Introduction
Optimizing surgical outcomes for severe peripheral nerve injuries remains a challenge for peripheral nerve surgeons. At regeneration rates of only 1–3 mm/day, the capacity of injured nerves to reinnervate target tissues, especially over long distances, is limited. Consequently, functional recovery of proximal injuries is often poor. Reference Seddon, Medawar and Smith1–Reference Sunderland3 The socioeconomic cost of peripheral nerve injuries is high, as these injuries are common in young otherwise healthy individuals. Reference Noble, Munro, Prasad and Midha4,Reference Midha5 Thus, identifying potentially modifiable factors to optimize surgical outcomes and improve time to surgical intervention for severe peripheral nerve injuries is of clinical importance.
The ideal timeline for surgical reconstruction of closed peripheral nerve injuries is not well established in the literature. While some groups advocate for very early surgical intervention within weeks of injury, others advocate waiting a minimum of three months to allow for the assessment of spontaneous recovery. Reference Birch6–Reference Jivan, Kumar, Wiberg and Kay8 Although timing of surgery has an important influence on clinical outcomes, most published studies have not taken other potential confounding factors into account. This represents an important knowledge gap. Similarly, there is very limited information in the literature regarding factors associated with delays to peripheral nerve reconstruction surgery. For example, treatment of initial injury at smaller hospitals was found to be significantly associated with surgical delays. Reference Dy, Baty, Saeed, Olsen and Osei9 Since nerve reconstruction surgery should not be performed beyond one year from injury, Reference Hoang, Chen and Seruya10 it is essential to delineate factors that could have a potential impact on the time to surgery within the first year of injury, as the majority of nerve reconstruction surgery is conducted within that time frame.
Therefore, the goal of this study is to determine the association of time to operative intervention, as well as patient and injury characteristics with Medical Research Council (MRC) strength outcomes after surgery. Reference Martin, Senders, DiRisio, Smith and Broekman11 To minimize risk of erroneous associations found in univariate analyses, a clustered multivariable logistic regression analysis was used to evaluate factors that could potentially impact or confound strength outcomes. Reference Sperandei12,Reference Zhang13 Additionally, we aimed to investigate patient and system factors that are associated with delays to operative intervention, utilizing a multivariable model to identify potentially modifiable factors that may be addressed to reduce delays in surgery.
Materials and Methods
We carried out a comprehensive retrospective cohort study using the provincial database of all adult patients in Alberta who underwent a nerve graft or transfer for treatment of a closed severe proximal (above the level of the elbow or knee) peripheral nerve injury between 2005 and 2018, in accordance with the STROBE guidelines. The patients were all treated with surgical intervention as the assessing surgical teams deemed neurological recovery impossible or very unlikely without surgical intervention based on their clinical assessments. The human research ethics boards at the University of Calgary and the University of Alberta approved this study.
Inclusion Criteria
All adult patients in Alberta with closed severe peripheral nerve injury requiring nerve surgery were treated in one of two regional programs. In Calgary, patient charts were identified using operative booking codes pertaining to peripheral nerve surgery and subsequently manually screened for those patients who underwent nerve transfer or graft. In Edmonton, patients who underwent nerve transfer and graft were identified directly using operative booking codes. All patients who underwent operative intervention had a preoperative MRC grade of 0 or 1, no meaningful clinical recovery, and documented severe denervation on electromyographic (EMG) study.
All patients who had a minimum of 12 months documented follow-up or demonstrated reinnervation resulting in MRC strength grade ≥ 3 prior to 12 months were included in the strength outcome analysis. Reinnervation following surgery was confirmed by EMG study. Pediatric patients, patients who required a primary free muscle-based reconstruction, patients able to undergo primary nerve repair, and those not undergoing surgical reconstruction of their peripheral nerve injury were excluded from analysis (Fig. 1).
Figure 1: Flow chart of Edmonton and Calgary patient identification.
Predictor Variables
The outcomes of interest for the respective statistical analyses were whether or not the patient achieved an MRC ≥ 3 at the last documented follow-up and time to surgical intervention. Demographic information recorded included characteristics of the patients, injuries sustained, and operative interventions performed (Table 1).
Table 1: Demographic information recorded included characteristics of the patients, injuries sustained, and operative interventions performed
Rural area was defined as centers with a population less than 30,000 based on the 2011 Canadian census data. Reference Martin, Senders, DiRisio, Smith and Broekman11 The patients’ driving distance was calculated using postal codes. A polytrauma, often sustained concurrently at the time of peripheral nerve injury, was documented if two or more moderate injuries were sustained in different anatomic regions or if they had an injury severity score ≥ 16. Reference Noble, Munro, Prasad and Midha4,Reference Pape, Lefering and Butcher14–Reference Butcher and Balogh16 The number of preinjury comorbidities was documented as a proxy of the overall preinjury health status. Reference Kim, Song, Lee, Han, Hyung and Cho17 This is a relevant consideration since patients with more chronic comorbidities were more likely to report difficulties accessing specialist care and had higher surgical complication rates. Reference Kim, Song, Lee, Han, Hyung and Cho17,Reference Harrington, Wilson, Rosenberg and Bell18 Alcohol use was defined as self-reported or documented alcohol use beyond that of the recommended daily limit. Reference Bondy, Rehm, Ashley, Walsh, Single and Room19 The patients’ preinjury employment was classified as intellectual, manual employment, or unemployed. Manual labor was defined as a physically demanding occupation. Intellectual labor was defined as occupational activities that involved mostly intellectual work, office work, or the pursuit of education. Reference Galobardes, Lynch and Smith20
Statistical Analysis
We carried out a descriptive analysis of patient characteristics, injury types, and time to operative intervention. The Jarque–Bera test was used to test for normality. Mean ± standard deviation was reported for parametric data and median and interquartile range (IQR) for non-parametric data. Univariate statistical comparisons of patient and injury characteristics were done using student T-testing for parametric data and Mann Whitney-U testing for non-parametric data. Chi-squared testing was used to compare frequency data between the two sites. Post hoc testing when necessary was conducted by utilizing the Bonferroni adjustment method.
For the clustered multivariate logistic regression model, the method of purposeful variable selection was used to ensure that all a priori identified clinically relevant variables were included. We used this to evaluate association of the identified variables with a binary MRC strength outcome (MRC < 3 or MRC ≥ 3) in addition to significant and confounding variables. Reference Sperandei12,Reference Zhang13,Reference Bursac, Gauss, Williams and Hosmer21 Using this method to account for the correlation of predictive factors within the same individual, patients who received multiple procedures at the same time could also be compared. Patients with missing data and insufficient follow-up time were eliminated from the multivariable logistic regression model.
Location of surgical site, age, time from injury to operation, and follow-up time after surgery were retained in the final model regardless of statistical significance as they are clinically relevant variables. Reference Pape, Lefering and Butcher14,Reference Wang, Rancy, Lee, Feinberg and Wolfe22 The type of injury the patient sustained confounded the effect of follow-up and age on MRC outcome; thus, type of injury was retained in the model. The following assumptions were evaluated to ensure logistic regression model validity: (i) influential value examination using Cook’s distance plots and values, (ii) variable inflation factor index to assess multicollinearity of the variables, (iii) a visual inspection of scatter plots for the logit transformation of the continuous variables to ensure the linearity assumption was met, (iv) a Hosmer–Lemeshow test for goodness of fit (p = 0.19). Statistical analysis was performed using STATA 15. As per statistical convention, p ≤ 0.05 was considered statistically significant.
A multivariable Cox proportional hazard regression model was chosen to evaluate the adjusted effect of Akaike Information Criterion (AIC) selected independent variables on time to operative intervention. Reference Akaike23 AIC variable selection allows for the selection of variables that balance model fit with the least number of parameters and avoids the bias and validity issues introduced by selecting variables based solely on p-values. Reference Steyerberg, Eijkemans and Habbema24 The final multivariable Cox model included 6 variables. Reference Wen, Zhang, Quan and Wang25 Patients with missing data in the final model were eliminated from the Cox proportional hazard regression analysis, resulting in the final model including 144 patients. The proportional hazard assumption was tested for the final model and each of the individual variables in the selected model using the “phtest” based on Schoenfeld residuals. Reference Grambsch26,Reference Therneau, Grambsch and Fleming27 Statistical analysis was performed on R version 3.4.0. Reference Wen, Zhang, Quan and Wang25,Reference Therneau, Grambsch and Fleming27
Results
Patient Demographics
There were 163 patients: 97 from Edmonton and 66 from Calgary who underwent a nerve graft or transfer for treatment of severe peripheral nerve injury between 2005 and 2018 identified that met the inclusion criteria (Fig. 1). The majority of patients who underwent surgery were male (81.8%) and the median age at time of surgery was 34 (IQR: 25) years. The time to surgical intervention ranged from 0 to 633 days after injury, 17 patients received their operation greater than 365 days from injury, and 13 received surgery within 2 weeks of injury. The characteristics of the patients undergoing surgery in Edmonton and Calgary were examined in a univariate analysis for significant differences between locations (Tables 2 and 3). The time to operative intervention was significantly longer in Edmonton than Calgary.
Table 2: Comparison of regional surgical site differences
Table 3: Comparison of regional surgical site patient and injury characteristics
Strength Analysis Outcomes
After those with missing data and insufficient follow-up time were eliminated from the multivariable logistic regression model, 129 patients who met the eligibility criteria were identified: 79 from Edmonton and 50 from Calgary (Fig. 1). A total of 186 procedures were performed with 39 patients undergoing multiple procedures. All nerve coaptations were performed using standard epineurial sutures and fibrin glue. The median time to surgery in this population was 212 (IQR:131) days. The mean length of post-surgical follow-up was 1.9 years. Details of the procedures, patient, and injury characteristics are summarized in Table 4.
Table 4: Distribution of patient and injury characteristics for MRC strength outcome
The age of those undergoing grafting was significantly younger than those who had nerve transfer: 31.5 ± 2.1 years versus 39.3 ± 1.2 years (p = 0.002). Additionally, the time to operative intervention in the group undergoing nerve grafting was significantly shorter (5.7 vs 7.5 months, p = 0.005), and these patients had significantly longer follow-up (26.6 vs 21.4 months, p = 0.005). All other demographic variables were nonsignificant.
The final model included 186 procedures and 129 patient clusters. The unadjusted beta coefficients and respective p-values for all potential explanatory variables are listed in Table 5. The final logistic regression model revealed that surgery location, patient age, and follow-up time were not associated with MRC strength outcome (Table 6).
Table 5: Unadjusted beta coefficients and respective P-values for all potential MRC outcome explanatory variables
Table 6: Final clustered multivariable logistic regression model for MRC strength outcome
Overall, 63% of patients achieved greater than antigravity strength outcome after surgery. In patients who received nerve transfer rather than nerve graft reconstruction, the estimated adjusted odds of achieving an MRC strength grade ≥ 3 increased by 388% (aOR = 4.88, p = 0.003). In patients who sustained pre-ganglionic injury, the estimated adjusted odds of the patient achieving an MRC strength grade ≥ 3 decreased by 65% (aOR = 0.35, p = 0.05). For every week of delay from injury to time of surgery, the estimated adjusted odds of the patient achieving an MRC strength grade ≥ 3 decreased by 3% (aOR = 0.97, p = 0.02). Thus, for an operative intervention with delay of 1, 3, or 6 months after injury the adjusted odds of a patient achieving greater than antigravity strength decreases by 22% (aOR = 0.88), 31% (aOR = 0.69), and 53% (aOR = 0.47), respectively. The adjusted odds of achieving greater than antigravity strength decreased by 50% if surgery was delayed 5.5 months (aOR = 0.50). These trends still held when cases with delays of greater than a year were excluded.
Factors Associated with Time to Peripheral Nerve Surgery
We identified three critical steps in the peripheral nerve injury management pathway prior to patients’ operative intervention: (i) referral source, (ii) an EMG study diagnosing and documenting the extent of injury, this was a necessary exam that must have been completed prior to the patient being assessed by the surgical team, and (iii) attending a surgical consultation appointment in the peripheral nerve clinic. The median time to a referral being made to the multidisciplinary peripheral nerve clinic was 66 (IQR: 116) days after injury, time to initial EMG testing before the patients could be triaged for the peripheral nerve clinic was 84 (IQR: 100) days, and the median time for initial consultation with the multidisciplinary team that consists of nerve surgeons, physiatrists, neurologists, and allied health was 135 (IQR: 111) days (Table 7).
Table 7: Summary of critical points in the peripheral nerve injury referral pathway
The unadjusted univariate hazard ratios for all potential independent variables examined are shown in Table 8. The final Cox proportional hazard regression model demonstrates an increase in preinjury comorbidities (any documented medical condition that was not a direct consequence of the peripheral nerve injury) was associated with longer overall time to surgery (adjusted harzard ratio (aHR) 0.84, 95% CI 0.74–0.95) and a referral to the peripheral nerve clinic made by a surgeon was associated with a shorter overall time to surgery as compared to a referral made by a general practitioner (aHR 1.87, 95% CI 1.14–3.06) (Table 9). Reference Akaike23 An increase of one medical comorbidity, regardless of the nature of the comorbidity, decreased the odds of surgery by 16%; after controlling for involvement of the dominant limb, where the patient lived, and referring specialty (p = 0.006). The expected odds of surgery for a patient referred by a surgeon is 1.87 times that a patient referred by a general practitioner (family physician), after controlling for the other variables in the model (p = 0.014). There is a nonsignificant trend toward patients living in an urban center having a 1.39 times greater odds of surgery as compared to those living in a rural center, when all other factors remain the same (p = 0.072).
Table 8: Univariate analysis of all potential independent variables Cox proportional hazard model
Table 9: Cox proportional hazard regression analysis results
Discussion
The principal finding in this study is that to achieve the best possible strength outcomes for patients with severe closed peripheral nerve injuries, operative intervention at the earliest possible time, while still allowing for assessment of spontaneous recovery by utilizing nerve transfers, when possible, is necessary. Improved functional outcomes are observed in patients who underwent surgery prior to 6 months from injury. Reference Flores28,Reference Ahmed-Labib, Golan and Jacques29 Indeed, in a recent systematic review of a heterogenous population of nerve injuries the median time to surgery was 4 months for patients achieving greater than anti-gravity strength, while 7 months elapsed prior to surgery for those who achieved less than anti-gravity strength. Reference Martin, Senders, DiRisio, Smith and Broekman11 A major reason is that over time there is decreased expression of regeneration-associated genes and viable Schwann cells as well as progressive fibrosis and proteoglycan scarring in the distal nerve stump; thus, the regenerative capacity of the peripheral nerve declines with time. Reference Chen, Yu and Strickland30–Reference Fu and Gordon32
Given that by 5.5 months after injury the adjusted odds of achieving greater than anti-gravity strength decreased by 50%, operative reconstruction should be undertaken as early as possible after spontaneous recovery is deemed improbable. However, since the determination of optimal time for surgical reconstruction requires surgical judgment and balancing the potential for spontaneous recovery particularly in closed traction injuries, the decision to operate must therefore be made on an individual basis. Reference Martin, Senders, DiRisio, Smith and Broekman11,Reference Giuffre, Kakar, Bishop, Spinner and Shin33,Reference Lim, Lee, Kim, Kim and Kwon34
A major strength of this study is the utilization of multivariate logistic regression analysis that controls for the effects of potential confounding factors. Even though it is well recognized that there are many influencing factors that could have an impact on the functional outcomes following nerve reconstructive surgery, to our knowledge, this has rarely been done in the literature. It represents an important omission as unnecessary delays in surgical intervention must be identified and mitigated. We demonstrate a greater number of comorbidities were significantly associated with an overall increased time to operative intervention. In contrast, a referral by a surgeon was significantly associated with an overall decreased time to operative intervention. An increased number of comorbidities have been associated with a longer time to operative intervention, independent of surgical specialty. Reference Vergara, Bilbao, Gonzalez, Escobar and Quintana35,Reference Kwon, Carey, Cook, Qiu and Paszat36 Therefore, to minimize perioperative risk, optimization of medically modifiable comorbidities by a multidisciplinary preanesthetic consultation team prior to surgery is critical. Reference Fong and Sweitzer37 Although increased age often correlates with increased comorbidities, age was not found to be a significant factor impacting time to operative intervention in the univariate analysis age. Reference Davis, Chung and Juarez38 This is likely due to the fact that the population who sustain severe peripheral nerve injury is young (median age 34 years). Thus, there is a systemic accessibility problem, in which those patients with a lower preinjury health status wait longer for surgery.
The association of a referral made by a surgeon with a shorter time to surgery has not been previously reported in the literature, this suggests accessibility is biased toward a patient population that already has access to a surgical specialist. Currently, there is no standardized referral system in Alberta. Thus, the referring physician decides what information is included, to whom and how the referral is sent. Incomplete referral letters have been associated with poor patient outcomes. Reference Jiwa, Arnet, Bulsara, Ee and Harwood39,40 It is possible that surgeons’ referrals contain more clinically pertinent and complete information allowing for more accurate triage of referral. Additionally, appropriate referral is reliant on the referring physicians’ knowledge of local resources, it may be that surgeons are more aware of other available surgical resources. Finally, it is likely surgeons and general practitioners have different experiences diagnosing peripheral nerve injuries as these injuries more commonly present to tertiary care centers. Reference Noble, Munro, Prasad and Midha4 Medical education to increase awareness of patterns of peripheral nerve injuries may serve to reduce diagnostic challenges and expedite referrals. The implementation of standardized referral process, enhancing clinician’s knowledge regarding resources available, and/or an increased emphasis of peripheral nerve injury diagnosis at medical education level may serve to alleviate the association between time to surgery and referral source. Reference Naseriasl, Adham and Janati41–Reference Tobin-Schnittger, O’Doherty, O’Connor and O’Regan43 Potential negative impacts of delays in surgical wait times are not unique to peripheral nerve injury. This may also apply to a wide range of diseases including prostate cancer treatment. Reference Zakaria, Couture and Nguyen44 To mitigate this, various strategies including close coordination with diagnostic facilities, Reference McKevitt, Dingee and Warburton45 and among different surgical specialties, have been proposed. Reference Wright and Menaker46 As these challenges are common to many clinical conditions, implementing system wide changes could potentially have widespread impacts through timely healthcare delivery and improve functional outcomes.
Limitations
Limitations of this study include its retrospective nature. We were constrained by the documentation contained within patient charts; thus, all possible confounders may not have been examined. Furthermore, we have combined the populations of two large academic centers with geographically different catchment areas, patients, injuries, and operative techniques. However, on balance, we felt that the utilization of provincial data is important as the findings reflect outcomes and practices in the entire province rather than just a single center.
In evaluating motor outcomes, we were unable to account for the effect of individual types of nerve transfers, graft lengths, and the distinct differences in functional outcomes of these procedures due to the concerns regarding model over-fitting. To address these would require conducting prospective multicentre studies in order to recruit a sufficiently large sample of patients to compare the effect of predictor variables within individual nerve transfer procedures. Additionally, MRC scoring is subjective, dependent on patient effort, does not offer a full functional assessment of motion, and there is inter-examiner variability in scoring. Reference Paternostro-Sluga, Grim-Stieger and Posch47 However, it should be noted all MRC grades were supported with correlating EMG studies. Ideally, a combination of patient-reported outcomes, quantitative force measurements, and range of motion data would comprise a more comprehensive outcome measure.
In order to assess the effect of location of surgery on the fit of the Cox proportional hazards model evaluating parameters affecting time to surgery, we compared the AIC values of the chosen model and the model including an additional location variable; the AIC scores differed by 0.06. When comparing models a lower AIC score indicates a model that better balances goodness of fit and number of parameters. An expert statistician deemed a difference in AIC scores of 0.06 negligible, thus including location in the model did not provide additional explanatory value over the chosen model.
Conclusion
The timing of operative intervention after a severe peripheral nerve injury, operative reconstructive technique, and whether or not a pre-ganglionic injury was sustained play an important role in post-surgical strength outcomes. These findings suggest that proximal injuries with long regeneration distances should be reconstructed as early as feasible and preferentially with nerve transfers when possible. Delays in operative intervention need to be mitigated in order to achieve optimal patient strength outcomes after peripheral nerve surgery. It is clear that the subspecialized nature and complex care pathway for the treatment of peripheral nerve injuries results in difficulties delivering timely surgical interventions. Reference Camp and Birch48 Based on these findings, we need to work on improving clinical pathways for peripheral nerve injury referrals and enhance awareness of injury patterns and clinical resources. We must also work to improve access to peripheral nerve surgery for patients living in small centers and those with a greater number of comorbidities.
Acknowledgments
We gratefully acknowledge funding support from CIHR and University of Alberta SPOR Support Unit.
Competing interests
None.
Statement of authorship
JCB – Manuscript preparation, data analysis.
AB – Manuscript preparation.
MJM – Manuscript preparation.
JLO – Manuscript preparation.
RTT – Manuscript preparation.
RM – Manuscript preparation.
CSMC – Manuscript preparation.
BW – Manuscript preparation.
KMC – Manuscript preparation, data analysis.
