The phantom paradox: Protective and risk factors for phantom limb pain following major lower extremity amputation

Abstract

Background

Approximately 50% to 80% of individuals undergoing amputation experience some form of phantom limb syndrome (PLS), a debilitating complication following major amputations that significantly diminishes patient quality of life, particularly PLS with pain (PLP). Yet, comprehensive studies exploring potential risk factors remain scarce. This large retrospective study identified major risk factors for PLP to help guide preoperative risk stratification and perioperative care strategies.

Methods

A national healthcare database was utilized to identify adults undergoing major lower extremity amputation (LEA) between 2012 and 2024. Patients were propensity score-matched across 31 demographic, clinical, and surgical variables. 18 distinct exposure cohorts, categorized into demographic, metabolic/vascular, substance use/medication history, and psychosocial/surgical factors, were analyzed to assess their associated risk for PLP within 1-year following LEA.

Results

Across 18 comparisons (n ≈ 12,000–88,000 per cohort), significant risk factors associated with increased PLP included preoperative limb pain (RR: 1.386), active tobacco use (RR: 1.372), hereditary and idiopathic neuropathy (RR: 1.261), peripheral arterial disease (RR: 1.117), and preoperative infection (RR: 1.160). Factors associated with reduced PLP risk included dementia (RR: 0.470), advanced age ≥65 (RR: 0.697), end-stage renal disease (RR: 0.771), and diabetes mellitus (RR: 0.790).

Conclusions

This study identified preoperative pain, neuropathy, tobacco use, peripheral arterial disease, and infection as risk factors, and dementia, dialysis, older age, diabetes, and above-knee amputation as protective against PLP. These findings support evidence-based preoperative risk stratification, targeted counseling, and optimization of modifiable risk factors. Prospective studies are needed to validate these associations, informing clinical prediction tools and targeted interventions aimed at improving patient outcomes.

Keywords

phantom limb pain risk factors protective factors lower extremity amputation

Introduction

Major lower extremity amputations (LEA) have an estimated incidence 185,000 new procedures performed annually with approximately 2.1 million individuals living with a major LEA in the United States,^1,2 a prevalence projected to rise significantly in the coming decades.^3,4 Among major amputations, lower extremity (LE) (above-knee (AKA) and below-knee (BKA), proximal to the ankle) procedures are the most frequently performed.⁵

A challenging postoperative complication following LEA is phantom limb syndrome, encompassing phantom limb syndrome without pain and phantom limb syndrome with pain (PLP). PLP, a chronic neuropathic condition, manifests as perceived pain originating from the amputated limb, often described as shooting or stabbing sensations. It affects an estimated 50% to 80% of major limb amputees,^6–9 while nearly all individuals experience some form of phantom limb sensation (PLS) post-amputation.¹⁰ Although the precise mechanisms underlying PLP remain incompletely understood, current hypotheses center on traumatic alterations to peripheral and central nervous system axons, leading to aberrant afferent signaling and subsequent maladaptive cortical reorganization.¹¹

Despite the emergence of over 50 distinct treatments for phantom limb pain (PLP) in recent decades, most lack robust clinical evidence or have demonstrated only modest success.^6,12,13 Given the substantial impact of PLP on patient quality of life, healthcare resource utilization, and rehabilitation outcomes, a critical clinical priority is the early identification of patients at greatest risk.^14,15 Accurate risk stratification not only supports personalized surgical planning and efficient use of healthcare resources but also informs targeted preventive strategies and more effective preoperative patient counseling.

While prior studies have explored individual risk factors in isolation, to the best of our knowledge no investigation has systematically evaluated a comprehensive range of demographic characteristics, metabolic and vascular comorbidities, substance use patterns, psychosocial and psychiatric factors, and surgical variables within a nationally representative cohort of sufficient scale to robustly identify predictors of PLP. Many existing investigations are further limited by small sample sizes and inconsistent approaches to confounder control, underscoring the need for large-scale, methodologically rigorous analyses to reliably predict PLP risk.

Therefore, this study aims to comprehensively investigate and identify demographic, clinical, psychosocial, and surgical factors associated with increased risk for phantom limb syndrome following LE amputations. Using rigorous propensity score matching and advanced statistical analysis of a large, nationally representative cohort, we seek to provide clinicians with actionable tools for preoperative PLP risk stratification. This evidence-based approach will ultimately guide targeted counseling for high-risk patients, optimize clinical decision-making regarding potential prophylactic interventions, and identify modifiable targets to improve long-term quality of life following major LE amputation.

Methods

Data source

This study utilized the TriNetX Research Network (https://trinetx.com; Cambridge, MA, USA), one of the largest federated health data platforms. This network integrates electronic medical records from over 100 healthcare organizations (HCOs) across the United States, Canada, and Western Europe, encompassing more than 150 million de-identified patient records. Additional patient-level insights were supplemented with claims data from over 100 commercial and government payers, including Medicare.^16,17

Study design

Adults aged 18 years and older who underwent a major LEA between 2012 and 2024 were identified, specifically those below-the-knee (CPT code 27880) or above-the-knee (CPT code 27590) proximal to the ankle. These inclusion criteria were applied uniformly across all analyses, except in the direct comparison of below-the-knee versus above-the-knee amputation for PLP risk. Foot and ankle amputations, hip disarticulations, and other minor or revision procedures were excluded.

From this patient population, we evaluated 18 candidate risk factors. Risk factor selection was guided by a combination of prior literature suggesting potential associations with phantom limb pain and biological plausibility based on known pathophysiologic mechanisms, including nociceptive priming, neuroplasticity, and ischemic neuropathy. In addition to reviewing prior literature and biological plausibility, we performed an initial series of univariate logistic regression models in the full amputation cohort to screen a broad set of demographic and clinical variables for association with the outcome (PLP). Variables that showed statistically significant or borderline associations with PLP, or were judged clinically important a priori, were considered for inclusion. From this set, 18 candidate factors were selected to define the exposed versus unexposed cohorts described above.

For each risk factor, two cohorts were established: one cohort included patients with the specific risk factor present in addition to either amputation CPT code, while the other cohort consisted of patients without the respective risk factor. For demographic factors (obesity with BMI ≥30 and advanced age ≥65) and metabolic/vascular comorbidities (Charcot neuroarthropathy, diabetes mellitus, hereditary and idiopathic neuropathy, peripheral arterial disease (PAD), and pre-operative limb pain, cellulitis, preoperative infection, end-stage renal disease + dialysis treatment, and foot ulcers), patients in the exposure cohorts were required to have documented diagnoses of these potential risk factors within 6 months preceding the amputation procedure, ensuring accurate reflection of the patient’s exposure at the time of surgery. For psychosocial and psychiatric variables (depression, anxiety, PTSD, and dementia), diagnoses were required within 1 week prior to surgery to reflect immediate preoperative status. Control groups were composed of patients without documented exposure to the respective variable during the defined window.

Primary outcome

The primary outcome, phantom limb syndrome with pain, was assessed 1 year postoperatively for all cohort comparisons following propensity matching. Patients were identified using the validated ICD-10 diagnostic code G54.6 (“Phantom limb syndrome with pain”).

Propensity-score matching

Propensity-score matching is a statistical technique used to reduce bias in observational studies by creating matched sets of subjects with similar covariates. This methodology involves computing propensity scores using logistic regression based on selected covariates, which predict the probability of a patient being in a particular cohort.¹⁸ In the current study, to appropriately select covariates for matching, an initial regression analysis was conducted to identify factors associated with both the exposure variable and the outcome of interest.

These 18 pairs of cohorts (one pair for each candidate risk factor) were then propensity matched based on the following variables: age, sex, race, active tobacco use, diabetes mellitus, Hemoglobin A1c levels in blood (stratified into three categories: ≤7%, 7.01-8.49%, ≥8.50%), overweight and obesity, peripheral vascular diseases, opioid dependence, chronic kidney disease (CKD), depressive episodes, anxiety disorder, Post-traumatic stress disorder (PTSD), hereditary and idiopathic neuropathy, pre-operative limb pain, essential (primary) hypertension, hyperlipidemia, congestive heart failure, cerebrovascular diseases, chronic obstructive pulmonary disease (COPD), problems related to care provider dependency, corticosteroid use, and alcohol abuse. However, when a variable under evaluation was included in the matching criteria, it was removed from the matching list (e.g., diabetes was removed from matching criteria when evaluating diabetes as a risk factor, obesity was removed when evaluating obesity, age was excluded from matching when assessing advanced age as a risk factor, etc.).

Covariate balance was quantified post-matching using standardized mean differences (SMD), with SMD <0.1 indicating adequate balance. This threshold was achieved for all variables in all 14 analyses. Detailed cohort characteristics before and after propensity score matching, matching diagnostics (including density plots), and full outcome statistics are available in the Supplemental File.

Statistical analysis

Risk ratios (RR) with 95% confidence intervals and absolute risk differences were calculated to compare PLP incidence between cohorts. Categorical variables were analyzed using Fisher’s exact or Chi-square tests, while continuous variables were assessed with Student's t-tests. Statistical significance was defined as p < 0.05 (two-tailed).

Software used for statistical analysis, validation, and data visualization

Data extraction utilized the TriNetX Live platform, with subsequent analysis and visualization performed using R (R Software, Vienna, Austria) and Microsoft Excel (Version 16.0, Microsoft Corp).

Results

Following propensity-score matching, several significant predictors for PLP were identified. Comprehensive statistical details for each risk factor are presented in Table 1, with a clear visual representation provided by the accompanying forest plot (Figure 1).

Table 1.

Risk factors for phantom limb pain following lower extremity amputation.

Variables evaluated	Exposure (+)	Control (−)	Risk ratio	95% CI	p
Demographics
Geriatric (≥65 Years vs 18–64)	1,097 (12.4%)	1,504 (17.8%)	0.697	(0.648, 0.748)	<0.05**
Obesity (BMI ≥30)	1,179 (16.5%)	1,185 (16.6%)	0.991	(0.920, 1.066)	0.80
Metabolic and vascular comorbidities
Charcot neuroarthropathy	164 (14.2%)	179 (15.5%)	0.919	(0.755, 1.117)	0.39
Diabetes mellitus	661 (11.4%)	833 (14.5%)	0.790	(0.718, 0.869)	<0.05**
Hereditary and idiopathic neuropathy	738 (20.1%)	610 (15.9%)	1.261	(1.145, 1.390)	<0.05**
Peripheral arterial disease (PAD)	1,140 (13.9%)	1,036 (12.4%)	1.117	(1.033, 1.208)	<0.05*
Pre-operative limb pain	999 (20.9%)	775 (15.0%)	1.386	(1.273, 1.509)	<0.05**
Pre-operative infection	568 (18.6%)	511 (16.0%)	1.160	(1.040, 1.293)	<0.05*
Foot ulcer	860 (13.7%)	874 (14.0%)	0.978	(0.896, 1.068)	0.62
Cellulitis	1,244 (13.7%)	1,341 (14.5%)	0.944	(0.879, 1.014)	0.11
End state renal disease on dialysis	297 (11.2%)	386 (14.5%)	0.771	(0.670, 0.889)	<0.05**
Substance use/Medication history
Opioid dependence	175 (24.5%)	167 (22.0%)	1.115	(0.926, 1.343)	0.25
Active tobacco use	427 (22.0%)	319 (16.0%)	1.372	(1.204, 1.564)	<0.05**
Psychosocial/Psychiatric factors
Depression	762 (17.2%)	768 (16.3%)	1.050	(0.958, 1.150)	0.30
Anxiety	655 (18.6%)	679 (17.6%)	1.055	(0.957, 1.163)	0.28
PTSD	64 (18.9%)	91 (22.6%)	0.836	(0.629, 1.112)	0.22
Dementia	63 (4.9%)	133 (10.5%)	0.470	(0.352, 0.628)	<0.05**
Level of amputation
Above knee vs below knee	1,525 (13.3%)	1,697 (14.4%)	0.922	(0.865, 0.983)	<0.05

Exposure (+) and Control (−) columns show the number and percentage of patients in each cohort who were diagnosed with phantom limb pain (ICD-10 G54.6) within 1 year after amputation. Risk ratio (RR) compares the incidence of PLP between the exposure and matched control cohorts. Significant values are bolded (p < 0.05). * = p < 0.01, ** = p < 0.001.

Figure 1.

Forest plot of risk factors for phantom limb pain.

Significant risk factors

Following propensity-score matching, all exposure and control cohorts showed excellent covariate balance (standardized mean difference <0.1 for all variables; Supplemental Table 1). Several preoperative and surgical characteristics were associated with increased PLP risk (Table 1, Figure 1). The strongest associations were observed for pre-operative limb pain and hereditary/idiopathic neuropathy, with more modest risk elevations for active tobacco use, pre-operative infection, and peripheral arterial disease.

Significant protective factors

In contrast, dementia, end-stage renal disease on dialysis, diabetes mellitus, age ≥65 years, and AKA (relative to BKA) were associated with lower 1-year PLP risk (Table 1). Chronic opioid dependence, obesity, Charcot neuroarthropathy, foot ulcers, cellulitis, depression, anxiety, and PTSD were not significantly associated with PLP after adjustment.

Non-significant associations

No statistically significant associations were observed for obesity, Charcot neuroarthropathy, chronic opioid use, depression, anxiety, PTSD, foot ulcers, or cellulitis.

Discussion

In this large, multicenter cohort of adults undergoing major lower-extremity amputation, we identified several patient- and procedure-level characteristics associated with 1-year PLP. Pre-operative limb pain, hereditary/idiopathic peripheral neuropathy, active tobacco use, pre-operative infection, and peripheral arterial disease were associated with an increased risk of PLP, whereas dementia, end-stage renal disease on dialysis, diabetes mellitus, age ≥65 years, and above-knee amputation were associated with a lower risk. Chronic opioid dependence, obesity, Charcot neuroarthropathy, depression, anxiety, PTSD, foot ulcer, and cellulitis were not significantly associated with PLP after adjustment. These findings add to existing evidence and may help clinicians stratify PLP risk and prioritize resources for higher-risk patients.

Pre-operative limb pain showed the largest relative increase in PLP risk. This aligns with prior work by Richardson et al.'s who demonstrated that higher levels of pre-amputation pain are associated with more frequent and severe PLP,⁶ supporting the concept that ongoing nociceptive input may prime the nervous system for persistent pain after amputation.^19–21 This phenomenon aligns with Richardson et al.'s observations that pre-amputation pain intensity directly correlates with phantom pain severity, likely mediated through sustained hyperexcitability in dorsal horn neurons and thalamocortical pathways.²¹ While pre-operative limb pain is not completely modifiable, particularly in patients with chronic ischemia or longstanding neuropathy, our results reinforce the importance of early recognition and optimization of pain management before surgery when possible, using multimodal analgesia and referral to pain specialists for complex cases.

The significant association between hereditary and idiopathic neuropathy and increased PLP risk further emphasizes the role of pre-existing neural pathology. Chronic nerve damage induces neural sensitization within both peripheral nociceptors and central pain pathways, lowering the threshold for aberrant signaling originating from the residual limb.^22–24 This heightened sensitization state is known to amplify maladaptive cortical reorganization following amputation, hence plausibly facilitating the development of PLP as observed in this study.²⁵ Importantly, this finding specifically pertains to hereditary and idiopathic neuropathies, as diabetes mellitus was explicitly retained as a matching variable, effectively controlling for diabetes-induced neuropathy and enabling clear differentiation between diabetic and non-diabetic neuropathy subtypes. This matching strategy allowed us to accurately isolate and evaluate the independent effects of hereditary and idiopathic neuropathies on PLP risk.

The significantly increased PLP risk in patients with a preoperative infection may result from interconnected pathophysiological mechanisms. Infectious states induce robust inflammatory cascades, releasing proinflammatory cytokines and neurotoxic substances that sensitize peripheral nociceptors and spinal dorsal horn neurons, priming neural pathways for subsequent neuropathic pain after amputation.^1,3,21 Additionally, certain chronic infections, particularly osteomyelitis or procedure-related infections, may involve direct bacterial infiltration and inflammation-related nerve injury, potentially causing structural nerve damage and ectopic impulse generation, thus facilitating maladaptive cortical reorganization.^3,21

Active tobacco use was another modifiable factor associated with elevated PLP risk. Mechanistically, nicotine-induced vasoconstriction and endothelial dysfunction exacerbate microvascular compromise in peripheral nerves.^26–28 In patients with vascular comorbidities, who constitute most amputee populations as detailed in the baseline patient characteristics in the supplemental file, this chronic ischemic insult potentiates axonal damage and impairs neural recovery, creating a neuropathic milieu primed for maladaptive signaling post-amputation. Our findings align with Lans et al.’s cohort analysis identifying smoking as an independent predictor of post-amputation neuropathic pain,²⁹ and support reinforcing screening for tobacco use and offering cessation interventions prior to amputation, especially when surgery can be scheduled electively.

Similarly, peripheral arterial disease (PAD) was also associated with a modestly higher PLP risk. Mechanistically, similar to tobacco use, the chronic ischemia inherent to PAD is associated with peripheral nerve hypoxia and damage. This ischemic neuropathy may therefore create a vulnerable neural substrate where the additional trauma of amputation could more readily trigger aberrant afferent signaling and subsequent maladaptive central reorganization, contributing to PLP development.^29–31 Additionally, substantial PAD may remain proximal to the level of amputation, potentially contributing to an element of ischemic pain.

A notable and initially counterintuitive finding was the significant protective association between end-stage renal disease (ESRD) patients on dialysis and PLP. Given that diabetes mellitus and glycemic control (HbA1c levels) were controlled through propensity matching, the observed protective effect is unlikely to be simply attributed to differences in diabetic neuropathy prevalence. Instead, a possible explanation is that chronic uremia in ESRD patients may induce profound generalized neuropathy and peripheral nerve dysfunction independent of diabetes severity.^1,6 Such extensive pre-existing nerve dysfunction could lead to neuropathic preconditioning, substantially limiting the capacity for additional aberrant nerve signaling following amputation.^14,20 Additionally, altered pain perception or reporting due to longstanding uremia and dialysis-related comorbidities may further contribute to lower reported incidence of PLP.

Interestingly, diabetes mellitus conferred significant protection against PLP development. This association persisted despite adjustment for HbA1c levels, diabetic complications, and relevant comorbidities. Two mechanisms may underlie this finding. First, severe pre-amputation diabetic neuropathy, distinct from the hereditary and idiopathic neuropathies analyzed separately, may provide neuropathic preconditioning: extensive peripheral and central deafferentation could reduce the additional aberrant signaling triggered by surgical transection, thereby dampening maladaptive central responses.³² Second, diabetes-related impairments in nerve regeneration, including diminished axonal regrowth and increased Schwann cell apoptosis, may limit neuroma formation and the initiation of neuropathic pain 3,32. These pathophysiological alterations align with clinical observations by Richardson et al. and Lans et al.,^3,6 who similarly reported reduced neuropathic pain in diabetic amputees, suggesting fundamental diabetes-mediated recalibration of pain processing pathways.

Another notable finding was that advanced age (≥65 years) demonstrated substantial protection against PLP, compared to younger patients. This finding aligns with established principles of age-related neuroplasticity constraints. Younger brains exhibit heightened cortical reorganization capacity that, while beneficial for learning, may facilitate maladaptive somatosensory remapping following deafferentation.^33,34 Age-related reductions in long-term potentiation and GABAergic cortical inhibition likely further constrain phantom percept generation,^35,36 though potential underreporting in elderly populations warrants consideration.

Preoperative dementia was also associated with a markedly reduced PLP risk relative to patients without dementia. This protective effect may be plausible given the core features of dementia including significant impairments in sensory processing, integration, memory, and communication. Neurodegeneration in these patients may disrupt integrative sensory processing, impairing the cognitive-perceptual frameworks necessary for phantom limb awareness.^37,38 Moreover, pathological changes in parietal and prefrontal cortices may also preclude the cortical remapping fundamental to PLP.³⁸ While underreporting due to communication deficits cannot be excluded, the magnitude of effect suggests potential inherent neurobiological mechanisms in PLP development.

Level of amputation is often constrained by vascular status and soft tissue viability but can sometimes be influenced by surgical planning. In our analysis, above-knee amputation was associated with a slightly lower risk of PLP compared with below-knee amputation. One possible explanation is that below-knee amputation preserves more distal nerve tissue, with a larger surface area for neuroma formation and aberrant signaling, whereas more proximal amputation may reduce the relative burden of symptomatic neuromas. This anatomical rationale is corroborated by Mioton et al.’s observations of higher residual limb pain prevalence following BKA.³⁹ These findings hold relevance for surgical decision-making when level selection is flexible, however, the magnitude of this association was modest, and decisions about amputation level should remain primarily driven by considerations of wound healing, function, and prosthetic potential rather than PLP risk alone.

Contrary to some prior reports^1,2,10,11 and surprising to us, our analysis demonstrated no significant association between PLP and depression, anxiety, PTSD, or chronic preoperative opioid use. This divergence may reflect methodological distinctions inherent to our study design. As the largest investigation of PLP risk factors to our knowledge, our nationally representative cohort provided large statistical power to detect associations. Comprehensive propensity-score matching across 27 demographic, clinical, and psychosocial variables, including socioeconomic proxies, enabled isolation of exposure effects compared to smaller cohorts. Within this rigorously controlled analytical framework, psychological factors did not independently predict PLP incidence. This suggests prior reported associations may reflect residual confounding or greater relevance to PLP chronicity than initial syndrome development. Similarly, neither obesity nor Charcot neuroarthropathy demonstrated significant associations with PLP risk. In some studies, variables that are not statistically significant may potentially be clinically meaningful. Consequently, we do not minimize the impact of anxiety, depression and PTSD in outcomes after LEA despite the statistics. Nevertheless, the complexity of PLP pathogenesis warrants further investigation through prospective longitudinal cohorts with granular phenotyping of psychological states, randomized trials of preemptive interventions targeting identified risk factors, and mechanistic studies exploring neurobiological interactions between protective variables (e.g., diabetes, aging) and pain pathways.⁴⁰

Notably, emerging techniques such as targeted muscle reinnervation (TMR) and regenerative peripheral nerve interfaces (RPNI) show promise by actively managing transected nerves at the time of amputation or in the postoperative period. These approaches have demonstrated potential to reduce phantom limb pain, decrease opioid reliance, and enhance prosthetic integration.^14,15 In parallel, a variety of non-surgical interventions, including pharmacologic therapies, neuromodulation, and behavioral approaches, are commonly employed in PLP management, although comparative effectiveness data across treatment modalities remain limited and heterogeneous. However, implementation of advanced surgical interventions such as TMR and RPNI requires substantial multidisciplinary resources, including access to specialized orthopaedic and peripheral nerve surgeons, and may increase procedural complexity and cost.

Accordingly, the present study was not designed to evaluate the effectiveness of specific PLP treatments, but rather to identify patient- and procedure-level factors associated with PLP development. The identification of these risk factors provides a critical foundation for risk-informed allocation of resource-intensive interventions such as TMR or RPNI. Prioritizing these advanced techniques for the highest-risk patients may optimize both clinical outcomes and healthcare efficiency. Importantly, these findings also establish a framework for future prospective studies that assess the effectiveness of available PLP treatments within well-defined, risk-stratified patient populations.

Clinically, these findings provide a base for evidence-backed preoperative risk stratification. Patients with significant pre-amputation pain, hereditary and idiopathic neuropathy, PAD, or active tobacco use, especially those undergoing BKA, constitute a higher-risk cohort who may benefit from targeted preoperative counseling and preemptive multimodal analgesia, with smoking cessation and pain optimization representing modifiable targets perioperatively. Conversely, the protective effects observed with age ≥65 years, diabetes, dementia, and AKA help delineate lower-risk profiles.

Limitations and strengths

Several limitations inherent to retrospective analyses of electronic health records warrant consideration. First, identification of PLP was based on the ICD-10 diagnostic code G54.6, which lacks granularity regarding symptom severity and does not distinguish variations in pain intensity. Although comprehensive propensity matching was performed, residual confounding from unmeasured factors, such as specific surgical techniques (e.g., nerve-handling protocols), rehabilitation intensity, or detailed psychosocial determinants, may remain. In addition, although phantom limb pain may fluctuate or develop beyond 1 year following amputation, the 1-year postoperative assessment window was selected to capture clinically established PLP rather than transient early postoperative symptoms, while balancing diagnostic reliability and follow-up completeness in a large EHR-based cohort. Additionally, our analysis relied on diagnosis and procedure codes recorded in electronic health records, which are subject to underreporting and misclassification. Pre-operative limb pain and psychiatric conditions, in particular, may be incompletely documented and thus underestimated. Medication exposure, including chronic opioid use, may also be under captured when prescriptions are filled outside participating health systems. Such misclassification would likely bias observed associations toward the null. Finally, as with all observational studies, causal inferences cannot be definitively established.

Nevertheless, this study possesses notable methodological strengths. The healthcare database platform provided unprecedented access to a geographically diverse, nationally representative cohort, enhancing generalizability beyond single-center studies. By focusing on major lower-extremity amputations (BKA/AKA), we reduced some of the clinical heterogeneity present in prior work that combined upper- and lower-limb procedures or minor amputations. Comprehensive assessment of demographic, clinical, psychosocial, and surgical variables through rigorous propensity-score matching achieved exceptional covariate balance (standardized mean difference <0.1 for all variables), substantially reducing selection bias. Precise temporal definitions of exposures relative to surgery mitigated protopathic bias, while the cohort scale afforded statistical power to detect clinically meaningful associations often unattainable in smaller studies.

Conclusions

In this large, multicenter cohort of patients undergoing major lower-extremity amputation, pre-operative limb pain, hereditary and idiopathic peripheral neuropathy, active tobacco use, peripheral arterial disease, and pre-operative infection were associated with modestly increased risk of PLP, whereas dementia, end-stage renal disease on dialysis, diabetes mellitus, age ≥65 years, and above-knee (vs below-knee) amputation were associated with lower observed PLP risk. These findings provide a crucial foundation for evidence-based preoperative risk stratification, enabling targeted counseling, optimization of modifiable risks (pain, smoking), informed surgical level selection, and resource allocation of advanced prophylactic interventions to high-risk patients. Prospective studies with detailed phenotyping and longer follow-up are needed to confirm these findings, clarify underlying mechanisms, and develop validated prediction tools for PLP risk after amputation.

Supplemental material

Supplemental Material - The phantom paradox: Protective and risk factors for phantom limb pain following major lower extremity amputation

Supplemental Material for The phantom paradox: Protective and risk factors for phantom limb pain following major lower extremity amputation by Sri Tummala, Brandon A Wood, Mehul M Mittal, Senthil N Sambandam, Dane K Wukich in Journal of Orthopaedic Surgery

Footnotes

ORCID iDs

Sri Tummala

Mehul M Mittal

Consent for publication

This retrospective study is exempt from informed consent. The data reviewed represent a secondary analysis of existing data, do not involve intervention or interaction with human subjects, and are de-identified per the de-identification standard defined in Section §164.514(a) of the HIPAA Privacy Rule. The process by which the data are de-identified is attested to through a formal determination by a qualified expert as defined in Section §164.514(b) (1) of the HIPAA Privacy Rule. This formal determination by a qualified expert was refreshed on December 2020.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

The data that support the findings of this study are available from TriNetX. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from with the permission from TriNetX.

Location of work performed

All work for this manuscript was performed on the premises of University of Texas Southwestern Medical School (5323 Harry Hines Blvd, Dallas, TX, USA, 75390).

Institutional review board statement

This study was exempt from the UT Southwestern IRB approval since the data were de-identified and publicly available.

Supplemental material

Supplemental material for this article is available online.

References

Hogan

Anderson

Kovoor

, et al. Phantom limb syndrome: assessment of psychiatric and medical comorbidities associated with phantom pain in 44,028 below knee amputees. Injury 2022; 53(11): 3697–3701. https://doi.org/10.1016/j.injury.2022.09.018

Darnall

Ephraim

Wegener

, et al. Depressive symptoms and mental health service utilization among persons with limb loss: results of a national survey. Arch Phys Med Rehabil 2005; 86(4): 650–658. https://doi.org/10.1016/j.apmr.2004.10.028

Lans

Groot

Hazewinkel

MHJ

, et al. Factors related to neuropathic pain following lower extremity amputation. Plast Reconstr Surg 2022; 150(2): 446–455. https://doi.org/10.1097/PRS.0000000000009334

Ziegler-Graham

MacKenzie

Ephraim

, et al. Estimating the prevalence of limb loss in the United States: 2005 to 2050. Arch Phys Med Rehabil 2008; 89(3): 422–429. https://doi.org/10.1016/j.apmr.2007.11.005

Mioton

Dumanian

Fracol

, et al. Benchmarking residual limb pain and phantom limb pain in amputees through a patient-reported outcomes survey. Plast Reconstr Surg - Glob Open 2020; 8: 1–7. https://doi.org/10.1097/GOX.0000000000002977

Richardson

Glenn

Horgan

, et al. A prospective study of factors associated with the presence of phantom limb pain six months after major lower limb amputation in patients with peripheral vascular disease [published correction appears in J Pain. 2007 Dec; 8(12):998]. J Pain 2007; 8(10): 793–801. https://doi.org/10.1016/j.jpain.2007.05.007

Dijkstra

Geertzen

JHB

Stewart

, et al. van der Schans CP: phantom pain and risk factors: a multivariate analysis. J Pain Sympt Manage 2002; 24: 578–585. https://doi.org/10.1016/s0885-3924(02)00538-9

Kooijman

Dijkstra

Geertzen

JHB

, et al. van der Schans CP: phantom pain and phantom sensations in upper limb amputees: an epidemiological study. Pain 2000; 87: 33–41. https://doi.org/10.1016/s0304-3959(00)00264-5

Richardson

Glenn

Nurmikko

, et al. Incidence of phantom phenomena including phantom limb pain 6 months after major lower limb amputation in patients with peripheral vascular disease. Clin J Pain 2006; 22: 353–358. https://doi.org/10.1097/01.ajp.0000177793.01415.bd

10.

Fuchs

Flor

Bekrater-Bodmann

. Psychological factors associated with phantom limb pain: a review of recent findings. Pain Res Manag 2018; 2018: 5080123. https://doi.org/10.1155/2018/5080123

11.

Padovani

Martins

MRI

Venâncio

, et al. Anxiety, depression and quality of life in individuals with phantom limb pain. Acta Ortop Bras 2015; 23(2): 107–110. https://doi.org/10.1590/1413-78522015230200990

12.

Halbert

Crotty

Cameron

. Evidence for the optimal management of acute and chronic phantom pain: a systematic review. Clinical J Pain 2002; 18: 84–92. https://doi.org/10.1097/00002508-200203000-00003

13.

Sherman

. What do we really know about phantom limb pain? Pain Rev 1994; 1: 261–274.

14.

Woo

Kung

Brown

, et al. Regenerative peripheral nerve interfaces for the treatment of postamputation neuroma pain: a pilot study. Plast Reconstr surgery Glob open 2016; 4(12): e1038. https://doi.org/10.1097/GOX.0000000000001038

15.

Dumanian

Potter

Mioton

, et al. Targeted muscle reinnervation treats neuroma and phantom pain in major limb amputees. Ann Surg 2018; 270(2): 1. https://doi.org/10.1097/sla.0000000000003088

16.

Palchuk

London

Perez-Rey

, et al. A global federated real-world data and analytics platform for research. JAMIA Open 2023; 6: ooad035. https://doi.org/10.1093/jamiaopen/ooad035

17.

TriNetX . Real world Data—Linked. Available online: https://trinetx.com/real-world-data/linked/ (Accessed Jun 10, 2025).

18.

TrinetX Help Center . In compare outcomes, how are patients matched when balancing cohorts? 2024. https://support.trinetx.com/hc/en-us/articles/360011978033-In-compare-outcomes-how-are-patients-matched-when-balancing-cohorts (Accessed 2024 Jun 10).

19.

Glare

Aubrey

Myles

. Transition from acute to chronic pain after surgery. Lancet 2019; 393(10180): 1537–1546. https://doi.org/10.1016/S0140-6736(19)30352-6

20.

Kandasamy

Price

. The pharmacology of nociceptor priming. Handb Exp Pharmacol 2015; 227: 15–37. https://doi.org/10.1007/978-3-662-46450-2_2

21.

Pace

Passavanti

De Nardis

, et al. Nociceptor plasticity: a closer look. J Cell Physiol 2018; 233(4): 2824–2838. https://doi.org/10.1002/jcp.25993

22.

Inquimbert

Bartels

Babaniyi

, et al. Peripheral nerve injury produces a sustained shift in the balance between glutamate release and uptake in the dorsal horn of the spinal cord. Pain 2012; 153(12): 2422–2431. https://doi.org/10.1016/j.pain.2012.08.011

23.

Blom

Pfister

Santello

, et al. Nerve injury-induced neuropathic pain causes disinhibition of the anterior cingulate cortex. J Neurosci 2014; 34(17): 5754–5764. https://doi.org/10.1523/JNEUROSCI.3667-13.2014

24.

Dubois

Poitras

Voisin

JIA

, et al. Effect of pain on deafferentation-induced modulation of somatosensory evoked potentials. PLoS One 2018; 13(10): e0206141. https://doi.org/10.1371/journal.pone.0206141

25.

Dijkstra

Geertzen

Stewart

, et al. Phantom pain and risk factors: a multivariate analysis. J Pain Symptom Manage 2002; 24(6): 578–585. https://doi.org/10.1016/s0885-3924(02)00538-9

26.

Rose

Krishnan-Sarin

Exil

, et al. Cardiopulmonary impact of electronic cigarettes and vaping products: a scientific statement from the American Heart Association. Circulation 2023; 148(8): 703–728. https://doi.org/10.1161/CIR.0000000000001160

27.

El-Mahdy

Abdelghany

Hemann

, et al. Chronic cigarette smoke exposure triggers a vicious cycle of leukocyte and endothelial-mediated oxidant stress that results in vascular dysfunction. Am J Physiol Heart Circ Physiol 2020; 319(1): H51–H65. https://doi.org/10.1152/ajpheart.00657.2019

28.

Maiuolo

Gliozzi

Musolino

, et al. The role of endothelial dysfunction in peripheral blood nerve barrier: molecular mechanisms and pathophysiological implications. Int J Mol Sci 2019; 20(12): 3022. https://doi.org/10.3390/ijms20123022

29.

Smolderen

Ujueta

Buckley Behan

, et al. Understanding the pain experience and treatment considerations along the spectrum of peripheral artery disease: a scientific statement from the American Heart Association. Circ Cardiovasc Qual Outcomes 2025; 18(3): e000135. https://doi.org/10.1161/HCQ.0000000000000135

30.

Weinberg

Simovic

Isner

, et al. Chronic ischemic monomelic neuropathy from critical limb ischemia. Neurology 2001; 57(6): 1008–1012. https://doi.org/10.1212/wnl.57.6.1008

31.

Nukada

van Rij

Packer

, et al. Pathology of acute and chronic ischaemic neuropathy in atherosclerotic peripheral vascular disease. Brain 1996; 119(Pt 5): 1449–1460. https://doi.org/10.1093/brain/119.5.1449

32.

Stenberg

Dahlin

. Gender differences in nerve regeneration after sciatic nerve injury and repair in healthy and in type 2 diabetic Goto-Kakizaki rats. BMC Neurosci 2014; 15(1): 1–10. https://doi.org/10.1186/1471-2202-15-107

33.

Mowery

Garraghty

. Adult neuroplasticity employs developmental mechanisms. Front Syst Neurosci 2023; 16: 1086680. https://doi.org/10.3389/fnsys.2022.1086680

34.

Hübener

Bonhoeffer

. Neuronal plasticity: beyond the critical period. Cell 2014; 159(4): 727–737. https://doi.org/10.1016/j.cell.2014.10.035

35.

Ding

Zheng

Liu

, et al. Changes in GABAergic markers accompany degradation of neuronal function in the primary visual cortex of senescent rats. Sci Rep 2017; 7(1): 14897. https://doi.org/10.1038/s41598-017-15006-3

36.

Eavri

Shepherd

Welsh

, et al. Interneuron simplification and loss of structural plasticity as markers of aging-related functional decline. J Neurosci 2018; 38(39): 8421–8432. https://doi.org/10.1523/JNEUROSCI.0808-18.2018

37.

Evered

Scott

Silbert

. Cognitive decline associated with anesthesia and surgery in the elderly: does this contribute to dementia prevalence? Curr Opin Psychiatry 2017; 30(3): 220–226. https://doi.org/10.1097/YCO.0000000000000321

38.

Larbig

Andoh

Huse

, et al. Pre- and postoperative predictors of phantom limb pain. Neurosci Lett 2019; 702: 44–50. https://doi.org/10.1016/j.neulet.2018.11.044

39.

Mioton

Dumanian

Fracol

, et al. Benchmarking residual limb pain and phantom limb pain in amputees through a patient-reported outcomes survey. Plast Reconstr Surg Glob Open 2020; 8(7): e2977. https://doi.org/10.1097/GOX.0000000000002977

40.

Fournier-Farley

Boudier-Revéret

Gagnon

. Frequency and localization of neuromas in transtibial residual limbs. Arch Phys Med Rehabil 2024; 105(10): 1993–1996. https://doi.org/10.1016/j.apmr.2024.07.004