Classification approaches for treating low back pain have small effects that are not clinically meaningful: a systematic review with meta-analysis

Key points

1. 01Classification systems (for example the McKenzie method, STarT Back Tool, and treatment-based classification) try to overcome the modest, generic effects of standard low back pain care by tailoring treatment to subgroups.
2. 02Classified treatment did beat generic care, but only by a small amount: pain SMD -0.31 and disability SMD -0.27 at end of treatment, both below clinically meaningful thresholds (20/100 for pain, 10/100 for disability).
3. 03Certainty of evidence was low, and the 95% prediction intervals were wide and included benefit for the generic comparators too.
4. 04No individual classification system was superior to generic care at any time point; only the McKenzie method showed a benefit for disability in the intermediate term, and even that was not clinically meaningful.
5. 05Most systems classify on pathoanatomical features only, whereas low back pain also involves central and psychological factors, which the authors flag as a key limitation of current approaches.

How it was conducted

Design

Systematic review with random-effects meta-analysis (Hartung-Knapp-Sidik-Jonkman adjustment), PROSPERO-registered; GRADE for certainty

Search

MEDLINE, Embase, CINAHL, Web of Science, CENTRAL from inception to June 21, 2021; 5209 records, 64 articles (58 trials) in narrative synthesis

Intervention

Nonsurgical LBP classification systems (McKenzie, STarT Back, treatment-based, movement system impairment, others) with treatment matched to subclass

Comparator

Generic interventions: active, passive, education/advice, usual care, and mixed care that did not subclassify

Outcomes

Patient-reported back pain intensity, leg pain intensity, and disability across end of treatment, short, intermediate, and long term

Analysis

Standardized mean difference (Hedges' g) with 95% CI; Cochrane RoB 2 risk of bias; subgroup and sensitivity analyses

What they found

• Back pain intensity at end of intervention: SMD -0.31 (95% CI -0.54 to -0.07; P = .014; n = 4416, 21 trials), low certainty, favoring classification but below the clinically meaningful threshold.
• Disability at end of intervention: SMD -0.27 (95% CI -0.46 to -0.07; P = .011; n = 4809, 24 trials), low certainty, also below the clinically meaningful threshold.
• Leg pain intensity at end of intervention: SMD -0.20 (95% CI -0.54 to 0.14), not significant, very low certainty.
• No specific classification system outperformed generic care; only the McKenzie method showed a disability benefit in the intermediate term (SMD -0.30, 95% CI -0.60 to 0.00), still not clinically meaningful.
• Risk of bias: among classification-system trials, 24% low, 43% some concerns, 33% high; subclass trials had none at low risk of bias.

Limitations

• Certainty of evidence was low to very low for the main outcomes, downgraded for risk of bias, inconsistency, and imprecision.
• Many included trials had high risk of bias or some concerns, and standard deviations had to be imputed or converted for several trials.
• Wide prediction intervals indicate the true effect in a new setting could favor either classification or generic care.
• Classification systems have only poor to good reliability, so patients may not have reliably received treatment truly matched to their subgroup.

Why it matters

For patients

Being formally sorted into a low back pain subgroup and given matched treatment is unlikely to relieve your pain or disability noticeably more than good general care.

For clinicians

There is not yet enough evidence to prefer any classification system over generic, evidence-based care for low back pain, though the small signal does not rule out future better-designed systems.

For readers

Tailoring low back pain treatment by classification produced real but clinically trivial gains over one-size-fits-all care.