Abstract
Introduction
The management of spinal neoplasia consists of surgical, radiation, and systemic options. Little data exist to guide management based on overall health status, which is particularly challenging when patients who could benefit from surgery may be too frail for it. This study's objective was to evaluate the 5-Item Modified Frailty Index (mFI-5) as a predictor of 30-day morbidity in patients undergoing instrumented resection for metastatic extradural spinal tumors.
Methods
Adults undergoing extradural tumor resection from the 2011 to 2019 National Surgical Quality Improvement Program datasets were identified by Current Procedural Terminology codes 63275-63278 with an adjunct instrumentation code (22840-22843). Patients were classified into frailty levels 0, 1, or 2+ based on mFI-5 scores of 0, 1, or 2-5, respectively. The primary outcome was morbidity. Secondary outcomes were readmission and reoperation. Multivariate modeling was utilized to analyze mFI-5 as a predictor of outcomes. The Akaike information criterion (AIC) was used to compare relative-model-fit based on frailty versus individual comorbidity variables to determine the optimal model. Chi-squared and Fisher's exact tests were used to establish significance between individual complications and frailty.
Results
There were 874 patients. Readmission, reoperation, and morbidity rates were 19.5%, 5.0%, 52.3%, respectively. In multivariate analyses, mFI-5=1 (OR: 1.45, SE: 0.31, p=0.036), mFI-5=2+ (OR: 1.41, SE: 0.40, p=0.036), operative time (OR: 1.18, SE: 0.03, p≤0.001), and chronic steroid use (OR: 1.56, SE: 0.42, p=0.037) independently predicted morbidity. Elective surgery (OR: 0.61) was protective. Frailty did not predict readmission or reoperation. Frailty was found to be significantly associated with respiratory complications, urinary tract infections, cardiac events, and sepsis/septic shock specifically.
Conclusions
mFI-5=1 independently predicted 45% increased odds of morbidity. mFI-5=2+ independently predicted 41% increased odds of morbidity. Further, every 30 additional minutes of operative time predicted 18% increased odds of morbidity, suggesting an increased risk of site-related complication events. Taken together, the mFI-5 serves as a valid predictor of morbidity in patients with extradural tumor undergoing instrumented excision.
Keywords: Frailty, Modified Frailty Index, mFI-5, Spinal Tumor, Extradural Tumor, Metastatic Tumor, Risk Stratification, Predictor
Introduction
Spinal tumors are a key contributor to the mortality and morbidity of patients with cancer. Estimates as high as 70% indicate that patients with systemic neoplasia harbor secondary spinal metastases at the time of autopsy1). Around 18,000 metastatic cases to the spine are diagnosed in North America every year2). This number is ever-growing as new developments have significantly prolonged the survival time of patients with cancer. Primarily lesions represent only ～10% of spinal tumors in comparison to the vast majority by metastatic spread3). For these reasons, metastatic spinal tumor management is of particular importance in mitigating damage to patients with cancer of all kinds.
Nonetheless, the management of metastatic spinal tumors presents a host of difficult challenges for the physician to navigate. For one, surgical intervention of the spine is complex and holds a substantial risk of vital and neural complications4,5). Furthermore, the supplemental chemotherapy, radiation, and spinal reconstruction that often comes with spinal tumor resection make postoperative complication rates higher than in any other type of spine surgery1,6-10). With this in mind, surgeons are being challenged to consider other forms of management with far fewer risks for the palliation of symptoms and prolongation of survival. Given this, a predictive tool to better assess the preoperative risk of surgery and better inform decision making in this patient population is needed.
One metric that has been attracting substantial attention is frailty. The 5-item Modified Frailty Index (mFI-5) is of particular interest given its conciseness and usability with the variables listed in the National Surgical Quality Improvement Program (NSQIP) database, making for easy use by clinicians and researchers alike11). The mFI-5 generates a frailty score based on the absence or presence of five comorbidities12) (Appendix A). This index has been proven to be a strong predictor of postoperative adverse outcomes in surgeries for cervical myelopathy, elective posterior lumbar fusions, and kyphoplasties12,13). Despite its potential to serve as a common clinical tool, its value in predicting outcomes following metastatic extradural tumor removal has not yet been fully elucidated. In this study, we evaluate the utility of the mFI-5 as a predictor of morbidity following metastatic extradural spinal tumor resection with instrumented fusion and, specifically, the types of complications it predicts. Further, we evaluate the mFI-5 as a predictor of readmission and reoperation in these patients.
Materials and Methods
Data collection and inclusion criteria
The American College of Surgeons' NSQIP database is a prospectively collected database that collects records from randomly selected patients across >700 hospitals for >1 million cases annually. More than 150 variables are collected, including demographics, comorbidities, intraoperative variables (e.g., OR time, intra-op labs), and 30-day postoperative outcomes. The NSQIP dataset encodes surgical procedures by Current Procedural Terminology (CPT) codes and diagnoses by the International Classification of Disease, Ninth/Tenth Revision, Clinical Modification (ICD-9/10-CM) codes. Quality and accuracy are maintained by Surgical Clinical Reviewers and interrater reliability audits. All patient information in the NSQIP is deidentified, therefore making this study exempt from Institutional Review Board approval.
CPT codes were used to search the NSQIP database for patients undergoing removal of extradural tumors from 2011 to 2019. All identified patients meeting the inclusion criteria were included for further analyses (Appendix B). Exclusion criteria was also applied (Appendix B).
Frailty index and outcomes definitions
Total mFI-5 scores were calculated with the five relevant NSQIP-provided variables. Scores for each patient were summed for a total between 0 and 5. Based on previous literature, patients included for final analyses were categorized into three groups for comparison: (1) mFI-5=0; (2) mFI-5=1; and (3) mFI-5=2+.
The NSQIP reports 30-day complication data for each patient. The 30-day morbidity was calculated using the NSQIP complication variables. These complications include superficial infections, deep infections, pneumonia, renal failure, pulmonary embolism, cardiac arrest, sepsis, and septic shock, among others. If any postoperative complication was present, the patient received a morbidity score of 1; if none were present, the patient received a score of 0. The 30-day readmission and reoperation variables are also reported in the NSQIP database and were collected to be used for analyses.
Statistical analyses
Descriptive statistics, including the mean and standard deviation for continuous variables and frequency and proportion for categorical variables, were calculated to characterize the cohort. Differences in baseline characteristics between those with and without a postoperative complication were compared with independent t-tests, chi-squared tests, or Fisher's exact test, where appropriate. Multivariate models for morbidity, readmission, and reoperation were constructed following a structured approach. Multiple logistic regression models were fit to sociodemographic data, followed by the addition of frailty and perioperative variables. This process was repeated with the comorbidity variables, leaving frailty out of the models. In both cases, variable selection was based on factors considered to be risk factors for morbidity. The Akaike information criterion (AIC) was used to compare the relative fit of a model based on frailty versus a model based on the comorbidity variables. Both model complexity and AIC values were used to determine the optimal model, with a difference in AIC >2 indicating a meaningful difference in support. Results are presented as odds ratios (ORs) with corresponding 95% confidence intervals. Models were constructed using the R program and glm package (version 3.4.3, R Foundation for Statistical Computing, Vienna, Austria).
To better characterize which complications were influenced by frailty, complication outcomes as listed in the NSQIP database were categorized into 10 groups: wound infections, wound disruption, respiratory complications, thrombotic events, renal complications, urinary tract infection (UTI), stroke/cerebrovascular accident (CVA), cardiac events, bleeding requiring transfusion, and sepsis/septic shock. A chi-squared test was used to ascertain a statistically significant correlation between the three aforementioned frailty levels and the incidence of these complication groups. However, in the case of complication types with low incidence rates (five or fewer patients in at least two frailty levels), Fisher's exact test was utilized because of the decreased accuracy of the chi-squared test under these statistical circumstances. Analyses were carried out using the R program.
Results
Univariate analyses
Following the application of inclusion and exclusion criteria, the NSQIP database yielded 874 patients who underwent instrumented extradural tumor resection. The majority of patients were male (61.0%) and white (81.8%), with a mean age of 61.1 years. The average patient was overweight (body mass index (BMI)=25-29.9 kg/m2), with a mean BMI of 27.7 kg/m2. The number of patients in each mFI-5 category were 395 (45.2%), 326 (37.3%), and 153 (17.5%) for mFI-5 of 0, 1, and 2+, respectively.
Univariate analyses revealed several variables that were significantly (p<0.05) associated with morbidity at baseline (Table 1). Elective vs. nonelective surgery, BMI, ASA ≥ vs.<3, operative time, and length of hospital stay were all found to be significantly associated with morbidity, among others, before adjustment. Notably, mFI-5=0, mFI-5=1, and mFI-5=2+ were not found to be significant in predicting morbidity at baseline, prior to adjustment.
Table 1.
Descriptive Statistics and Univariate Analyses of Patient Characteristics.
| Patient Characteristics | Variable | Morbidity–No | Morbidity–Yes | p-Value |
|---|---|---|---|---|
| Demographics | ||||
| Sex | Female (%) | 173 (41.4) | 168 (36.8) | 0.191 |
| Male (%) | 245 (58.6) | 288 (63.2) | ||
| Race | Non-White (%) | 67 (16.0) | 67 (14.7) | 0.702 |
| White (%) | 290 (69.4) | 312 (68.4) | ||
| Ethnicity | Hispanic (%) | 26 (6.2) | 31 (6.8) | 0.793 |
| Non-Hispanic (%) | 331 (79.2) | 367 (80.5) | ||
| Age (years) | Mean (SD) | 60.9 (12.3) | 61.4 (12.2) | 0.561 |
| BMI | Mean (SD) | 28.2 (5.7) | 27.1 (5.6) | 0.008 |
| Comorbidities | ||||
| CHF | No (%) | 417 (99.8) | 450 (98.7) | 0.126 |
| Yes (%) | 1 (0.2) | 6 (1.3) | ||
| COPD | No (%) | 399 (95.5) | 428 (93.9) | 0.371 |
| Yes (%) | 19 (4.5) | 28 (6.1) | ||
| HTN | No (%) | 232 (55.5) | 232 (50.9) | 0.193 |
| Yes (%) | 186 (44.5) | 224 (49.1) | ||
| Diabetes | No (%) | 369 (88.3) | 395 (86.6) | 0.526 |
| Yes (%) | 49 (11.7) | 61 (13.4) | ||
| Functional status | No (%) | 388 (92.8) | 409 (89.7) | 0.131 |
| Yes (%) | 30 (7.2) | 47 (10.3) | ||
| Frailty | mFI-5=0 (%) | 204 (484) | 191 (41.9) | 0.109 |
| mFI-5=1 (%) | 148 (35.4) | 178 (39.0) | ||
| mFI-5=2+ (%) | 66 (15.8) | 87 (19.1) | ||
| ASA | <3 (%) | 354 (84.7) | 424 (93.2) | <0.001 |
| ≥3 | 64 (15.3) | 31 (6.8) | ||
| Weight loss | No (%) | 398 (95.2) | 406 (89.0) | 0.001 |
| Yes (%) | 20 (4.8) | 50 (11.0) | ||
| Ventilator | No (%) | 417 (99.8) | 447 (98.0) | 0.022 |
| Yes (%) | 1 (0.2) | 9 (2.0) | ||
| Perioperative | ||||
| Elective | No (%) | 261 (62.4) | 344 (75.4) | <0.001 |
| Yes (%) | 157 (37.6) | 112 (24.6) | ||
| Preop sepsis | No (%) | 393 (94.0) | 407 (89.3) | 0.016 |
| Yes (%) | 25 (6.0) | 49 (10.7) | ||
| Preop transfusion | No (%) | 413 (98.8) | 437 (95.8) | 0.007 |
| Yes (%) | 5 (1.2) | 19 (4.2) | ||
| Emergency | No (%) | 364 (87.1) | 374 (82.0) | 0.049 |
| Yes (%) | 54 (12.9) | 82 (18.0) | ||
| Operative time | Mean (SD) | 7.3 (2.9) | 8.9 (3.8) | <0.001 |
| LOS | Mean (SD) | 9.0 (6.2) | 12.5 (9.1) | <0.001 |
| Discharge disposition | Home (%) | 262 (62.7) | 196 (43.0) | <0.001 |
| Rehab (%) | 150 (35.9) | 234 (51.3) | <0.001 | |
| Preoperative labs | ||||
| INR | Mean (SD) | 1.1 (0.1) | 1.1 (0.2) | 0.003 |
| HCT | Mean (SD) | 38.1 (5.1) | 35.2 (5.5) | <0.001 |
Bold p-values are statistically significant (p<0.05). BMI: Body Mass Index; CHF: Congestive Heart Failure; COPD: Chronic Obstructive Pulmonary Disease; HTN: Hypertension; ASA: American Society of Anesthesiology Score; LOS: Length of Stay; INR: International Normalized Ratio; HCT: Hematocrit
Multivariate analyses
Multivariate logistic regression analyses demonstrated an increase in morbidity in patients with an mFI-5 level of 1 as compared to 0 (OR: 1.45, SE: 0.31, p=0.036). The same model demonstrated an increase in morbidity in patients with an mFI-5 level of 2+ as compared to 0 (OR: 1.41, SE: 0.40, p=0.036). Multivariate analyses further demonstrated that elective surgery (OR: 0.061, SE: 0.012, p=0.004), increasing OR time (OR: 1.180, SE: 0.03, p≤0.001), and chronic steroid use (OR: 1.56, SE: 0.42, p=0.037) predicted morbidity (Table 2).
Table 2.
Multivariate Logistic Regression Analyzing Predictors of Morbidity.
| Independent variable | Odds ratio | Standard error | p-Value |
|---|---|---|---|
| Demographics | |||
| Age | 1.00 | 0.01 | 0.869 |
| Sex (males) | 1.38 | 0.27 | 0.052 |
| BMI | 0.96 | 0.01 | 0.004 |
| Comorbidities | |||
| mFI-5=1 | 1.45 | 0.31 | 0.036 |
| mFI-5=2+ | 1.41 | 0.40 | 0.036 |
| Weight loss | 1.38 | 0.61 | 0.297 |
| Smoker (yes) | 1.18 | 0.29 | 0.411 |
| Steroids (yes) | 1.56 | 0.42 | 0.037 |
| Perioperative | |||
| Preop transfusion | 1.71 | 2.01 | 0.336 |
| Elective (yes) | 0.61 | 0.12 | 0.004 |
| Operative time | 1.18 | 0.03 | <0.001 |
| Preoperative labs | |||
| Creatinine | 0.90 | 0.18 | 0.510 |
| WBC | 1.03 | 0.02 | 0.136 |
| HCT | 0.90 | 0.01 | <0.001 |
All listed variables controlled for in the multivariate model. Bold p-values are statistically significant (p<0.05). BMI: Body Mass Index; WBC: White Blood Cells; HCT: Hematocrit
Multivariate logistic regression demonstrated no significance in readmission or reoperation between mFI-5 levels 1 and 2+ as compared to 0.
The AIC was analyzed to determine which model best fits the data. The AIC for the model based on frailty was calculated to be 1012.1, whereas the AIC for the model based on the mFI-5 comorbidity variables was calculated to be 1015.4.
Chi-squared test and Fisher's exact test
The observed rates of the different complication types as listed in the NSQIP are noted (Table 3). Expected values were calculated and recorded (Table 4).
Table 3.
Observed Incidence of Complication Categories among the Three Frailty Groups.
| mFI-5=0 | mFI-5=1 | mFI-5=2+ | Total | |
|---|---|---|---|---|
| Wound infection | 14 | 10 | 9 | 33 |
| Including superficial wound infection, deep incisional SSI, organ/space SSI | ||||
| Wound disruption | 5 | 1 | 1 | 7 |
| Respiratory | 21 | 19 | 19 | 59 |
| Including pneumonia, unplanned intubation, ventilator >48 hours | ||||
| Thrombotic events | 31 | 22 | 12 | 65 |
| Including pulmonary embolism, DVT | ||||
| Renal events | 0 | 2 | 2 | 4 |
| Includingpr ogressive renal insufficiency, acute renal failure | ||||
| UTI | 8 | 11 | 10 | 29 |
| Stroke/cerebrovascular accident | 1 | 3 | 1 | 5 |
| Cardiac events | 1 | 9 | 4 | 14 |
| Including cardiac arrest requiring CPR, myocardial infarction | ||||
| Bleeding requiring transfusion | 155 | 149 | 74 | 378 |
| Sepsis & septic shock | 8 | 9 | 12 | 29 |
| No listed complication | 152 | 96 | 11 | 259 |
| Totals | 396 | 331 | 155 | 882 |
All observed instances of complication categorized in groups among frailty groups. SSI: surgical site infection; DVT: Deep Vein Thrombosis; UTI: Urinary Tract Infection; CPR: Cardiopulmonary Resuscitation
Table 4.
Expected Incidence of Complication Categories among the Three Frailty Groups.
| mFI-5=0 | mFI-5=1 | mFI-5=2+ | Total | |
|---|---|---|---|---|
| Wound infection | 14.81633 | 12.38435 | 5.79932 | 33 |
| Including superficial wound infection, deep incisional SSI, organ/space SSI | ||||
| Wound disruption | 3.142857 | 2.626984 | 1.230159 | 7 |
| Respiratory | 26.4898 | 22.14172 | 10.36848 | 59 |
| Including pneumonia, unplanned intubation, ventilator >48 hours | ||||
| Thrombotic events | 29.18367 | 24.39342 | 11.4229 | 65 |
| Including pulmonary embolism, DVT | ||||
| Renal events | 1.795918 | 1.501134 | 0.702948 | 4 |
| Includingpr ogressive renal insufficiency, acute renal failure | ||||
| UTI | 13.02041 | 10.88322 | 5.096372 | 29 |
| Stroke/cerebrovascular accident | 2.244898 | 1.876417 | 0.878685 | 5 |
| Cardiac events | 6.285714 | 5.253968 | 2.460317 | 14 |
| Including cardiac arrest requiring CPR, myocardial infarction | ||||
| Bleeding requiring transfusion | 169.7143 | 141.8571 | 66.42857 | 378 |
| Sepsis & septic shock | 13.02041 | 10.88322 | 5.096372 | 29 |
| No listed complication | 116.2857 | 97.19841 | 45.51587 | 259 |
| Totals | 396 | 331 | 155 | 882 |
Expected instances of complication categorized in groups among frailty groups. SSI: surgical site infection; DVT: Deep Vein Thrombosis; UTI: Urinary Tract Infection; CPR: Cardiopulmonary Resuscitation
Chi-squared tests and Fisher's exact tests were used on an individual complication category basis to better illustrate the correlation of frailty with each different complication type. Notably, wound disruption, renal events, stroke/CVA, and cardiac events had very low incidence rates, necessitating use of Fisher's exact test because of the reduced efficacy of the chi-squared test when incidence counts are fewer than 5 in more than a single category. These statistical tests established significance (p<0.05) between some individual complication groups and frailty, including respiratory complications, UTI, cardiac events, and sepsis/septic shock. These analyses demonstrated no significance between the other complication groups and frailty (Table 5).
Table 5.
Individual Chi-squared and Fisher’s Exact Tests for Each Complication Group among Frailty Groups.
| Complication category | p-Value |
|---|---|
| Wound infection | 0.31 |
| Wound disruption | 0.64* |
| Respiratory | 0.0091 |
| Thrombotic | 0.82 |
| Renal | 0.092* |
| UTI | 0.032 |
| Cerebrovascular accident | 0.26* |
| Cardiac events/MI | 0.0028* |
| Bleeding requiring transfusion | 0.11 |
| Sepsis/Septic shock | 0.0025 |
Frailty as an independent predictor of specific complication groupings. Bold p-values are statistically significant (p<0.05). Asterisks (*) indicate analysis with Fisher’s exact test. UTI: Urinary Tract Infection; MI: Myocardial Infarction
Discussion
Our analyses sought to illustrate the ability of the mFI-5 in predicting morbidity following instrumented metastatic extradural tumor resection. After controlling for the patient-related and procedural factors found in Table 2, we found that a mFI-5 of 1 independently predicted 45% increased odds of morbidity as compared to the control (mFI-5 of 0) and that a mFI-5 of 2+ predicted 41% increased odds of morbidity as compared to the control (mFI-5 of 0). Given that the AIC for the model based on frailty (1012.1) was smaller than the model based on the mFI-5 comorbidities (1015.4) and that there was a difference >2, there is evidence of a better fit from the frailty model to the data and a meaningful difference between the two regression analyses.
Although it is common sense that patients with comorbidities are at an increased risk of complications, such a markedly elevated risk with just one of the five comorbidities making up this index is of significance. These findings suggest that frailty is a strong predictor for morbidity in this patient population. However, given the risk of complication with one comorbid condition, intuitively, one would theorize that additional comorbidities should further increase the risk of complication. Nonetheless, this was not supported by the data. The apparent reversal of the expected ORs may be explained by the possibility that several comorbidities do not confer any more risk than just one. Or it may be explained by the low number of patients in the mFI-5=2+ category. Lastly, the difference between the theorized trend in ORs and the trend observed may be best explained by the limitations of the mFI-5 index. It is likely these patients had other medical conditions not captured by the mF-5 that may be obscuring the data and subsequent results.
We also found that elective surgeries of this nature as compared to nonelective surgeries were significant predictors of postoperative morbidity, decreasing the risk by 39%. This may be a result of the acuity of treatment (or lack thereof) and/or the condition of elective vs. nonelective patients with metastatic spread to the spine. Nonelective cases may have been in situations of acute cord compression due to the tumor itself or a tumor-related fracture, with retropulsion into the canal. Without specificity in the NSQIP database, we cannot know for sure. This unexplained finding is a limitation of this study.
Further, we found an 18% increased risk of postoperative morbidity for every 30 additional minutes of operative time. This is likely due to the increased infection risks conferred by the lengthened OR time. Moreover, the operative time likely increases proportionally with tumor size. A larger neoplasm may result in greater bleeding risk, requiring transfusion or larger surgical wounds with an increased risk of disruption which may also explain this finding. Additionally, we found that steroid use for chronic conditions prior to surgery was also a significant predictor of morbidity, best explained by steroids' adverse effects on bone density.
It is also of note that neither an mFI-5 of 1 or 2+ was significantly associated with readmission or reoperation. The complexity of patients with cancer and their management likely contributes to this insignificance. Additionally, we would expect that this population in particular has high rates of readmission and reoperation for reasons not related to their spinal tumor resection. As the NSQIP variables do not specify why patients seek readmission or reoperation, it is fair to assume that unassociated instances of additional care that do not relate to the patients' frailty as described by the mFI-5 are captured in this dataset. This is another limitation of this study.
Additionally, the chi-squared analyses and Fisher's exact tests used to determine whether frailty is a significant predictor of the development of a given complication group showed that respiratory complications, UTIs, cardiac events, and sepsis/septic shock are significantly influenced by frailty. Although it is difficult to determine why some complications were statistically influenced by frailty and others not, we appreciate the limitations of small sample sizes and encourage future research to identify a greater number of patients with morbidity. Higher powered studies, for example, may find significant differences for other complication categories not observed in these data. Nevertheless, frailty should be considered in patients undergoing extradural tumor resection. Admission to higher observational units, such as the ICU or sub-ICU, postoperatively for patients with mFI-5s >0 should be considered to better mitigate morbidity.
Our results add to the emerging body of evidence that the mFI-5 is a useful predictor of postoperative outcomes in spine surgery and orthopedics at large. Weaver et al. found that the mFI-5 was successful in predicting postsurgical complications and outcomes in frail patients who had undergone elective posterior lumbar fusions13). Further, Segal et al. concluded that there is a significant correlation between frailty and adverse outcomes after kyphoplasty12). Moreover, Zreik et al. established that the mFI-5 is significantly associated with 30-day adverse outcomes following anterior cervical discectomy and fusion14). Although it has fallen out of favor because of its difficulty implementing clinically and its lack of synchrony with the NSQIP database, others have proven the predictive value of frailty by utilizing the 11-Item Modified Frailty Index in spine surgery and orthopedics11,15-17). Despite the overwhelming evidence of its predictive power, there are some instances where the mFI-5 index has been proven to be a poor prognostic tool. Elsamadicy et al., for example, found that frailty, as measured by mFI-5, does not significantly predict length of stay, 30-day adverse events, or 30-day unplanned readmission in patients undergoing lumbar spinal decompression and fusion for spondylolisthesis18). Thus, it is important to continue to assess the predictive value of the mFI-5 in various spine surgeries and otherwise.
Assessing surgical outcomes in the management of metastatic spread to the spine is important for clinical decision making - understanding postoperative complication risk is particularly useful for counseling and guiding treatment decisions in these complex surgeries and patients. Risk stratification can aid in proper patient selection. Although age has often been used as a measure of one's health, we now better understand that it does not always fully depict one's physiologic state. This is the niche that frailty has begun to fill in both current literature and clinical practice alike. Given that its predictive power has been proven in various orthopedic operations and that it can be easily applied in practice with the accessibility of the variables that make up the index, the mFI-5 is a strong clinical tool for orthopedists and spine surgeons. The risk of morbidity following extradural tumor resection can now be evaluated with more accuracy preoperatively.
This study has several limitations inherent to the database queried and otherwise. First, there is a possibility of ICD and CPT coding error and reporting biases. Second, the generalizability of our findings may be limited by the retrospective nature of the study. Additionally, unequal sample sizes in mFI-5 groups and possible confounders unaccounted for in the database may have obscured the results. Furthermore, morbidities characteristic of spine surgery, such as implant-related complications or neurological paralysis, are not recorded in the NSQIP database and therefore, cannot contribute to the computed morbidity scores. We recognize that without these outcomes accounted for in the dataset, true complication risk may be obscured and that morbidity rates may be understated. Finally, the NSQIP only records complication rates within a 30-day window, meaning that some cases of morbidity, readmission, and reoperation may have been missed given this limited timeframe. We encourage that more studies be conducted on the utility of the mFI-5 in predicting the postoperative morbidity, readmission, and reoperation in this patient population and others with a larger sample size and with complications characteristic of spine surgery accounted for. Regardless, the findings described throughout this study should shed some light on the predictive value of frailty, measured by the mFI-5 index, on postoperative complication rates.
Our study indicates that the mFI-5 is a strong predictor of 30-day complications following the resection of metastatic, extradural tumors with instrumented fusion. We found that, while adjusting for confounders, there was a 45% increase in morbidity of patients with an mFI-5=1 and a 41% increase in morbidity of patients with an mFI-5=2+ as compared to the mFI-5=0 control group. The mFI-5 is a straightforward risk stratification tool that can be used to correlate the physiologic condition of patients preoperatively with morbidity following surgical resection of the metastatic, extradural tumor. As it relies on readily accessible variables, it can be applied efficiently and effectively. Given its predictive value and ease of application, it would be no surprise to see the mFI-5 index become clinically commonplace, especially in the field of orthopedics and spine surgery.
Conflicts of Interest: Dr. Jeff Silber receives teaching fees for Stryker.
Dr. David Essig receives consulting fees for Stryker and DePuy.
For all remaining authors, none was declared.
Sources of Funding: No funding was received in connection to this study.
Author Contributions: A.S., A.K., M.S., S.S., J.S., R.V., S.V., J.S., and D.E. all provided contributions to the conception and design of the study. The authors all helped analyze data, draft the work, or provided critical revisions necessary for the conception of the research.
Ethical Approval: IRB approval not required for de-identified NSQIP research with no direct patient involvement.
Informed Consent: Consent was not required for de-identified NSQIP research with no direct patient involvement.
Supplementary Material
References
- 1.Jacobs WB, Perrin RG. Evaluation and treatment of spinal metastases: an overview. Neurosurg Focus. 2001;11(6):1-11. [DOI] [PubMed] [Google Scholar]
- 2.Lawton AJ, Lee KA, Cheville AL, et al. Assessment and management of patients with metastatic spinal cord compression: a multidisciplinary review. J Clin Oncol. 2019;37(1):61-71. [DOI] [PubMed] [Google Scholar]
- 3.Ahmed AK, Goodwin CR, De la Garza-Ramos R, et al. Predicting short-term outcome after surgery for primary spinal tumors based on patient frailty. World Neurosurg. 2017;108:393-8. [DOI] [PubMed] [Google Scholar]
- 4.Kimmell KT, Algattas H, Joynt P, et al. Risk modeling predicts complication rates for spinal surgery. Spine. 2015;40(23):1836-41. [DOI] [PubMed] [Google Scholar]
- 5.Patil CG, Patil TS, Lad SP, et al. Complications and outcomes after spinal cord tumor resection in the United States from 1993 to 2002. Spinal Cord. 2008;46(5):375-9. [DOI] [PubMed] [Google Scholar]
- 6.Fehlings MG, Nater A, Zamorano JJ, et al. Risk factors for recurrence of surgically treated conventional spinal schwannomas: analysis of 169 patients from a multicenter international database. Spine. 2016;41(5):390-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Fisher CG, Versteeg AL, Dea N, et al. Surgical management of spinal chondrosarcomas. Spine. 2016;41(8):678-85. [DOI] [PubMed] [Google Scholar]
- 8.Fourney DR, Gokaslan ZL. Current management of sacral chordoma. Neurosurg Focus. 2003;15(2):1-5. [DOI] [PubMed] [Google Scholar]
- 9.Hsu W, Kosztowski TA, Zaidi HA, et al. Multidisciplinary management of primary tumors of the vertebral column. Curr Treat Options Oncol. 2009;10(1-2):107-25. [DOI] [PubMed] [Google Scholar]
- 10.Kaloostian PE, Zadnik PL, Etame AB, et al. Surgical management of primary and metastatic spinal tumors. Cancer Control. 2014;21(2):133-9. [DOI] [PubMed] [Google Scholar]
- 11.Zreik J, Alvi MA, Yolcu YU, et al. Utility of the 5-item modified frailty index for predicting adverse outcomes following elective anterior cervical discectomy and fusion. World Neurosurg. 2021;146:e670-7. [DOI] [PubMed] [Google Scholar]
- 12.Segal DN, Wilson JM, Staley C, et al. The 5-item modified frailty index is predictive of 30-day postoperative complications in patients undergoing kyphoplasty vertebral augmentation. World Neurosurg. 2018;116:e225-31. [DOI] [PubMed] [Google Scholar]
- 13.Wilson JRF, Badhiwala JH, Moghaddamjou A, et al. Frailty is a better predictor than age of mortality and perioperative complications after surgery for degenerative cervical myelopathy: an analysis of 41,369 patients from the NSQIP Database 2010-2018. J Clin Med. 2020;9(11):3491. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Weaver DJ, Malik AT, Jain N, et al. The modified 5-item frailty index: a concise and useful tool for assessing the impact of frailty on postoperative morbidity following elective posterior lumbar fusions. World Neurosurg. 2019;124:e626-32. [DOI] [PubMed] [Google Scholar]
- 15.Flexman AM, Charest-Morin R, Stobart L, et al. Frailty and postoperative outcomes in patients undergoing surgery for degenerative spine disease. Spine J. 2016;16(11):1315-23. [DOI] [PubMed] [Google Scholar]
- 16.Lakomkin N, Zuckerman SL, Stannard B, et al. Preoperative risk stratification in spine tumor surgery: a comparison of the modified charlson index, frailty index, and ASA score. Spine. 2019;44(13):E782-7. [DOI] [PubMed] [Google Scholar]
- 17.Ali R, Schwalb JM, Nerenz DR, et al. Use of the modified frailty index to predict 30-day morbidity and mortality from spine surgery. J Neurosurg Spine. 2016;25(4):537-41. [DOI] [PubMed] [Google Scholar]
- 18.Elsamadicy AA, Freedman IG, Koo AB, et al. Modified-frailty index does not independently predict complications, hospital length of stay or 30-day readmission rates following posterior lumbar decompression and fusion for spondylolisthesis. Spine J. 2021;21(11):1812-21. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.