Table 5 Overview of Significant Physical Prognostic Factors: synthesis across included studies [bivariate and multivariable analyses when reported are documented here for consistency - reporting was inconsistent across studies]

Oswestry Disability Index [ODI]

Silverplats et al., 2010

LOW risk of bias

Bivariate analyses:

● Patients with worse pre-operative ODI scores were more likely to report improvement in leg pain [dichotomized as improvement versus no improvement / worse]. Patients with improved leg pain had pre-operative mean ODI 52 compared to 42 in no improvement/worse group [p = 0.040].

● Patients with worse pre-operative ODI scores were more likely to report improvement in back pain [dichotomized as improvement versus no improvement / worse]. Patients with improved back pain had pre-operative mean of 52 compared to 44 in no improvement/worse group [p = 0.040].

Multivariable analyses:

● ODI was not a significant predictor when using the full model of potential predictors [no measure of association reported] at 2 years or long term follow up.

Pre-operative ODI was not significant as a prognostic factor for leg pain or for back pain at 2 years or long term follow up [mean 7.3 ± 1.0 years].

+

Very low

Using GRADE, there is very low level evidence that ODI is not associated with patient outcome.

Solberg et al., 2005

HIGH risk of bias

Multivariable analyses:

● Using change in ODI score as a dichotomous variable (deterioration or no deterioration of score) in binary stepwise logistic regression analyses, a low pre-operative ODI score was an independent risk factor for ‘deterioration’ [β [age adjusted] 0.087, p = 0.011; β [independent risk factor] –0.216, p = 0.013].

● Using ODI raw score at 12 months as a dichotomous variable [“good” ODI score > 39, or “poor” outcome] pre-operative ODI was not an independent risk factor for a “poor” outcome [no measure of association reported].

Lower ODI score [β = − 0.0442, p < 0.001] pre-operatively was a predictor of less improvement in ODI score.

Pre-operative ODI was significant as a prognostic factor for post-operative disability [ODI] at 12 months.

Higher pre-operative ODI predicts better outcome [lower ODI] at 12 months.

Duration of back pain

Nygaard et al., 2000

HIGH risk of bias

● Multivariable analyses:

Multiple linear regression analysis demonstrated that pre-operative duration of back pain was not predictive of clinical overall score [COS]; coefficient β [Standard error] = − 0.26 [0.16], t test − 1.65, p = 0.100.

Pre-operative duration of back pain was not significant as a prognostic factor for COS at 12 months.

+

Very low

Using GRADE, there is very low level evidence that duration of back pain is not associated with patient outcome.

Solberg et al., 2005

HIGH risk of bias

Multivariable analyses:

● Using change in ODI score as a dichotomous outcome variable in binary stepwise logistic regression analyses, duration of back pain was not an independent risk factor for ‘deterioration’ [β [age adjusted] 0.001, p = 0.304].

● Using ODI raw score at 12 months as a dichotomous variable [“good” ODI score > 39, or “poor” outcome] in multivariate analyses, duration of back pain was not an independent risk factor for a “poor” outcome [no measure of association reported].

Pre-operative duration of back pain was not significant as a prognostic factor for disability [ODI] at 12 months.

Duration of leg pain

Fischer et al., 2004

HIGH risk of bias

Multivariable analyses:

Patients with longer pre-operative duration of leg pain were more likely to report less improvement in Pain Disability Score [PDS] [p = 0.026] after adjustment for gender, age and pre-operative PDS. Mean change PDS 24.4 for duration 0–3 months, 20.0 for duration 3.1–9 months, 13.1 for duration > 9 months [no measures of association reported].

Pre-operative duration of leg pain was significant as a prognostic factor for PDS at 12 months.

Shorter pre-operative duration of leg pain predicts better outcome [lower PDS] at 12 months.

a

am

Lewis et al., 1987 and Weir et al., 1979

HIGH risk of bias

Bivariate analyses:

● Duration leg pain < 17 months associated with complete relief of back pain in 43/71 cases [61%] at 1 year; 39/65 cases [60%] at 5–10 years.

● Duration leg pain ≥17 months associated with complete relief of back pain in 12/19 cases [63%] at 1 year; 9/15 cases [60%] at 5–10 years.

● Duration leg pain < 17 months associated with complete relief of leg pain in 54/71 cases [76%] at 1 year; 43/65 cases [66%] at 5–10 years.

● Duration leg pain ≥17 months associated with complete relief of leg pain in 12/19 cases [63%] at 1 year; 6/15 cases [40%] at 5–10 years.

● Significant association [chi-square of Fisher’s exact test] at 1-year follow-up review between duration leg pain and relief of back or leg pain [above]. Shorter duration of leg pain before surgery is associated with relief of leg pain following surgery. Not significant at 5–10 years [results not reported].

Pre-operative duration of leg pain was not significant as a prognostic factor for leg pain and for back pain at 12 months [no multivariable analyses].

Nygaard et al., 2000

HIGH risk of bias

Multivariable analyses:

● Patients with longer pre-operative duration of leg pain were more likely to report less improvement in COS. Multiple linear regression analysis, coefficient β [Standard error] = 0.98 [0.3], t test 3.23, p = 0.0016.

Pre-operative duration of leg pain was significant as a prognostic factor for COS at 12 months.

Shorter pre-operative duration of leg pain predicts better outcome [lower COS] at 12 months.

Silverplats et al., 2010

LOW risk of bias

Bivariate analyses:

● Patients with longer pre-operative duration of leg pain were more likely to report improvement in leg pain. Pre-operative short duration [< 6 months] of leg pain predicts good outcome on MacNab [dichotomized outcome] classification [p = 0.039] at 2-year follow up and predicts patient satisfaction with treatment [p = 0.019] at long term follow-up [mean 7.3 ± 1.0 years].

Multivariable analyses:

● Duration of leg pain was not a significant predictor when using the full model of potential predictors [no measure of association reported].

Pre-operative duration of leg pain was not significant as a prognostic factor for leg pain or health-related quality of life [EQ-5D] at 2 year and long term follow up [mean 7.3 ± 1.0 years].

Silverplats et al., 2011

LOW risk of bias

Multivariable analyses:

● Duration of leg pain was not a significant predictor for EuroQol-5 Dimension, EQ-5D at 2 years [no measure of association reported].

Solberg et al., 2005

HIGH risk of bias

Multivariable analyses:

● Duration of leg pain was not an independent risk factor for ‘deterioration’ [β [age adjusted] 0.008, p = 0.006; β [independent risk factor] 0.005, p = 0.572]; using change in ODI score as a dichotomous outcome variable (deterioration or no deterioration).

● Using ODI raw score at 12 months as a dichotomous outcome variable [“good” ODI score > 39, or “poor” outcome] duration of leg pain was not an independent risk factor for a “poor” outcome [no measure of association reported].

Pre-operative duration of leg pain was not significant as a prognostic factor for disability [ODI] at 12 months.

Severity leg pain

Divecha et al., 2014

HIGH risk of bias

Bivariate analyses:

● Patients with worse pre-operative leg pain were more likely to report improvement in functional outcome. Pearson’s correlation coefficient between pre-operative leg pain [%] and Core Outcome Measures Index [COMI] score at 12 months was −0.394 (95% CI -0.653, − 0.053; p = 0.0256].

Multivariable analyses:

● Patients with higher pre-operative leg pain had significantly lower COMI [R² = 0.155, p = 0.03] at 12 months.

Pre-operative severity of leg pain was significant as a prognostic factor for functional outcome [COMI] at 12 months.

Higher severity pre-operative leg pain predicts better outcome [lower COMI] at 12 months.

++

Low

Using GRADE, there is low level evidence that higher severity of pre-operative leg pain predicts better Core Outcome Measures Index at 12 months and better post-operative leg pain at 2 and 7 years.

Silverplats et al., 2010

LOW risk of bias

Bivariate analyses:

● Patients with higher pre-operative leg pain were more likely to report improvement in leg pain. Patients with improved leg pain had higher leg pain pre-operatively on VAS [60 versus 47, p = 0.008]

Multivariable analyses:

● For improvement in leg pain the only significant predictor among all potential predictors was pre-operative VAS leg pain (p = 0.039). Pre-operative VAS leg pain was also the first and only predictor selected by the stepwise procedure [no measure of association reported].

Pre-operative severity of leg pain was significant as a prognostic factor for leg pain at 2 years and long term follow up [mean 7.3 ± 1.0 years].

Pre-operative severity of leg pain was not significant as a prognostic factor for EQ-5D at 2 years or long term follow up [mean 7.3 ± 1.0 years].

Higher severity pre-operative leg pain predicts better outcome [lower leg pain] at 2 years and long term follow up [mean 7.3 ± 1.0 years].

Silverplats et al., 2011

LOW risk of bias

Bivariate analyses:

Patients with higher pre-operative leg pain were more likely to report improvement in health-related quality of life. Pre-operative VAS leg pain was correlated with change in EQ-5D at 2-year follow-up [r = 0.33, p = 0.002] and at 7-year follow up [r = 0.23, p = 0.04].

Multivariable analyses:

VAS leg pain was not identified as a significant predictor of EQ-5D [no measure of association reported].

Solberg et al., 2005

HIGH risk of bias

Bivariate analyses:

Patients with higher pre-operative leg pain were more likely to report improvement in disability. Pre-operative VAS leg pain mean [SD; 95%CI] was 63.4 [27.5; 59.3 to 67.4], and at 12 months was 16.8 [21.1; 13.7 to 20.0]. Improvement was 46.5 [33.4, 41.6 to 51.4].

Multivariable analyses:

Using change in ODI score as a dichotomous outcome variable, VAS leg pain was not an independent risk factor for ‘deterioration’ [β [age adjusted] -0.009, p = 0.481] at 12 months. Using ODI raw score at 12 months as a dichotomous outcome variable [“good” ODI score > 39, or “poor” outcome] VAS leg pain was not an independent risk factor for a “poor” outcome [no measure of association reported].

Pre-operative severity of leg pain was not significant as a prognostic factor for disability [ODI] at 12 months.

Severity back pain

Silverplats et al., 2010

LOW risk of bias

Bivariate analyses:

Patients with higher pre-operative back pain were more likely to report improvement in back pain. Patients with improved back pain had higher VAS back pain pre-operatively [53 versus 36, p = 0.001].

Multivariable analyses:

Pre-operative back pain was not a significant predictor when [no measure of association reported] at 2 years or long term follow up [mean 7.3 ± 1.0 years].

Pre-operative severity of back pain was not significant as a prognostic factor for back pain or EQ-5D at 2 years or long term follow up [mean 7.3 ± 1.0 years].

+ Very low

Using GRADE, there is very low level evidence that severity of back pain is not associated with patient outcome.

Silverplats et al., 2011

LOW risk of bias

Bivariate analyses:

Back pain at baseline was not significantly correlated with change in EQ-5D at any follow-up.

Multivariable analyses:

Back pain was not identified as a significant predictor of EQ-5D at 2 years follow up [no measure of association reported].

Solberg et al., 2005

HIGH risk of bias

Bivariate analyses:

Baseline VAS back pain (0–100 points) mean [SD; 95%CI] = 51.7 [29.3; 47.4, 56.0]. 12 months 21.3 [22.6; 18.0, 24.6]. Improvement 31.4 [35.6, 25.2–35.6]. VAS back pain pre-operatively not predictive of follow up ODI score at 12 months.

Multivariable analyses:

VAS back pain was not an independent risk factor for ‘deterioration’ [β–[age adjusted] 0.003, p = 0.800]; using change in ODI score as a dichotomous variable [deterioration or no deterioration].

Using ODI raw score at 12 months as a dichotomous variable [“good” ODI score > 39, or “poor” outcome], VAS back pain was not an independent risk factor for a “poor” outcome [no measure of association reported].

Pre-operative severity of back pain was not significant as a prognostic factor for disability [ODI] at 12 months.

Health-related quality of life

[EuroQol-5 Dimension, EQ-5D]

Silverplats et al., 2011

LOW risk of bias

Bivariate analyses:

Patients with lower pre-operative EQ-5D were more likely to report improvement in health-related quality of life [EQ-5D]. Pre-operative EQ-5D was correlated with change in EQ-5D at 2-year or 7-year follow-ups [r = −0.70, p < 0.001 and r = − 0.71, p < 0.001].

Multivariable analyses:

The only significant predictor of outcome was pre-operative EQ-5D score. The influence of baseline EQ-5D score was estimated [β = − 1.0, 95% CI: − 1.2, − 0.8] at 2 years.

Pre-operative EQ-5D was significant as a prognostic factor for health-related quality of life [EQ-5D] at 2 years.

Lower pre-operative EQ-5D predicts better outcome [lower EQ-5D] at 2 years.

+

Very low

Using GRADE, there is very low level evidence that a lower pre-operative EQ-5D predicts better EQ-5D at 2 years.

Ipsilateral Straight Leg Raise [SLR]

Lewis et al., 1987 and Weir, 1979

HIGH risk of bias

Bivariate analyses:

Positive ipsilateral SLR associated with complete relief of back pain in 47/75 cases [63%] at 1 year; 41/69 cases [59%] at 5–10 years. Negative ipsilateral SLR associated with complete relief of back pain in 9/16 cases [56%] at 1 year; 8/12 cases [67%] at 5–10 years.

Positive ipsilateral SLR associated with complete relief of leg pain in 59/75 cases [79%] at 1 year; 42/69 cases [61%] at 5–10 years. Negative ipsilateral SLR associated with complete relief of leg pain in 8/16 cases [50%] at 1 year; 8/12 cases [67%] at 5–10 years.

Significant association [chi-square of Fisher’s exact test] at 1-year follow-up review between ipsilateral SLR and relief of back or leg pain [above]. Positive ipsilateral SLR before surgery is associated with relief of back and leg pain following surgery. Not significant at 5–10 years [results not reported].

Pre-operative ipsilateral SLR was not significant as a prognostic factor for back pain or leg pain at 5–10 years [no multivariable analyses].

+

Very low

Using GRADE, there is very low level evidence that straight leg raise is not associated with patient outcome.

Forward bend

Lewis et al., 1987 and Weir, 1979

HIGH risk of bias

Bivariate analyses: Forward bend to knee associated with complete relief of back pain in 41/58 cases [71%] at 1 year; 33/50 cases [66%] at 5–10 years. Forward bend to mid tibia or floor associated with complete relief of back pain in 15/33 cases [45%] at 1 year; 16/31 cases [52%] at 5–10 years.

Forward bend to knee associated with complete relief of leg pain in 48/58 cases [83%] at 1 year; 34/50 cases [68%] at 5–10 years. Forward bend to mid tibia or floor associated with complete relief of leg pain in 19/33 cases [58%] at 1 year; 13/31 cases [42%] at 5–10 years.

Significant association [chi-square of Fisher’s exact test] at 1-year follow-up review between forward bend and relief of back or leg pain [above]. Positive forward bend to knee before surgery is associated with relief of back and leg pain following surgery. Not significant at 5–10 years [results not reported].

Pre-operative forward flexion was not significant as a prognostic factor for back pain or leg pain at 5–10 years [no multivariable analyses].

+

Very low

Using GRADE, there is very low level evidence that forward bend is not associated with patient outcome.