LOW BACK PAIN IN ADULTS WITH TRANSFEMORAL AMPUTATION: A RETROSPECTIVE POPULATION-BASED STUDY

Abstract

Background:

Low back pain (LBP) is common among individuals with transfemoral amputation (TFA) and has a negative impact on quality of life. Little is known about healthcare utilization for LBP among this population and whether utilization varies by amputation etiology.

Objective

To determine if individuals with TFA have an increased likelihood of seeking care or reporting symptoms of acute or chronic LBP during physician visits post-amputation compared to matched individuals without amputation.

Design

Retrospective Cohort

Setting

Olmsted County, Minnesota (2010 population: 144,248).

Participants

All individuals with incident TFA (N=96), knee disarticulation and transfemoral amputation, residing in Olmsted County between 1987 and 2014. Each was matched (1:10 ratio) with non-TFA adults on age, sex, and duration of residency. Individuals were divided by etiology of amputation: dysvascular and trauma/cancer.

Interventions

Not applicable.

Main Outcome Measures

Death and presentation for evaluation of LBP (LBP event) while residing in Olmsted County. LBP events were identified using validated ICD-9 codes and corresponding Berkson, HICDA, and ICD-10 diagnostic codes. Hurdle and competing risk Cox proportional hazard models were used.

Results:

Having a TFA of either etiology did appear to correlate with increased frequency of LBP events, though this association was only statistically significant within the dysvascular TFA-cohort (dysvascular TFA-cohort: RR 1.80, 95% CI 1.07 – 3.03, median follow-up 0.78 years; trauma/cancer TFA cohort RR 1.14, 95% CI 0.58 – 2.22, median follow-up 7.95 years). In time to event analysis, dysvascular TFA had an increased risk of death and event. Obesity did not significantly correlate with increased frequency of LBP events or time to event for either cohort. At any given point in time, individuals with TFA of either etiology who had phantom limb pain were 90% more likely to have a LBP event (HR 1.91, 95% CI 1.11 – 3.31). Conditional on not dying and no LBP event within the first 2.5 years, individuals with prosthesis had a decreased risk of LBP events in subsequent years.

Conclusions:

Risk of LBP events appears to vary by TFA etiology. Obesity did not significantly correlate with increased frequency of LBP event or time to event. Phantom limb pain correlated with decreased time to LBP event post amputation. The association between prosthesis receipt and LBP events is ambiguous.

INTRODUCTION

Low back pain (LBP) affects 52–89% of those who undergo lower extremity amputation [1–5] compared to an estimated mean global prevalence of 31%[6] and can have an even greater impact on quality of life than phantom limb pain [2]. In 2005, there were 1.6 million Americans living with limb loss, most commonly due to diabetes and peripheral arterial disease [7, 8], with the lower extremity amputation rate reaching 5 per 1,000 adult patients with diabetes per year in 2014 [9, 10]. Due in part to the aging population and an increase in prevalence of those living with diabetes, the number of Americans with an amputation is projected to double by the year 2050 [8, 11]. Reported risk factors for affecting the presence and intensity of LBP that are unique to individuals with amputation include musculoskeletal imbalance secondary to altered gait mechanics and postural changes, deconditioning, poor prosthetic fit, leg-length discrepancy, amputation level, multiple comorbidities, and presence of phantom limb or residual pain [12–17]. Results are mixed however the majority of studies report that individuals with TFA tend to have a higher prevalence of LBP compared to those with transtibial amputation [1, 2, 4, 5].

Between 1996 and 2013, diabetes and the combination of low back and neck pain accounted for the first and third, respectively, highest health care spending in the U.S. and together were the conditions with the largest increase in health care spending over the 18 year time period [18]. While individuals with amputation, specifically TFA, have reported higher prevalence of low back pain compared to those without, little is known about the pattern of healthcare utilization for back pain specific to this population and whether utilization varies by amputation etiology. The purpose of this study was to perform a population-based longitudinal evaluation to determine if individuals with TFA have an increased likelihood of seeking care for acute or chronic LBP or of reporting LBP during a physician visit (LBP event) as compared to matched individuals without amputation. These findings will help to identify if there is increased healthcare utilization for LBP among the amputee population, which may lead to further study on improving health care delivery to this high risk and growing population. It may also help to identify a subset of the amputee population at higher risk as well as additional risk factors that providers and patients should be aware of and if modifiable, address early after amputation. We hypothesized that individuals with TFA, especially those of dysvascular etiology, have an earlier and higher frequency of LBP events compared to the general population.

METHODS

Data Source and Study Population

Individuals with TFA residing in Olmsted County, MN, were identified using the Rochester Epidemiology Project (REP) resources. The REP is a population-based longitudinal dataset of the pooled medical records within Olmsted County health care providers including Mayo Clinic, Olmsted Medical Center, and their affiliates [19]. The REP includes all individuals regardless of setting (community-dwelling, assisted living, long-term care etc.). The Olmsted County population is similar to that of the Upper Midwest but is less diverse, wealthier, and more highly educated than the general U.S. population, yet results have been found to be generalizable [20]. Using the resources of the REP, incident patients (patients who had an amputation while residing in Olmsted County) were identified using the ICD-9 diagnostic and procedure codes for amputations (84.17 for a TFA procedure or V49.76 if an individual had a TFA). Each individual with TFA was matched (1:10 ratio) with non-TFA adults on age, sex, and duration of residency in Olmsted County. Patients who had denied research authorization were excluded. This study was approved by both the Mayo Clinic and Olmsted Medical Center Institutional Review Boards.

Medical records for individuals with TFA were reviewed to confirm amputation status and level. Additional data obtained included gender, race, amputation etiology (dysvascular or trauma/cancer), year of amputation (index date), pre and post-amputation comorbidities, and receipt of prosthesis. Comorbidities were extracted from administrative data and classified using modified Charlson comorbidities via the icd9 package in R[21–22]. The outcome of interest was whether an individual presented for evaluation of LBP or reported symptoms of acute or chronic LBP (LBP event) while residing in Olmsted County. LBP events were identified using previously validated ICD-9 codes and corresponding Berkson, HICDA, and ICD-10 diagnostic codes [23]. Any LBP event within 30 days of the previous event was not counted in the analysis in order to avoid overestimating by capturing follow-up encounters. Sixty and 120 day intervals were tested and results did not change.

Statistical Analysis

Individuals were divided into two cohorts by etiology of amputation: dysvascular disease versus trauma/cancer. Two separate modeling techniques were used to explore the relationship between having a TFA and the time from amputation to presentation for LBP (LBP event) and subsequent frequency of visits to a health care provider for LBP. A competing risk Cox regression model was used to evaluate time to first presentation with LBP at a physician visit due to the high mortality rates among individuals with TFA. By using this model, the higher risk of mortality in individuals with TFA was considered and the baseline hazard rate was allowed to vary. This is important in the analysis of progressive conditions that take time to develop such as LBP, when an individual is at a higher risk of developing a comorbid condition (in this case death) that would preclude its development [24]. Not accounting for this additional risk factor could over-estimate risk of LBP events. An early and late event indicator for both dysvascular and trauma/cancer cohorts was created with the early and late time period corresponding to the first 2.5 and 5 years after index date, respectively. These respective specifications were used to ensure that the proportional hazard condition was met throughout all follow-up time and was determined empirically from raw survival curves for each group. To examine the frequency at which an individual sought care related to LBP, a Hurdle model was used to accounted for the fact that some individuals never seek medical care for LBP [25]. Due to wide variation in follow-up time, an offset accounting for follow-up time was included in the model. Additionally, for issues of convergence, follow-up was capped at 30 years. Both cohorts were analyzed separately for time to first LBP related physician visit and frequency of LBP related visits. Effect of prosthesis was evaluated among all individuals with TFA, not divided by etiology due to the small sample size of the trauma/cancer cohort. Variable selection for each model was performed using boosting. This is similar to the more traditional stepwise regression where each variable’s correlation with the model is evaluated one at a time, but reduces the likelihood of spurious variable inclusion by evaluating the variable’s predictive power as opposed to a statistical significance cutoff value and offers superior predictive models [26]. Statistical significance was reported by calculating 95% confidence intervals. All statistical analysis was conducted using R version 3.3.2 [27].

RESULTS:

Characteristics of Study Sample

The study population included 96 individuals with TFA: 84 with amputation due to dysvascular etiology and 12 secondary to trauma or cancer. Mean index age was 76.4 ± 11.3 years for those with dysvascular TFA and 48.6 ± 20.8 years for those with TFA due to trauma or cancer (p < .001). Those with dysvascular amputations had a significantly higher mortality rate at one and five years compared to those with TFA secondary to trauma/cancer (p < .001, Table 1). Individuals with dysvascular TFA had significantly higher Charlson Comorbidity Indices and were significantly less likely to receive a prosthesis (Table 1).

Table 1.

Characteristics of study sample, TFA cohort and matched non-TFA cohort

	Dysvascular			Trauma or cancer			p value, Dysvascular vs Trauma/Cancer
	TFA (N=84)	Non-TFA (N=840)	p value	TFA (N=12)	Non-TFA (N=120)	p value
Female N, (%)	43 (51.2)	430 (51.2)	-	3 (25)	30 (25)	-	.17
Mean index age (SD)	76.4 (11.2)	76.4 (11.3)	-	48.6 (20.0)	48.6 (20.8)	-	<.001
Mortality, mean years after index date (SD)	2.7 (4.2)	7.5 (6.2)	<.001	10.5 (7.8)	7.3 (6.2)	.11	<.001
Mortality rate, %, one year after index date (SD)	56 (50)	6 (25)	<.001	8 (29)	1 (10)	.04	.09
Mortality rate, %, five years after index date (SD)	85 (36)	28 (45)	<.001	17 (39)	7 (25)	.22	.002
Charlson Comorbidity Index, mean (SD)	6.5 (2.7)	3.9 (2.4)	<.001	1.7 (2.6)	1.2 (1.9)	.42	<.001
Obesity %	29	24	.39	8	11	.79	.14
PLP %	14	-	-	50	-	-	.003
Received prosthesis, N (%)	17 (20)			7 (58)			.004
LBP event prior to index date (%)	54	54	.97	25	28	.85	.07
Percent (SD) with at least one LBP event post index date	31 (47)	52 (50)	<.001	58 (51)	40 (49)	.22	-
Median follow-up, years (IQR)	0.8 (0.06–3.5)	7.9 (4.1–14.3)	-	8.0 (5.4–19.5)	11.1 (7.1–21.8)	-	-
Age at first LBP event, years (SD)	65.4 (13.1)	68.6 (14.8)	.11	47.0 (17.3)	51.0 (19.6)	.58	.001
Number of LBP events prior to index date, mean (SD)	1.4 (2.6)	1.3 (3.3)	.88	0.2 (0.6)	0.5 (1.6)	.45	.11
Number of LBP events post index date, mean (SD)	0.8 (1.8)	2.2 (4.6)	.008	1.8 (3.4)	1.1 (2.7)	.41	.15

PLP = Phantom Limb Pain.

Over a median follow-up of 0.78 years (IQR 0.06 – 3.5; range 0.0–24.3), 31% (N=26) of individuals with dysvascular TFA had at least one LBP event. Fifty-two percent (N=437) of the matched non-TFA cohort had at least one event over a median follow-up of 7.94 years (IQR 4.1 – 14.3; range 0.01–30.3). Over a median follow-up of 7.95 years (IQR 5.4 – 19.5; range 0.25 – 29.7), 58% (N=7) of individuals with TFA due to trauma/cancer had at least one event. Forty percent (N=48) of the matched non-TFA cohort had at least one LBP event over a median follow-up of 11.12 years (IQR 7.1–21.8; range 0.24 – 29.7) (Table 1).

Frequency of LBP Events

Having a TFA of either etiology did appear to correlate with increased frequency of LBP events, though this association was only statistically significant within the dysvascular TFA-cohort (dysvascular TFA-cohort: RR 1.80, 95% CI 1.07 – 3.03; trauma/cancer TFA cohort RR 1.14, 95% CI 0.58 – 2.22; Table 2). Gender, index age, and Charlson Comorbidity Index did not independently correlate with frequency of LBP events for either cohort (Table 2). There was no significant difference in number of pre-amputation events between the dysvascular or trauma/cancer TFA cohorts or between the TFA cohorts and their respective non-TFA cohorts (Table 1). The frequency of LBP events prior to index date among both cohorts correlated with a slightly increased post-amputation frequency of events (Table 2).

Table 2.

Adjusted Frequency of LBP events

	Dysvascular etiology TFA cohort and matched controls			Trauma or Cancer TFA cohort and matched controls			All Individuals with TFA
	RR	95% CI		RR	95% CI		RR	95% CI
Individual with TFA vs. Controls	1.80	1.07	3.03	1.14	0.58	2.22	-	-	-
Female vs. male	1.06	0.84	1.34	1.19	0.78	1.82	1.13	0.63	2.02
White vs. non-white	1.09	0.81	1.47	1.05	0.58	1.88	1.25	0.63	2.49
Index Age (years)	1.01	0.99	1.02	1.00	0.98	1.02	0.99	0.96	1.01
Charlson Comorbidity Index (per point)	0.99	0.91	1.07	1.14	0.95	1.38	1.03	0.88	1.21
Trauma Cancer etiology vs.	-	-	-	-	-	-	0.27	0.12	0.63
Obese vs. non-Obese	0.97	0.76	1.23	1.16	0.68	1.96	-	-	-
Age first diagnosed as obese (years)	-	-	-	1.00	0.98	1.01	-	-	-
# of LBP events prior to index date	1.09	1.06	1.12	1.16	1.03	1.30	-	-	-

TFA = Transfemoral Amputation

RR = Relative Risk

CI = Confidence Interval

Time to LBP

Compared to the non-TFA cohort, at any given point in time, the dysvascular TFA cohort had a 9.7 fold increase in risk of death and 2.4 fold increase in the risk of a LBP event within the first 2.5 years after index date (Figure 1, Table 3). Risk of a LBP event among the trauma/cancer cohort were 3.6 times higher within the first 5 years of index date (Table 3). At any given point in time, individuals with dysvascular TFA and increased Charlson Comorbidity Index were 12% more likely to present for LBP while those with trauma/cancer TFA and increased Charlson Comorbidity Index were 21% more likely to have a LBP event (Table 3). At any given point in time, individuals with TFA of either etiology who had phantom limb pain were 90% more likely to have a LBP event (HR 1.91, 95% CI 1.11 – 3.31; Table 3).

Figure 1.

Probability of low back pain (LBP) or mortality in individuals with transfemoral amputation (TFA) due to dysvascular disease and 16 matched non‐TFA cohort

Table 3.

Adjusted Time to Incident LBP Event or Death.

	Dysvascular etiology TFA cohort			Trauma or Cancer TFA cohort			All Individuals with TFA
Individuals with TFA vs Matched Controls	HR	95% CI		HR	95% CI		HR	95% CI
Early Death, Individuals with TFA vs Matched Controls	9.71	6.77	13.94	5.03	1.00	25.24	-	-	-
Late Death, Individuals with TFA vs Matched Controls	2.36	1.04	5.40	1.90	0.24	14.97
Early LBP, Individuals with TFA vs Matched Controls	2.41	1.51	3.85	3.61	1.32	9.92	-	-	-
Late LBP, Individuals with TFA vs Matched Controls	1.57	0.64	3.86	1.82	0.41	8.15	-	-	-
Time to first LBP event
Female vs. male	0.96	0.84	1.11	1.17	0.66	2.05	2.15	1.37	3.39
Index Age (years)	1.03	1.02	1.04	1.01	0.99	1.03	1.01	0.99	1.03
White vs non-White	0.82	0.70	0.95	2.02	1.15	3.54	1.03	0.65	1.63
Charlson Comorbidity Index (per point)	1.12	1.08	1.16	1.21	1.01	1.46	1.11	1.01	1.23
Trauma Cancer Etiology	-	-	-	-	-	-	0.58	0.28	1.19
Obese vs. non-Obese	1.15	0.98	1.35	-	-	-	-	-	-
Age first diagnosed as obese (years)	-	-	-	0.99	0.98	1.00	-	-	-
# of LBP events prior to index date	1.10	1.07	1.12	1.19	1.04	1.36	-	-	-
Individuals prescribed a prosthesis vs not prosthesis
Early Death, Individuals prescribed a prosthesis vs not prosthesis	-	-	-	-	-	-	0.14	0.04	0.45
Early LBP, Individuals prescribed a prosthesis vs not prosthesis	-	-	-	-	-	-	2.00	0.88	4.54
Late Death, Individuals prescribed a prosthesis vs not prosthesis	-	-	-	-	-	-	0.14	0.01	1.36
Early LBP, Individuals prescribed a prosthesis vs not prosthesis	-	-	-	-	-	-	2.00	0.88	4.54
Late LBP, Individuals prescribed a prosthesis vs not prosthesis	-	-	-	-	-	-	0.06	0.01	0.62
Individuals with PLP vs. those without PLP	-	-	-	-	-	-	1.91	1.11	3.31

PLP = Phantom Limb Pain. TFA = Transfemoral Amputation.

Early Death/Early LBP: within 2.5 years for dysvascular cohort and All individuals with TFA cohort; and 10 years for Trauma/Cancer cohort.

Late Death/Late LBP: greater than 2.5 years for dysvascular cohort and All individuals with TFA cohort; and 10 years for Trauma/Cancer cohort.

HR = Hazard Ratio

CI = Confidence Interval

Prosthesis effect

Among individuals with TFA who received a prosthesis (N= 24), and conditional on not dying and no LBP event within the first 2.5 years of index date, there was a decreased risk of presenting for LBP afterwards (HR 0.06, 95% CI 0.01–0.62; Figure 2). Individuals with prosthesis were 84% less likely to die within the first 2.5 years of the index date compared to those who did not receive a prosthesis (Table 3). As previously reported, all individuals who received a prescription for prosthesis received one within the first seven months of amputation (prosthesis fitting occurred an average of 105 days post amputation) [28]. In sensitivity analysis, we eliminated the individuals who died within the first 7 months of index date (N = 41) and the protective effect of prosthesis on death hazard rate disappeared. This indicates that it is likely confounding by indication and illness severity and we are unable to explain with our current construct given that patients needed to be healthy enough to qualify for prosthesis receipt.

Figure 2.

Probability of LBP or mortality in individuals with TFA who received a prosthesis compared to individuals who did not receive a prosthesis.

DISCUSSION

This unique, population-based study suggests that LBP events that rise to the level of requiring medical attention in the form of physician office visits appear to vary by TFA etiology and are influenced by obesity and the presence of phantom limb pain. Time to first visit for LBP is not shortened as a result of undergoing TFA. Mortality rate among individuals with dysvascular TFA in our population was 56% at one year and 85% at five years, which is similar to reported ranges of 43–54% mortality at one year and 40–90% at five years [29–31]. Those with dysvascular amputations had a significantly higher mortality rate than those with TFA secondary to trauma/cancer, which is also consistent with the literature and contributed to shorter term follow-up.[31–35] Consistent with the literature, those with dysvascular amputation were also significantly older compared to those with traumatic amputation[4], which one may suspect would lead to the former having a higher likelihood of having LBP events. We found, however, that age of amputation did not independently correlate with frequency of events. Our findings are likely due in part to higher mortality rates.

This study demonstrates that among individuals with TFA of either dysvascular or trauma/cancer etiology, obesity was not associated an increased frequency of LBP and it did not correlate with decreased time to first visit for LBP. These findings differ from several studies of the general population, including one surveying individuals from nine countries, which reports that individuals who are obese have a significantly increased risk of developing LBP [36–40], are more likely to develop chronic LBP, [38–40] and are more likely to seek care for LBP [38].

Across all three model specifications, comorbidity burden did not appear to be independently correlated with LBP. This differs from findings by Devan et al in which individuals with nondysvascular TFA or transtibial amputations who had two or more comorbid conditions had increased odds of chronic LBP[14]. Studies of the general population also report that those with chronic LBP have a higher comorbidity burden compared to controls [40, 41]. Of note, we were unable to distinguish between acute, chronic, or nonspecific LBP, which may partly explain the conflicting results.

Regarding previous history of presentation for LBP, having an amputation of either etiology did not correlate with number of pre-amputation LBP events and there was no significant difference in number of pre-amputation events between the dysvascular or trauma/cancer TFA cohorts or their respective non-TFA cohorts. As one might expect, however, frequency of events prior to amputation did correlate with higher frequency of LBP events post-amputation for both cohorts. Phantom limb pain correlated with decreased time to LBP event post amputation. This supports previous studies in which individuals with phantom limb pain had an increased risk of also having LBP [14, 43]. Together, they suggest that addressing phantom limb pain early in the post-operative time period may be an important component in managing LBP.

The data in this study demonstrated that time to first visit is not shortened as a result of undergoing TFA. Further, the correlation of prosthesis receipt and LBP events is unclear. It appears that receipt of a prosthesis is protective against LBP events after 2.5 years of amputation. Stam et al. evaluated 240 individuals with TFA and prosthesis and found no relationship between LBP and time since amputation. They also did not find a correlation between LBP and level of physical activity [44]. In a study of 92 individuals with lower extremity amputation who used a prosthesis at least 5 days per week, there was no correlation between time since amputation and occurrence of LBP [2].

This study has several strengths. First, the unique population-based data facilitated by the REP and access to complete medical records of all study subjects over an extended time period makes it possible to identify TFA events and pre and post-operative comorbidities, which is typically not possible in other settings. Median follow-up for dysvascular subjects was 0.78 years, most likely due to higher mortality, and was nearly eight years for those with trauma/cancer TFA. As opposed to studies that use self-reported surveys, the data used in this study directly measured healthcare utilization in terms of physician visits related to LBP.

There are several potential limitations of this study. First, data on other reasons for physician visits were not obtained. Individuals may not seek care for LBP or not report it at visits. Second, data on daily use of prosthesis, prosthetic design, and the characteristics of the LBP (i.e. acute or chronic, presence of associated radiculopathy) were not available; future study of a larger population that addresses these factors would be beneficial. While the population of Olmsted County closely resembles that of the Upper Midwest, generalizability of our results to the U.S. population as a whole may be limited. It is also a relatively small cohort, which could limit the ability to find correlations between factors that may predispose individuals to LBP events. Nevertheless, our ability to collect accurate, validated exposure and outcome data from an entire community population over an extended time period is unique and the internal validity of this study is high.

CONCLUSION

This study suggests that risk of LBP events varies by TFA etiology. Phantom limb pain correlated with decreased time to LBP event post amputation. The association between prosthesis receipt and LBP events is ambiguous, but may be protective in the long term, or at least does not suggest that prosthetic restoration leads to more provider intervention for LBP.

Abbreviations:

LBP

low back pain

TFA

transfemoral amputation