Quantitative sensory testing and predicting outcomes for musculoskeletal pain, disability, and negative affect: a systematic review and meta-analysis

Abstract

Hypersensitivity due to central pain mechanisms can influence recovery and lead to worse clinical outcomes, but the ability of quantitative sensory testing (QST), an index of sensitisation, to predict outcomes in chronic musculoskeletal disorders remains unclear. We systematically reviewed the evidence for ability of QST to predict pain, disability, and negative affect using searches of CENTRAL, MEDLINE, EMBASE, AMED, CINAHL, and PubMed databases up to April 2018. Title screening, data extraction, and methodological quality assessments were performed independently by 2 reviewers. Associations were reported between baseline QST and outcomes using adjusted (β) and unadjusted (r) correlations. Of the 37 eligible studies (n = 3860 participants), 32 were prospective cohort studies and 5 randomised controlled trials. Pain was an outcome in 30 studies, disability in 11, and negative affect in 3. Meta-analysis revealed that baseline QST predicted musculoskeletal pain (mean r = 0.31, 95% confidence interval [CI]: 0.23-0.38, n = 1057 participants) and disability (mean r = 0.30, 95% CI: 0.19-0.40, n = 290 participants). Baseline modalities quantifying central mechanisms such as temporal summation and conditioned pain modulation were associated with follow-up pain (temporal summation: mean r = 0.37, 95% CI: 0.17-0.54; conditioned pain modulation: mean r = 0.36, 95% CI: 0.20-0.50), whereas baseline mechanical threshold modalities were predictive of follow-up disability (mean r = 0.25, 95% CI: 0.03-0.45). Quantitative sensory testing indices of pain hypersensitivity might help develop targeted interventions aiming to improve outcomes across a range of musculoskeletal conditions.

Figure 1

PRISMA flow chart of the study selection process

Figure 2

Forest plots showing the overall association (r-correlations and β-coefficients) between QST and follow-up pain. CI: Confidence Interval, CPM: Conditioned Pain Modulation, CPT: Cold Pain Detection Threshold, DNIC: Diffuse Noxious Inhibitory Control EPT: Electrical Pain Threshold, EST: Electrical Sensation Threshold, MEP: Motor Evoked Potentials, Observ.: Observational Cohort Study PDT: Pressure Detection Threshold, PPT: Pain Pressure Detection Threshold, QST: Quantitative Sensory Testing, RCT: Randomised Controlled Trial TS: Temporal Summation. Forest plot showing the pooled unadjusted (0.31, 95%CI: 0.23 to 0.38) and adjusted correlation (0.18, 95%CI: 0.11 to 0.25) of QST modalities with musculoskeletal pain. The Unadjusted Correlation plot has been derived through the incorporation of Correlation-Coefficient data (Pearson’s or Spearman’s r) expressing a univariate association unadjusted by other factors while the Adjusted Correlation plot has been derived through the incorporation of β-coefficient data from linear or logistic regressions expressing a multivariate association.

Figure 3

Funnel plots for QST studies (n=18) examining the capacity of pain hypersensitivity (as measured by QST) to predict or associate with pain at follow-up depicting A: unadjusted (r-correlation) data with little or no indication of publication bias due to their symmetrical presentation and B: adjusted (β-coefficient) data with an indication of publication bias due to asymmetry. The axes on both graphs are different scales.

Figure 4

Forest plot showing the overall association (r-correlations and β-coefficients) between QST and follow-up disability in musculoskeletal conditions. CI: Confidence Interval, CPM: Conditioned Pain Modulation, CPT: Cold Pain Detection Threshold, EPT: Electrical Pain Threshold, HNCS: Heterotopic Noxious Counter-Stimulation, Observ.: Observational Cohort Study PPT: Pain Pressure Detection Threshold, QST: Quantitative Sensory Testing, RCT: Randomised Controlled Trial, THPR: Tonic Heat Pain Response. Forest plot showing the pooled unadjusted (0.30, 95%CI: 0.19 to 0.40) and adjusted correlation (0.35, 95%CI: 0.21 to 0.49) of QST modalities with musculoskeletal disability. The Unadjusted Correlation plot has been derived through the incorporation of Correlation-Coefficient data (Pearson’s or Spearman’s r) expressing a univariate association unadjusted by other factors while the Adjusted Correlation plot has been derived through the incorporation of β-coefficient data from linear or logistic regressions expressing a multivariate association.