Comparison of Clinical and Biomechanical Characteristics between Individuals with Lower Limb Amputation with and without Lower Back Pain: A Systematic Review and Meta-Analysis

Abstract

Background

Lower back pain is a debilitating condition common to individuals with lower limb amputation. It is unclear what risk factors contribute to the development of back pain. This study systematically reviewed and analyzed the available evidence regarding the clinical and biomechanical differences between individuals with amputation, with and without lower back pain.

Methods

A literature search was conducted in PubMed, Web of Science, Scopus, and CINAHL databases in November 2020 and repeated in June 2021 and June 2022. Studies were included if they reported comparisons of demographic, anthropometric, biomechanical, and other clinical variables between participants with and without LBP. Study quality and potential for reporting bias were assessed. Meta-analyses were conducted to compare the two groups.

Findings

Thirteen studies were included, with aggregated data from 436 participants (239 with LBP; 197 pain free). The median reporting quality score was 37.5%. The included studies enrolled participants who were predominantly male (mean=91.4%, range=77.8-100%) and with trauma-related amputation. Meta-analyses showed that individuals with LBP exhibited moderate (3.4 out of 10) but significantly greater pain than those without LBP. We found no between-group differences in age, height, weight, BMI, and time since amputation (p=0.121-0.682). No significant differences in trunk/pelvic kinematics during gait were detected (p=0.07-0.446) between the groups.

Interpretation

Demographic, anthropometric, biomechanical, and simple clinical outcome variables may be insufficient for differentiating the risk of developing back pain after amputation. Investigators should be aware of the existing gender bias in sampling and methodological limitations, as well as to consider incorporating psychosocial measures when studying LBP in this clinical population.

1. Introduction

Lower back pain (LBP) is one of the most common health conditions that can develop following lower limb amputation (LLA). Up to 90% of the individuals with LLA develop LBP, among them 50% report moderate to severe pain that is persistent and bothersome.^1,2 LBP, even at a moderate level, can interfere with daily activities in individuals with LLA,³ regardless of their functional⁴ and physical activity levels.⁵ The etiology of LBP after LLA has been widely speculated but is still currently unclear^6-8.

While the development of LBP after LLA is generally recognized as multifactorial, the contributors have been broadly categorized into biomechanical and psycho-social factors.⁷ Empirical and clinical evidence suggested that changes in movement patterns after LLA play a possible role in the development of LBP symptoms.^6,9,10 In a series of studies by Hendershot et al., the researchers demonstrated that alterations in trunk movement and neuromuscular control were commonly observed in individuals with LLA during a variety of locomotor tasks including walking and sit-to-stand.^11-19 Therefore, characteristics such as posture, gait, muscle morphology and activation patterns that are typically affected by LLA have been the primary focus of investigations.⁸ However, given that both LLA and LBP can independently affect movement patterns, direct comparisons of the biomechanical characteristics within individuals with LLA, with and without LBP, is a logical step to understanding LBP in this unique population.

In a recent systematic review, Highsmith et al. examined the strength of evidence related to the presence and severity of LBP secondary to LLA for the purpose of formulating empirical evidence statements to guide practice and future research. With respect to the etiology of LBP after LLA, the authors concluded that the confidence level of evidence regarding risk factors such as amputation level, leg length, posture, and spinal/pelvic kinematics are low with conflicting results from different studies.⁶ Biomechanical research concerning the LLA population is often criticized for being underpowered.²⁰ To this end, there is value in a pooled meta-analysis to examine the existing body of literature pertaining to the comparison of characteristics between those with and without LBP.

The purpose of this study was to systematically and quantitatively assess, synthesize, and review the currently available studies on the clinical and biomechanical differences between individuals with LLA, with and without LBP. Specifically, a series of meta-analyses were conducted for pooled comparisons between the two groups, based on data obtained from published studies. Information gained from this study will contribute to the understanding of LBP after LLA and the development of prevention and treatment options for this prevalent and debilitating condition. Assessing the quality of current research in a systematic way can also help identify gaps in the existing literature and research methodologies to facilitate future investigations.

2. Methods

This review was conducted based on the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA). The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO #CRD42021223787).^21,22

2.1. Eligibility criteria

In order to conduct a comprehensive analysis based on the full range of available evidence on this topic, we searched published literature for peer-reviewed original research. Studies were included if they met the following criteria: 1) inclusion of individuals with major LLA defined as loss of at least the ankle joint, 2) quantification of demographic, anthropometric, biomechanical, and/or other clinical variables associated with LBP, and 3) comparison of participants with and without LBP. Other inclusion criteria of the selected articles were full-length academic journal articles written in English and published after the year 2000. Articles were excluded if they were conference abstracts, case reports/series, dissertations, review articles, full text not available in English, or did not recruit human participants.

2.2. Search Strategy

Three of the authors met to formulate the research question and identify the PECO parameters prior to the literature search. The parameters included Population: in individuals with LLA; Exposure: clinical and/or biomechanical examination; Comparison: demographic, anthropometric, biomechanical, and other clinical characteristics between individuals with LLA with and without LBP; Outcome: identifying risk factors for LBP.²³ The subsequent literature search was conducted using the PubMed, Web of Science Core Collection, Scopus, and CINAHL databases. Three groups of search terms or keywords were queried in all databases: (1) low back pain and synonyms; (2) lower limb amputation and synonyms; (3) biomechanical, clinical measures, and synonyms (Table 1). The search was guided by Boolean operators AND and OR with the intention to narrow down to the relevant contents in the search fields of title, abstract, and keywords. For example, the search terms for one of the queries were: (low* back pain OR lumbago OR Sciatica OR back ache) AND (amput* OR limb loss). The initial search was conducted in November 2020 and then repeated in June 2021 and June 2022. Reference lists from all accessed articles were screened and cross-referenced for additional eligible studies. After duplicate records were removed, one author (SPL) screened the titles and abstracts of all articles based on the inclusion criteria described above. Full-text manuscripts of the remaining articles were retrieved and additionally screened by two authors (SPL and JC; Figure 1).

Table 1:

Three Groups of Search Terms used

Search term groups
Low back pain and synonyms	Lower limb loss/amputation and synonyms	Biomechanical, clinical measures and synonyms
back pain	amput*	biomech*
sciatica	limb loss	kinemat*
backache	prosthe*	motion
spinal pain		kinet*
spinal disease		electromyograph*
spinal disorder		EMG
spinal dysfunction		strength
lumbar pain		force
lumbago		postur*
		gait
		walk*
		risk factor
		predictor
		outcome
		measur*
		function*

Figure 1:

Flow chart summarizing the study selection process

2.3. Quality Assessment

Reporting quality and potential for bias of the included studies were assessed using an adapted 13-item checklist by van der Worp et al. (Table 2)²⁴, which has previously been used for systematically reviewing musculoskeletal disease research.^24-27 The checklist was designed to assess study quality in the following areas: study population, measurements and outcome assessment, as well as data analysis and presentation. Recently our group has used this checklist to perform a systematic review of gait biomechanical studies comparing individuals with and without LBP.²⁸ Two authors (SPL and JC) assessed and scored all eligible articles (13 studies, Figure 1). When there were discrepancies in scores, the two scoring authors discussed each article to reach a consensus. A third author (JAS) was consulted when a consensus could not be reached.

Table 2:

Checklist used for assessing methodological quality and potential for bias of the included studies

Domain	Item #	Description	CS	CC
Study objective
	1	Positive, if the study had a clearly defined objective.	Not scored
Study population
	2	Positive, if the main features of the study population are described (sampling frame and distribution of the population according to age andsex).	Not scored
	3	Positive, if cases and controls are drawn from the same population and a clear definition of cases and controls is given and if participants who developed LBP within the past 3 months are excluded from the control group.		+
	4	Positive, if the participation rate is at least 80% or if the participation rate is 60–80% and the non-response is not selective (data shown).	+	+
Measurements
	5	Positive, if data on participant history of LBP and LLA are collected and included in the statistical analysis.	+	+
	6	Positive, if the outcome is measured in an identical manner among cases and controls.		+
	7	Positive, if the outcome assessment is blinded with respect to LBP status (must be stated specifically).	+	+
	8	Positive, if the outcome is assessed at a time before the occurrence of the LBP.		+
Assessment of the outcome
	9	Method for assessing LBP status: physical examination blinded to exposure status (+); self-reported: specific questions relating to LBP symptoms/severity/use of manikin (+), single question (−).	+	+
Analysis and data presentation
	10	Positive, if the measures of association were presented including odds ratio/risk ratio and confidence intervals, and/or group comparisons were presented including confidence intervals in the results.	+	+
	11	Positive, if the analysis is controlled for confounding or effect modification:individual factors.	+	+
	12	Positive, if the analysis is controlled for confounding or effect modification: other factors (examples: controlling for internal validity threats such as in instrumentation and testing).	+	+
	13	Positive, if the number of cases in the final multivariate model was at least 10 times the number of independent variables in the analysis.	+	+
Total possible score			8	11

CS: cross-sectional study design; CC: case-control study design

Included study types based on our search and screen results were either cross-sectional or case-control design. Specifically, a cross-sectional design designation was given if the study recruited a sample of participants with LLA (not knowing whether they have LBP), and the identification/assessment for their LBP status was part of the investigation. Each article was given a total quality score as the sum of all quality assessment criteria met (8 possible points for cross-sectional studies and 11 possible points for case-control studies). A percentage of the possible maximum score was calculated for each article, with high quality of reporting defined as if the study scored 50% or higher.^24,25,29

2.4. Data extraction

The following data were extracted from all included studies: study design, sample size, inclusion/exclusion criteria, participant characteristics including age, height, weight, BMI, sex, as well as time since, cause, and level of amputation. LBP-related clinical outcomes (i.e., reported pain level and Oswestry Disability Index score) and biomechanical measures (i.e., spinal/pelvic kinematics) were extracted. Data from specific outcome measures were reviewed, synthesized, and analyzed if at least three unique studies reported the variable. In cases where group means, standard deviations, and/or individual participant data were absent from published papers, attempts were made to contact the study authors when the data required for analysis were not available in the publication.

2.5. Meta-analysis

For the pooled meta-analyses, aggregated comparison groups (i.e. LBP vs. No-LBP) were synthesized based on data from at least three studies reporting the same variable. We contacted the authors who have published more than one study to ensure that data from the same samples were not double-counted. Authors were also contacted when group means and standard deviations were not provided in the published articles. The weighted mean and pooled standard deviation were computed based on the sample size of each study to summarize anthropometric and biomechanical characteristics among the included studies. For the meta-analysis of each variable, a random effects model was used to calculate standardized mean differences (SMD) and 95% confidence intervals between the LBP and No-LBP groups. A heterogeneity measure, I² statistic, was computed to describe the percentage of variation across studies due to heterogeneity rather than chance in each model.³⁰ The analyses were conducted and organized using RStudio v1.3.1093 (RStudio Team, Boston, MA) by a blinded biostatistician. The significance level was set to 0.05.

3. Results:

Our searches of scientific databases yielded 465 articles (Figure 1). After removing duplicates, titles and abstracts of 373 articles were screened, resulting in 67 articles. Further examination of the full text of the 67 articles led to the exclusion of 55 articles, mostly due to the lack of LBP versus pain-free (No-LBP) group comparison (n=36). One recently published study (2022) was manually entered. Overall, 13 articles were included in this systematic review.

The median score for research quality was 37.5% (range=25 to 75%). Six studies scored 50% or higher. Ten of the 13 articles were cross-sectional studies.^4,5,20,31-37 Among all studies, most used samples of convenience and did not report the participation rate, or that the response rate was low. Blinding of the investigators regarding participants’ LBP status was only implemented in 3 studies.

Of the 13 reviewed studies, a total of 436 participants were included and compared based on their LBP status (n=239 with LBP and n=197 without LBP). Study authors were approached to avoid double-counting of participants. All of the included studies reported participants’ amputation levels (i.e. transfemoral vs. transtibial); 69.2% of the participants in the studies (n=302) had LLA at the transfemoral level. In Table 3, we provided a summary of the sample population, amputation etiology, LBP status criteria, age and sex of the participants, sample size, and the primary outcome measures. The sampled populations included individuals in the general population and military personnel/Veterans. Most of the included studies enrolled participants who were predominantly male (mean=91.4%, range=77.8 to 100%) and with trauma-related amputation (Table 3). The criteria used to determine LBP status varied but most studies assessed pain duration and frequency, as well as pain severity using a visual analog scale.

Table 3:

Summary of the included studies

Author, year	Sampled population (amputation etiology)	LBP status criteria	Age (years)	Sex (% male)	Sample size	Amputation level (% transfemoral)	Primary outcome measures
Acasio et al., 2022	NR (trauma only)	LBP occurring at least half the days for > 6 months	LBP=35.8 No-LBP=33.9	100%	19 LBP 16 No-LBP	31.4%	Pain (0-10; current and last 7 days), spinal and pelvic kinematics during gait, model-based segmental and muscle forces
Butowicz et al., 2019	NR	LBP > 3 months and pain on at least half the days in 6 months	LBP=35.8 No-LBP=33.1	100%	19 LBP 13 No-LBP	29.4%	Pain (VAS), LBP-related disability (ODI), stability and trunk muscle activation during sitting
Butowicz et al., 2020	Military or Veterans (trauma only)	Duration > 3 months and pain on at least half the days in the last 6 months	LBP=35.8 No-LBP=33.1	100%	20 LBP 13 No-LBP	30.3%	Pain (NPRS), LBP-related disability (ODI), chronic LBP impact (RTF-Impact), and psychometric measures (pain catastrophizing, anxiety, depression, and kinesiophobia)
Devan et al., 2017	General population (trauma, congenital, or tumors only, excluding dysvascular causes)	LBP > 12 months	LBP=47.2 No-LBP=42.3	83.3%	9 LBP 9 No-LBP	33%	Spinal and pelvic kinematics during gait
Devan et al., 2012	General population (trauma only)	Having ADL-restricting LBP in the last 4 weeks	LBP=53.4 No-LBP=58.3	82.8%	89 LBP 51 No-LBP	100%	Pain level (0-10) and physical activity level
Fatone et al., 2016	General population (all causes)	LBP more than once during the last 30 days	LBP=47.5 No-LBP=48.3	82.6%	12 LBP 11 No-LBP	100%	Spinal and pelvic kinematics during gait
Friel et al., 2005	Amputee support group and referrals from prosthetists (all causes)	Pain level (VAS) within last week (>2 or 5 out 10)	LBP=48.2 No-LBP=64.1	83.3%	LBP No-LBP	42.1%	Pain (VAS), LBP-related disability (ODI), lumbopelvic muscle lengths, back extensor endurance
Kulkarni et al., 2005	Patients from a rehabilitation clinic (trauma only)	Presence of back pain (with VAS)	NR	NR	20 LBP* 20 No-LBP*	50%	Lumbar ranges of motion, disc degeneration score, gait (time and GRF), and standing balance
Mahon et al., 2020	Military medical center (specific causes NR)	Presence of back pain (with VAS)	NR	100%	6 LBP† 23 No-LBP†	27.5%	Spinal and pelvic kinematics during gait
Matsumoto et al., 2019	Civilian and Veteran medical clinics (mostly trauma, 82.3%)	Persistent back pain since amputation (with chronic pain grade 0-10)	LBP=55 No-LBP=48	100%	9 LBP 8 No-LBP	100%	Lumbar lordosis and sacral inclination angles
Morgenroth et al., 2010	Civilian and Veteran medical clinics (mostly trauma, 82.3%)	Persistent back pain since amputation (with chronic pain grade 0-10)	LBP=52 No-LBP=49	NR	9 LBP 8 No-LBP	100%	Spinal and pelvic kinematics during gait
Morgenroth et al., 2009	Civilian and Veteran medical clinics (specific causes NR)	Persistent back pain since amputation (with chronic pain grade 0-10)	LBP=52 No-LBP=49	NR	9 LBP 8 No-LBP	100%	Static and dynamic leg-length discrepancy
Russell-Esposito and Wilken, 2014	NR (trauma only)	Frequency of LBP > 1/week	LBP=32.1 No-LBP=28.4	100%	9 LBP 7 No-LBP	100%	Spinal and pelvic kinematics and coordination (continuous relative phase) during gait
Total					239 LBP 197 No-LBP

LBP: low back pain; NR: not reported; VAS: visual analog scale; NPRS: numeric pain rating scale; ODI: Oswestry Disability Index; RTF: NIH Research Task Force Minimal Dataset questionnaire for Chronic Low Back Pain⁵⁶

^*:a subset of the 202 participants in the study who received additional biomechanical testing.²⁹

^†:a subset of the 32 participants in the study whose data was available at the 2-month time point.²⁸

In addition to LBP-related clinical outcomes, reported variables varied greatly among studies, ranging from lab-based biomechanical measures,^{20,31-33,36-38} pain-related psychometrics,³⁵ physical activity level,⁵ and other physical performance and anatomical measures (Table 3).^4,33,39,40

3.1. Comparison of Participants’ Clinical and Anthropometric Characteristics

Clinical and anthropometric characteristics were compared between study participants with and without LBP (Table 4). The pooled analysis showed that self-reported pain level was significantly higher in the LBP group (p<0.001, SMD=1.88, 95% CI=1.23–2.54; Figure 2a). Time since amputation was on average over a year longer in the LBP group, albeit this difference was not statistically significant (p=0.121, 95% CI=0.06–0.42; Figure 2b). No significant differences were detected in study participant’s age, height, weight, BMI, level of amputation, and Oswestry Disability Index score (Table 4).

Table 4:

Summary of the pooled meta-analysis findings. Weighted mean and standard deviation were reported based on the pooled analysis for each variable.

	LBP		No-LBP		p- value	# of study that included the variable
	N	Mean±SD	N	Mean±SD
Clinical and anthropometric variables
Age (years)	213	46.9±12.4	154	47.3±11.3	0.377	13
Height (cm)	67	179.2±7.1	49	178.9±5.3	0.682	4
Weight (kg)	67	90.7±15.6	49	88.3±13.8	0.284	4
BMI (kg/m2)	153	27.8±4.2	103	26.9±4.7	0.128	9
Time since amputation (years)	197	12.2±6.2	138	10.8±7.4	0.121	13
Level of amputation (% of transfemoral)	239	74.6±8.1	197	62.9±8.2	0.113	13
Oswestry Disability Index score (out of 100)	48	27.4±22.4	36	6.2±8.2	0.108	3
Pain level (out of 10)	97	3.4±1.8	80	0.6±0.9	<0.001	8
Biomechanical variables during gait
Preferred walking speed (m/s)	38	1.1±0.2	35	1.2±0.2	0.446	3
Trunk (thorax) sagittal RoM (°)	34	4.0±1.1	46	3.4±1.0	0.102	3
Trunk (thorax) frontal RoM (°)	34	7.3±2.5	46	5.4±1.7	0.244	3
Trunk (thorax) transverse RoM (°)	43	9.6±2.5	54	7.8±2.4	0.232	4
Pelvis sagittal RoM (°)	46	6.6±2.3	57	4.5±1.8	0.089	5
Pelvis frontal RoM (°)	46	7.9±2.7	57	7.1±2.5	0.070	5
Pelvis transverse RoM (°)	55	8.7±3.4	65	9.7±3.6	0.293	5

LBP: low back pain; BMI: body mass index; SD: standard deviation; RoM: range of motion

Figure 2:

Forest plots depicting the group comparisons of personal and anthropometric characteristics

3.2. Comparison of Participants’ Trunk and Pelvic Biomechanics during Gait

In the included studies, biomechanical characteristics were quantified during various activities including gait,^{20,31-34,37,38} sitting,³⁶ and during clinical assessment.^4,33,39,40 Due to our criterion of including variables that were examined and reported by at least three studies, our pooled analyses were focused on 5 of the 7 studies that examined trunk and pelvic kinematics during gait from which the original data were available (Table 4).^{20,31,32,37,38}

The testing procedures used to conduct gait analysis varied among these studies. Three studies allowed the participants to walk at a self-selected speed, which varied from 1.03 to 1.3 m/s^20,32,38 while 2 other studies used prescribed walking speeds from 1.0 to 1.4 m/s.^31,37 The number of walking trials collected ranged from 3 to 24. Overground walking was used in all of the included gait studies. The more commonly reported trunk/pelvic kinematics variables were peak-to-peak ranges of motion of the upper trunk (thorax) and pelvis, modeled as rigid bodies with reference to the global coordinate system (3 of the 5 gait studies). Discrepancies existed in thoracic spine marker placement. For example, 2 studies defined the thorax using markers placed on C7 and T8 spinal levels,^31,32 while 1 study used C7 and T10.³⁷

One study defined the spine in 3 segments (cervical, thoracic, lumbar) using markers at C1, C7, T3, and L1 spinal levels, and marker triads placed superficial to C5, T7, and L3 spinous processes.²⁰ The spinal kinematics from this study were excluded from the meta-analysis due to the fundamental difference in modeling approach to other studies that used two-segment models, however pelvic kinematic data were retained.

Our analyses showed that individuals with LLA and LBP exhibited generally greater trunk and pelvic ranges of motion (5 out of 6 variables) when compared to those without LBP. However, none of the biomechanical variables in our pooled analyses reached statistical significance (Table 4 and Figure 3).

Figure 3:

Forest plots of selected biomechanical variables

4. Discussion

To our knowledge, this systematic review and meta-analysis is currently the most comprehensive examination of published studies that directly compare demographic, anthropometric, clinical, and biomechanical characteristics between individuals with LLA, with and without LBP. This type of comparative research is important for understanding clinical and biomechanical presentations of LBP following LLA. Our pooled analyses encompassed available data from 436 participants in the 13 included studies (239 with LBP and 197 without). Our results show that individuals with LBP in general report modest levels of pain (3.4 out of 10). However, there were no other statistically significant differences observed between the LBP and No-LBP groups in terms of demographics, anthropometrics, and presence of co-morbidities. Our meta-analyses of the biomechanical variables did not detect any statistically significant difference in trunk and pelvic kinematic variables between the groups.

Our use of a validated checklist for assessing methodological quality afforded us a unique opportunity to examine gaps and weaknesses in this field of research. The median quality score of 37.5% demonstrated that there certainly is room for improvement. Specifically, few studies reported participation rate, estimated statistical measures of association, or blinding of investigators to participants’ LBP status. On the other hand, most of the included studies, particularly the more recent ones, used previously suggested standardized criteria to determine participants’ LBP status including duration, frequency, and severity.⁵ We also noticed a lack of studies with longitudinal design. The study by Mahon et al. was the only one that followed a group of individuals with LLA over a period of one year. Future prospective studies are needed to adequately describe the natural history of developing LBP after amputation, and how therapeutic interventions may alter the course of disease progression.⁶

The pooled data enabled unique analyses that have not previously been possible. We found that age, height, mass, and BMI were not significantly different between participants with and without LBP. This corroborated the survey study results from Devan et al.⁵ that these factors are not predictive of LBP status in individual with traumatic LLA. However, this finding is in contrast with the non-LLA population, where greater BMI is frequently observed in individuals with LBP.^41-43 This result may be in part due to the high representation of relatively younger military personnel and Veterans in the included studies. Our meta-analysis also demonstrated that longer time since amputation may be associated with the presence of LBP. Among the 11 studies that reported time since amputation (n=369), the mean durations for both groups were over 10 years (13.3 vs 11.9 years, LBP vs. No-LBP, respectively) and it should be noted that only two included studies examined LBP in participants who were on average within 5 years of their amputations.^31,32 While the time since amputation as reported by the reviewed studies did not represent the timing of LBP diagnosis, this finding may still be indicative of how prevalence of secondary musculoskeletal issues typically increases over time following amputation.⁴⁴ In a study of young military service members with LLA, Farrokhi et al.,⁴⁴ reported the incidence of new LBP development to be as high as 29% within the first year after limb loss, indicating a direct relationship between LLA, prosthesis use, and the development of LBP. Given that it is generally expected that the rates of LBP will increase with continued prosthesis use, the observation of a slightly greater time since amputation in our cohort with LBP is not unexpected. This further highlights the importance to examine the chronicity of LBP development and progression in this population.

It has long been suspected that the altered gait pattern secondary to amputation and prosthesis use could be a predominant cause of developing LBP after amputation, particularly in those with higher levels of LLA.^38,40 Over the last two decades, evidence in clinical biomechanics research regarding this theory has accumulated, which enabled our meta-analysis. We found that individuals with LLA and LBP exhibit greater trunk and pelvis RoM during gait in 5 of the 6 variables in comparison to those with no LBP. However, none of the apparent differences reached statistical significance and only trunk frontal and transverse RoM, and pelvic sagittal RoM exceeded the minimal detectable change values established in non-LLA individuals.⁴⁵ The small differences in these kinematic variables may be due to that walking utilize only a small proportion of the available spinal/pelvic RoM.⁴⁶ Given the high variability of spinal/pelvic motion in individuals with different levels of LLA and prosthetic designs,²⁰ caution should be exercised when interpreting functional relevance of miniscule spinal angular displacement differences in this population⁴⁷ and the generalizability of these patterns as related to LBP.²⁸ To further examine the findings in our meta-analyses of biomechanical variables, we found substantial heterogeneity across studies in the findings for frontal and transverse plane motion in the trunk (I² values of 68% and 76% respectively). This heterogeneity may in part be due to differences in gait testing and analysis protocols, including the varying spinal kinematic modeling techniques, which we will discuss further below.

When examining the gait analysis protocols used by the reviewed studies, we found that spatiotemporal gait parameters such as step and stride length and step width were almost never reported. In non-LLA populations, individuals with LBP demonstrate both reduced preferred walking speed and reduced stride length.²⁹ Including measurement of step/stride length and width, and step width asymmetry in future studies may help paint a more complete picture of the association between gait variations and LBP in individuals with LLA. In regards to the modelling of spinal biomechanics during gait in the five studies that reported these variables,^{20,31,32,37,38} we found that the analysis protocols generally were lacking in detail and, although arguably sufficient for the purposes of understanding each study, complicated the interpretation of our pooled data. While marker placement was described, coordinate system definition was mentioned in few studies. The effect of using T8^31,32 versus T10³⁷ in the definition of the trunk is unclear without knowledge of the coordinate system definitions, which were not reported. Similarly, the extent of any discrepancies that may arise from using a posteriorly-mounted marker triad for definition and tracking³⁸ in comparison to anterior and posterior markers on the sternum and spine used in other studies, is not known. Differences between ranges of motion observed across studies may partially be explained by inconsistencies in the tri-axial rotation sequences applied, which differed across at least two of the included studies.^20,38 Although the rotation order may have a negligible effect when used to compare data from non-pathological cohorts where ranges of motion are small,⁴⁵ it is unclear whether this assumption holds for individuals with LLA who sometimes exhibit increased ranges of trunk and pelvic motion. The angles reported about the three orthogonal axes are interlinked by nature of the angle calculation, in which the angle about each of the axes is computed sequentially as the segment is rotated about the first axis, then the second, then the third. Theoretically, a true difference in range about the first axis could result in an observed difference about the second and third axes that is purely artefactual. Lastly, the use of segment optimization procedures that can alter the effect of soft tissue artefact⁴⁸ was not reported in any of the reviewed articles. For future studies comparing biomechanics during walking and other functional tasks, we recommend providing access to detailed biomechanical methods including marker placement, coordinate system definition, segmental tracking markers, rotation sequences, and use of segment optimization to enable accurate and consistent kinematic assessment in this area of investigation. An example may be found in the supplementary information of a previous biomechanical study of individuals with LLA.⁴⁹

Although the etiology of LBP after LLA has historically been attributed to alterations in biological factors (e.g., gait and spinal biomechanics, muscle strength/flexibility, body weight, etc.), it has been recently suggested that inclusion of psychological and social assessments will further allow identification of additional intervention targets.⁷ In general, it has been reported that individuals with limb loss have high levels of psychosocial health concerns, which require holistic care approaches for improving quality of life outcomes.⁵⁰ To this end, as an individual with LLA reintegrates within the community, psychological factors may affect the risk for LBP development and its eventual chronicity. For example, chronic and recurrent LBP have been previously associated with more severe depressive symptoms, post-traumatic stress disorders, and anxiety in military service members with traumatic LLA.^35,51 Similarly, development of secondary overuse musculoskeletal injury such as LBP after LLA also appear to be related to social and personal factors such as tobacco use and alcohol consumption, which appear to be prevalent in this population.⁵²

It is critically important to understand how psychosocial factors may interact with biological factors in the development and persistence of LBP following LLA and to recognize the extent to which the relative contribution of psychosocial and biological factors to LBP may vary across individuals. To demonstrate the clinical efficacy of a comprehensive intervention in managing patients with amputation, a previous study showed that significant gait and posture improvements can be achieved in patients with transfemoral LLA after a combined intervention strategy that included psychological and physiotherapeutic treatments.⁵³ Additionally, psychologically informed practices that address the behavioral aspects of pain (i.e., peoples’ responses to pain) by identifying individual expectations, beliefs, and feelings as prognostic factors for clinical and occupational outcomes have also been successfully implemented within the chronic pain populations that warrant consideration in patients with LLA.⁵⁴

Despite the growing evidence supporting the positive effects of biopsychosocial treatment approaches for managing LBP with regards to pain, disability, and health-related quality of life,⁵⁵ their utilization in management of LBP in patients with LLA remains largely overlooked, which necessitates further research with an emphasis on clinical translation and implementation.

In recognition that the development and persistence of LBP is a multifactorial phenomenon, research agencies such as the National Institute of Health have begun to promote research standards and recommendations for a minimum dataset that should be reported in all LBP studies.^56,57 This minimum dataset encompasses work status, education level and comorbidities as well as self-report measures of physical function, depression, quality of sleep and pain catastrophizing. Although some studies in this review included measures of psychosocial characteristics, the lack of standardized assessment precluded synthesis of these variables. We recommend implementation of the minimum dataset approach in future studies of individuals with LLA and LBP. Adopting this research standard will promote a more comprehensive understanding of contributory factors to LBP after LLA and will facilitate meta-analyses. Importantly, none of the studies included in the present review reported if participants had a history of LBP prior to their LLA. In individuals without LLA, previous LBP is one of the strongest predictors of future LBP.⁵⁸ It is currently not clear to what extent this is also true in individuals with LLA. Additionally, we were not able to determine if there is a higher risk of LBP in individuals who are experiencing residual limb or phantom limb pain. In individuals with one or both comorbid chronic pain conditions, LBP may be more associated with sensitization and reorganization of the central nervous system than with biomechanical factors.⁵⁹ Factors related to prosthesis fit and use pattern may also be relevant to the development of LBP. For example, changes in coupling between the prosthesis and residual limb,^60,61 the availability and configuration of swing and stance phase knee control mechanisms including microprocessor control,^62-65 and alterations in limb and prosthetic alignment^66,67 have all been reported to result in measurable differences in gait kinematics and kinetics. Although it is plausible that increases in trunk and pelvic RoM may be responsible for the onset of LBP, it seems equally likely that underlying factors that lead to the changes in movement pattern are culpable. Further detail on prosthetic fitting, suspension, and components alongside measures such as socket fit comfort score may facilitate the identification of further factors related to the development of LBP.

The limited prevalence of LLA has always been a challenge for studying this clinical population, especially for laboratory-based biomechanical studies. This is evident in the studies included in this review, of them fewer than half (6 out of 13, 46%) enrolled more than 10 participants in both the LBP and No-LBP groups. At least one of the included studies expressed concerns about being underpowered.²⁰ Furthermore, the sampling of the current studies appears to be biased toward participants who were younger (mean age=47), male (46.1% of the studies enrolled males only), and with traumatic cause of amputation (69.2%). In fact, our pooled analysis showed that male participants accounted for 89.7 and 93% of the participants with and without LBP, respectively. These biases may be due to the fact that many of the included studies were focused on military personnel or Veterans. While it is true that the prevalence of amputation is approximately two times greater in males,⁶⁸ the disproportionately greater male representation in this line of research is still not justifiable. These biases represent a challenge in generalizing the study findings to females and older adults with LLA due to dysvascular causes, which constitute a large proportion of individuals with LLA in the general population.⁶⁸ As an example, Smith et al. reported higher back pain intensity in individuals with LLA who were older.⁶⁹ Using methods such as purposive sampling in place of convenience samples has been suggested for enrolling a more diverse and representative range of participants in future studies.²⁰

5. Conclusion

The findings of this systematic review and meta-analysis suggest that when compared to individuals with LLA without LBP, individuals with LLA and LBP experienced moderately higher levels of pain and slightly greater but non-significant differences in time since amputation and trunk and pelvic RoM during gait. However, the current body of evidence is limited by convenient sampling of predominantly younger and male participants with traumatic amputation. Methodological issues such as inconsistent gait testing procedure and biomechanical modeling of the spine/pelvis, as well as a general lack of examining psychosocial factors, hindered the current understanding of LBP in individuals with LLA. Prospective research focusing on the chronicity of LBP development following LLA will provide insight into the development and progression of LBP in this population.

Highlights:

1. Lower back pain is a common complaint associated with lower limb amputation.

2. We reviewed studies that compared persons with amputation with and without pain.

3. The average reported pain level is moderate (3.4 out of 10).

4. No between-group differences in age, BMI, comorbidity, and trunk/pelvic kinematics.

5. Determining lower back pain risk may require more comprehensive assessment.