METHODS: Pre-post prospective cohort study in routine clinical practice, carried out in primary care centers in the Spanish National Health Service in the Balearic Islands (Ib-Salut). Eight-hundred and seventy-one subacute and chronic NP, TP and LBP patients treated in Ib-Salut, who underwent NRT during 2011. A shared risk contract (SRC) was developed, where payments for NRT were linked to results on patients' clinical evolution, reduction in medication and proportion of patients undergoing spinal surgery. Main outcome measures were local pain (NP, TP or LBP), referred pain, LBP-related disability and NP-related disability, measured using previously validated instruments at referral and 3 months later, use of medication assessed at referral and discharge, and rates of spinal surgery prescription after undergoing NRT. RESULTS: Median improvements at discharge corresponded to 57.1% of baseline value for local pain, 75.0% for referred pain, 53.8% for LBP-related disability and 45.0% for NP-related disability. Patients taking medication at discharge represented 29.0% of those taking it at referral. The proportion of patients in whom spinal surgery was prescribed after undergoing NRT was 0%. These results were consistent with those from previous randomised controlled trials (RCTs) and studies in routine practice, and complied with the standards set in the SRC. CONCLUSIONS: It is feasible and effective to enhance post adoption surveillance methods with risk sharing tools to improve quality control and support value-based reimbursement decisions for NRT. The feasibility of generalising this approach to other settings and to other non-pharmacological treatments should be explored.

The Management of Metastatic Spinal Cord Compression in Routine Clinical Practice.

(1) Background: Whether clinical management of spinal metastatic disease (SMD) matches evidence-based recommendations is largely unknown. (2) Patients and Methods: A questionnaire was distributed through Spanish Medical Societies, exploring routine practice, interpretation of the SINS and ESCC scores and agreement with items in the Tokuhashi and SINS scales, and NICE guideline recommendations. Questionnaires were completed voluntarily and anonymously, without compensation. (3) Results: Eighty specialists participated in the study. A protocol for patients with SMD existed in 33.7% of the hospitals, a specific multidisciplinary board in 33.7%, 40% of radiological reports included the ESCC score, and a prognostic scoring method was used in 73.7%. While 77.5% of the participants were familiar with SINS, only 60% used it. The different SINS and ESCC scores were interpreted correctly by 57.5-70.0% and 30.0-37.5% of the participants, respectively. Over 70% agreed with the items included in the SINS and Tokuhashi scores and with the recommendations from the NICE guideline. Differences were found across private/public sectors, hospital complexity, number of years of experience, number of patients with SMD seen annually and especially across specialties. (4) Conclusions: Most specialists know and agree with features defining the gold standard treatment for patients with SCC, but many do not apply them.

Automatic Variable Selection Algorithms in Prognostic Factor Research in Neck Pain.

This study aims to compare the variable selection strategies of different machine learning (ML) and statistical algorithms in the prognosis of neck pain (NP) recovery. A total of 3001 participants with NP were included. Three dichotomous outcomes of an improvement in NP, arm pain (AP), and disability at 3 months follow-up were used. Twenty-five variables (twenty-eight parameters) were included as predictors. There were more parameters than variables, as some categorical variables had >2 levels. Eight modelling techniques were compared: stepwise regression based on unadjusted p values (stepP), on adjusted p values (stepPAdj), on Akaike information criterion (stepAIC), best subset regression (BestSubset) least absolute shrinkage and selection operator [LASSO], Minimax concave penalty (MCP), model-based boosting (mboost), and multivariate adaptive regression splines (MuARS). The algorithm that selected the fewest predictors was stepPAdj (number of predictors, p = 4 to 8). MuARS was the algorithm with the second fewest predictors selected (p = 9 to 14). The predictor selected by all algorithms with the largest coefficient magnitude was "having undergone a neuroreflexotherapy intervention" for NP (ß = from 1.987 to 2.296) and AP (ß = from 2.639 to 3.554), and "Imaging findings: spinal stenosis" (ß = from -1.331 to -1.763) for disability. Stepwise regression based on adjusted p-values resulted in the sparsest models, which enhanced clinical interpretability. MuARS appears to provide the optimal balance between model sparsity whilst retaining high predictive performance across outcomes. Different algorithms produced similar performances but resulted in a different number of variables selected. Rather than relying on any single algorithm, confidence in the variable selection may be increased by using multiple algorithms.

The Efficacy and Effectiveness of Education for Preventing and Treating Non-Specific Low Back Pain in the Hispanic Cultural Setting: A Systematic Review.

A systematic review was conducted to assess the efficacy and effectiveness of education programs to prevent and treat low back pain (LBP) in the Hispanic cultural setting. Electronic and manual searches identified 1148 unique references. Nine randomized clinical trials (RCTs) were included in this review. Methodological quality assessment and data extraction followed the recommendations from the Cochrane Back Pain Review Group. Education programs which were assessed focused on active management (3 studies), postural hygiene (7), exercise (4) and pain neurophysiology (1). Comparators were no intervention, usual care, exercise, other types of education, and different combinations of these procedures. Five RCTs had a low risk of bias. Results show that: (a) education programs in the school setting can transmit potentially useful knowledge for LBP prevention and (b) education programs for patients with LBP improve the outcomes of usual care, especially in terms of disability. Education on pain neurophysiology improves the results of education on exercise, and education on active management is more effective than "sham" education and education on postural hygiene. Future studies should assess the comparative or summatory effects of education on exercise, education on pain neurophysiology and education on active management, as well as explore their efficiency.

Catastrophizing, state anxiety, anger, and depressive symptoms do not correlate with disability when variations of trait anxiety are taken into account. a study of chronic low back pain patients treated in Spanish pain units [NCT00360802].

OBJECTIVES: To assess the influence of pain severity, catastrophizing, anger, anxiety, and depression on nonspecific low back pain (LBP)-related disability in Spanish patients with chronic LBP. Study Design. Cross-sectional correlation between psychological variables and disability. Methods. One hundred twenty-three patients treated for chronic LBP in pain units within nine Spanish National Health Service Hospitals, in eight cities, were included in this study. Intensity of LBP and pain referred to the leg, disability, catastrophizing, anger, state anxiety, trait anxiety, and depression were assessed through previously validated questionnaires. The association of disability with these variables, as well as gender, age, academic level, work status, and use of antidepressants, was analyzed through linear regression models. RESULTS: Correlations between LBP, referred pain, disability, catastrophizing, anger, state anxiety, trait anxiety, and depression were significant, except for the ones between anger and LBP and between anger and referred pain. The multivariate regression model showed that when variations of trait anxiety were taken into account, the association of the other psychological variables with disability was no longer significant. The final model explained 49% of the variability of disability. Standardized coefficients were 0.452 for trait anxiety, 0.362 for intensity of LBP, 0.253 for failed back surgery, and -0.140 for higher academic level. CONCLUSION: Among Spanish chronic LBP patients treated at pain units, the correlation of catastrophizing, state anxiety, anger, and depression with disability ceases to be significant when variations of trait anxiety are taken into account. Further studies with LBP patients should determine whether anxiety trait mediates the effects of the other variables, explore its prognostic value, and assess the therapeutic effect of reducing it.