Variability in Strength, Pain, and Disability Changes in Response to an Isolated Lumbar Extension Resistance Training Intervention in Participants with Chronic Low Back Pain

Abstract

Strengthening the lumbar extensor musculature is a common recommendation for chronic low back pain (CLBP). Although reported as effective, variability in response in CLBP populations is not well investigated. This study investigated variability in responsiveness to isolated lumbar extension (ILEX) resistance training in CLBP participants by retrospective analysis of three previous randomized controlled trials. Data from 77 participants were available for the intervention arms (males = 43, females = 34) 37 participants data (males = 20, females = 17) from the control arms. Intervention participants had all undergone 12 weeks of ILEX resistance training and changes in ILEX strength, pain (visual analogue scale; VAS), and disability (Oswestry disability index; ODI) measured. True inter-individual (i.e., between participants) variability in response was examined through calculation of difference in the standard deviation of change scores for both control and intervention arms. Intervention participants were classified into responder status using k-means cluster analysis for ILEX strength changes and using minimal clinically important change cut-offs for VAS and ODI. Change in average ILEX strength ranged 7.6 Nm (1.9%) to 192.1 Nm (335.7%). Change in peak ILEX strength ranged −12.2 Nm (−17.5%) to 276.6 Nm (169.6%). Participants were classified for strength changes as low (n = 31), medium (n = 36), and high responders (n = 10). Change in VAS ranged 12.0 mm to −84.0 mm. Participants were classified for VAS changes as negative (n = 3), non-responders (n = 34), responders (n = 15), and high responders (n = 19). Change in ODI ranged 18 pts to −45 pts. Participants were classified for ODI changes as negative (n = 2), non-responders (n = 21), responders (n = 29), and high responders (n = 25). Considerable variation exists in response to ILEX resistance training in CLBP. Clinicians should be aware of this and future work should identify factors prognostic of successful outcomes.

A Systematic Review Examining the Approaches Used to Estimate Interindividual Differences in Trainability and Classify Individual Responses to Exercise Training

“…These large CIs were likely attributable to the small sample sizes of these two studies. Interestingly, only the study by Steele et al (2017) , which examined strength responses to resistance training, revealed consistent evidence of interindividual differences in trainability as all ES IR 90% CIs lay above zero.…”

“…However, strikingly few studies (∼9.5%; 8/84 studies) published in 2017 onward [i.e., after Atkinson et al’s (2019) SD IR paper] statistically estimated interindividual differences in trainability. Seven of these studies ( Steele et al, 2017 ; Williamson et al, 2017 , 2018 ; Bonafiglia et al, 2019a ; Hammond et al, 2019 ; Walsh et al, 2020 ; Yu et al, 2020 ) used the SD IR , and one study explored different SD IR approaches using a variety of statistical parameters ( Hecksteden et al, 2018 ). The 2016 study by Leifer et al (2016) used Levene’s tests to compare variability in observed responses between control and exercise groups – an approach that follows the same principles as the SD IR ( Bonafiglia et al, 2019a ).…”

“…Our ES IR calculations highlighted mixed evidence for the presence of interindividual differences in trainability ( Figure 3 ). For instance, resistance training in the Steele et al (2017) study appeared to cause variability in strength responses, whereas aerobic training in the Leifer et al (2016) re-analysis may not have led to interindividual differences in cardiometabolic responses because the 90% CIs crossed zero. Interestingly, fasting insulin from the Leifer et al (2016) study and peak strength from the Steele et al (2017) study had similar ES IR values, but only peak strength revealed evidence of interindividual differences in trainability because its 90% CI lay fully above zero ( Figure 3 ).…”

“…For instance, resistance training in the Steele et al (2017) study appeared to cause variability in strength responses, whereas aerobic training in the Leifer et al (2016) re-analysis may not have led to interindividual differences in cardiometabolic responses because the 90% CIs crossed zero. Interestingly, fasting insulin from the Leifer et al (2016) study and peak strength from the Steele et al (2017) study had similar ES IR values, but only peak strength revealed evidence of interindividual differences in trainability because its 90% CI lay fully above zero ( Figure 3 ). Because fasting insulin had a much larger sample size than peak strength ( n = 1188 vs. n = 114), its larger 90% CI may reflect how blood-based physiological outcomes have larger random measurement error compared with strength-based performance outcomes.…”

“…For studies that statistically estimated interindividual differences in trainability, we extracted sample sizes and standard deviations (SD) of baseline values and change scores. We used WebPlotDigitizer ( Rohatgi, 2018 ) to calculate SDs of change scores from two studies ( Steele et al, 2017 ; Hammond et al, 2019 ) that did not report these values but presented individual data in their figures. For studies that classified individual responses, we extracted mean changes, thresholds used to classify “responders,” and response rates.…”