Keywords mortality - vertebral compression fracture - osteoporosis - cement augmentation
Introduction
Each year, over two million fragility fractures occur in the United States, hospitalizing
over 500,000 elderly patients and costing the health care system over $5.1 billion
annually.[1 ]
[2 ] Among these fragility fractures, roughly one third are osteoporotic vertebral compression
fractures (OVCFs).[3 ] OVCFs commonly cause dysfunction, severe pain, and postural changes that may be
treated via cement augmentation procedures.[4 ]
[5 ] Considering the rising incidence of osteoporotic fractures in the population[6 ] and the frailty commonly associated with OVCF patients,[7 ] literature regarding the clinical management of this condition is becoming progressively
more relevant. Treatment options for OVCF include benign neglect, bracing, medication,
and procedures such as cement augmentation. While the treatment plan for a given patient
is based on various clinical factors, several studies have shown that vertebral cement
augmentation may offer better long-term pain relief than medical management alone
for a subset of patients.[8 ]
[9 ]
[10 ]
[11 ] However, prior to recommending an invasive procedure, clinicians must consider the
risk-profile of each patient. In this study, we attempt to determine the relationship
between 15 comorbidities and the likelihood of developing complications, mortality,
or 30-day readmission following an invasive procedure for an OVCF. In doing so, we
hope to help guide clinical decision-making regarding treatment plans and help clinicians
manage patients’ expectations following vertebral cement augmentation.
Materials and Methods
In this retrospective study, data from the American College of Surgeons National Surgery
Quality Improvement Project (ACS-NSQIP) from 2007 through 2014 was reviewed. Because
information within this database is deidentified, this study was exempt from institutional
review board (IRB) approval. Following a methodology utilized by others,[12 ]
[13 ]
[14 ] patients who sustained an OVCF in the lumbar or thoracic region of the spine were
identified using current procedural terminology (CPT) codes and International Classification
of Disease (ICD-9 or ICD-10) codes. ICD-9 codes 733.13, 805.2, and 805.4 (utilized
for the years 2007–2013) and ICD-10 codes M48.56XA, S22.009A, S22.068A, and S22.089A
(used for 2014) were also included in this study. To ensure all target patients in
the ACS-NSQIP database were captured, patients assigned any of the following CPT codes
were also included: 22510, 22511, 22512, 22513, 22514, or 22515; these CPT codes refer
to vertebral cement augmentation procedures.
Patients with an OVCF in the cervical, sacral, or unclassified region of the spine,
and/or those with confirmed spinal or central nervous system (CNS) tumors were excluded.
For purposes of conducting a logistic regression analysis, the following patient characteristics
were considered: gender, body mass index (BMI), functional status prior to procedure,
preoperative comorbidities and preoperative serum albumin level, American Society
of Anesthesiologists (ASA) status, postoperative complications, mortality, reoperations,
and 30-day readmissions. If greater than 10 percent of patients were missing data
for a given variable, that variable was excluded from the study.
Outcome measures for this study included minor postoperative complications, major
postoperative complications, patient mortality, 30-day readmission due to any cause,
and 30-day readmission related to OVCF. As in a previous study by Chung et al,[14 ] the following issues were considered minor postoperative complications: pneumonia,
urinary tract infection (UTI), deep vein thrombosis (DVT), or incision site complications.
Major postoperative complications included cardiac arrest, acute myocardial infarction
(MI), sepsis, septic shock, stroke, pulmonary embolism (PE), acute renal failure,
a coma lasting more than 24 hours, and reintubation.
Statistical Analysis
Using multivariate logistic regression, odds ratios (OR) with corresponding p values and 95% confidence intervals (CI) were calculated. This statistical tool was
used to determine the direct correlation between 15 comorbidities and primary outcome
measures, while also accounting for possible confounders such as age, gender, or preoperative
serum albumin levels. While most comorbidities were considered dichotomous, preoperative
functional status was treated as a categorical variable. Patients exhibiting an inability
to perform at least one activity of daily living (ADL) were considered partially dependent;
if a patient was unable to perform any ADLs, he or she was classified as completely
dependent. Preoperative serum albumin level was treated as a continuous variable.
To determine which variables should be utilized in the logistic regression, a series
of bivariate analyses between age and the variable in question were conducted; those
that generated p values less than 0.05 were included in the multivariate analysis. Any variable for
which fewer than five incidences occurred within the patient dataset was excluded.
For statistical purposes, a patient listed as greater than 90 years of age in the
ACS-NSQIP database was treated as a 90-year old.
Results
Upon application of inclusion and exclusion criteria, 1979 patients were found in
the ACS-NSQIP database. [Table 1 ] provides a demographic overview of our study population. Patients were on average
approximately 74 years of age at the time of vertebral cement augmentation. More than
two-thirds were female and the average BMI was slightly under 27. Approximately one
out of seven patients had a smoking history within 1 year of the index procedure;
the same was true regarding chronic steroid use and diminished functional status prior
to intervention. On average, patients had a preoperative albumin in the healthy range
(> 3.5).[7 ]
Table 1
Patient characteristics
Total (n = 1979)
SD or %
Abbreviations: BMI, body mass index; SD, standard deviation.
a For statistical purposes, patients age 90+ are assumed to be 90 years of age.
Parameter
Age (years) a
73.92
10.82
Sex (female)
1328
67.1%
BMI
26.74
6.45
Mean preoperative albumin
3.58
0.65
Steroid use for chronic condition
293
14.8%
Smoking history
270
13.6%
On dialysis
26
1.3%
Functional status prior to surgery
Independent
1661
85.3%
Partially dependent
265
13.6%
Totally dependent
22
1.1%
Unknown
31
–
[Table 2 ] summarizes the prevalence of various comorbidities as well as the distribution of
ASA status in the study population. Nearly two-thirds of patients had a history of
hypertension and more than one-quarter were obese; almost 80% of patients had an ASA
class of III or higher, highlighting the frailty of OVCF patients. [Table 3 ] summarizes postoperative outcomes in our study population. Mortality was seen in
4.6% of patients within 30 days of the index procedure, and 13.1% of patients were
readmitted. Overall, 5.1% of patients experienced a major complication, most commonly
sepsis and reintubation, and 7.2% of patients suffered minor complications.
Table 2
Comorbidities and distribution of ASA status across 1979 patients
Diagnosis
# Patients
%
Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; CHF,
congestive heart failure; COPD, chronic obstructive pulmonary disorder; CVA, cerebrovascular
accident; MI, myocardial infarction.
Comorbidity
CHF
56
2.8%
COPD
307
15.5%
Coagulopathy
173
8.7%
CVA
16
0.8%
Diabetes
385
19.5%
Dialysis
26
1.3%
Hypertension
1317
66.5%
Ascites
3
0.2%
MI < 6 months
0
0.0%
Obesity (BMI ≥ 30)
497
25.1%
Renal Failure
4
0.2%
Smoker (w/in last year)
270
13.6%
Steroid use for chronic condition
293
14.8%
Weight loss
66
3.3%
ASA Status
ASA Class I
11
0.6%
ASA Class II
390
19.7%
ASA Class III
1297
65.6%
ASA Class IV
274
13.9%
ASA Class V
4
0.2%
None listed
3
–
Table 3
Postoperative outcomes among 1979 patients
Diagnosis
# Patients
%
Abbreviations: DVT, deep vein thrombosis; MI, myocardial infarction; PE, pulmonary
embolism; UTI, urinary tract infection; VCF, vertebral compression fracture.
Mortality
91
4.6%
Reoperations
83
4.2%
Readmissions
259
13.1%
Readmission related to VCF
169
8.5%
Mean length of Hospital Admission
5.93
13.89
Major complications
Acute MI
7
0.4%
Cardiac arrest
10
0.5%
Sepsis
38
1.9%
Septic shock
18
0.9%
Stroke
10
0.5%
PE
16
0.8%
Acute renal failure
5
0.3%
Coma > 24 hours
1
0.1%
Reintubation
31
1.6%
Prolonged intubation (> 7 days)
17
0.9%
Overall
101
5.1%
Minor Complications
DVT
31
1.6%
Pneumonia
64
3.2%
UTI
63
3.2%
Surgical site complication
39
2.0%
Overall
143
7.2%
[Fig. 1 ]
summarizes the results of our multivariate logistic regression. To determine which
factors to integrate into the analysis, a series of bivariate tests were conducted.
Of the 15 comorbidities recorded in our database, 11 bore a statistical impact (p value < 0.05) on at least one of the primary outcome measures. These included a history
of congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), coagulopathy,
cerebrovascular accident (CVA), diabetes mellitus (DM), dialysis, obesity, long-term
steroid use, weight loss exceeding 10% of body weight over the past 6 months, ASA
class > 2, and preoperative functional status. Ascites and renal failure had fewer
than five incidences and therefore were excluded from analysis. In addition, hypertension
and smoking were also excluded from the analysis because the bivariate analyses for
these variables failed to demonstrate statistical significance. To further validate
our results, age, gender, and preoperative serum albumin levels were also integrated
in our analysis.
Fig. 1 Multivariate logistic regression analysis for minor complications, major complications,
mortality, and 30-day readmissions. Variables with statistically significant p-values
are highlighted in orange. Ascites and renal failure (highlighted in red) were excluded
from logistic regression because of small sample size for these populations.
Results from
[Fig. 1 ]
show that preoperative functional status (p = 0.03 and 0.05), ASA class > 2 (p = 0.01), and/or a history of CVA (p = 0.02) were statistically correlated with minor complications. Preoperative functional
status (p < .01), a history of CVA (p = 0.05), and a history of coagulopathy (p = 0.03) were associated with major complications. Notably, a history of CVA also
exhibited the highest OR (5.36 and 4.60, respectively) among any comorbidity considered
in this study. Our data also show that preoperative functional status (p < .01), chronic steroid use (p = 0.03), and an ASA class > 2 (p < 0.01) were statistically correlated with mortality; the latter exhibited the highest
impact on mortality (OR = 14.65). Readmission within 30 days of the index procedure
was associated with a history of COPD (p < 0.01), obesity (p = 0.01), significant weight loss (p = 0.04), and an ASA class > 2 (p = 0.02).
Discussion
As the most common type of fragility fracture, OVCFs occur every 22 seconds on average
around the globe.[15 ] The prevalence of OVCF is estimated to be 25% in postmenopausal women over age 50,
and 40% of women over age 80, thus making the management of OVCF highly relevant to
patient care and health care costs.[16 ]
[17 ]
[18 ] As a result of the high prevalence, demographics, interventions, and patient outcomes
following procedures for OVCF have been well-studied.[19 ]
[20 ]
[21 ] Toy et al found an association between postoperative complications and an unfavorable
ASA class, in addition to inpatient status prior to procedure.[22 ] Similarly, we found that patient classification by ASA status is in some regards
a “catch-all” for the general health of a patient,[23 ] helping clinicians identify whether a patient is healthy, has mild systemic disease,
severe systemic disease, or life-threatening disease.[24 ] To assign an ASA class to a patient, many of the comorbidities from this study are
considered.[24 ]
[25 ] Categorizing a patient as “mildly sick” or “moderately sick,” while helpful, may
not give the specific insight that our data provides. The same can be said of preoperative
functional status, which was included in our analysis to be as comprehensive as possible;
however, like ASA status, it may not provide the specific insight that formally diagnosed
medical conditions provide. Thus, the motivation for our logistic regression analysis
was to see which disease processes were specifically impactful in the treatment of
OVCF.
While we found that patient ASA class was associated with complications, we also identified
correlations with a history of CVA and coagulopathy. These comorbidities are, not
surprisingly, among those factors that would lead to an ASA classification greater
than 2. At the same time, other comorbidities that may also place patients in the
same category, such as CHF or hypertension, were not found to be associated with any
adverse outcomes following cement augmentation for OVCF. By providing this specific
quantitative data from a large sample size, we hope to identify which comorbidities
predict complications, mortality, or readmission, potentially enabling physicians
to form a better treatment plan.
Our data shows that the most significant predictor of complications, both in terms
of statistical significance and OR, is a history of CVA (p = 0.02 and OR = 5.36 for minor complications; p = 0.05 and OR = 4.60 for major complications). Coagulopathy was also associated with
major postoperative complications (p = 0.03 and OR = 1.89), however neither coagulopathy nor a history of CVA was found
to be statistically correlated with mortality or readmission. To explain this, we
considered the long-term treatment of CVA, which commonly involves antiplatelet or
anticoagulant therapy to prevent reoccurrence.[26 ]
[27 ] Prior to intervention, CVA patients often discontinue their anticoagulation medication.[28 ] In an elderly and frail OVCF patient population, the lack of anticoagulation may
disproportionately increase the risk of DVT postoperatively, which was considered
a minor complication in our study. Additionally, a prior CVA may increase the likelihood
of a second ischemic event following the procedure–this was considered a major complication.[29 ]
[30 ] Our theory is supported by the literature, which suggests that 25% of CVA are recurrent,
thereby increasing the likelihood of such an event in patients with a history of CVA.[31 ]
In terms of mortality, we found a statistical association in patients using steroids
and those with ASA status > 2. This correlation may be due to the association of steroids
with sepsis and septic shock. While these complications are somewhat rare (1.9% experienced
sepsis and 0.9% endured septic shock), they are often life-threatening in the frail
and elderly OVCF population.[32 ]
[33 ] In one study relating to colorectal surgery, patients using steroids chronically
were shown to have a higher risk of malnutrition, diabetes, bleeding disorders, and
shock[34 ]; postoperatively, they exhibited higher rates of mortality and morbidity.[34 ] Singla et al showed that following lumbar spine fusion, chronic steroid use was
associated with incisional site infections in patients over the age of 65 as well
as increased mortality rates.[35 ] Our study corroborates this finding and highlights the potential perils of chronic
steroid use in patients undergoing vertebral cement augmentation for OVCF.
A statistical correlation was seen in readmission rates in patients with obesity,
COPD, and recent weight loss exceeding 10% of body weight. Ilyas et al also showed
that obesity was associated with an increased rate of readmission following lumbar
spine surgery,[36 ] while Elsamadicy found a correlation between BMI and elective spine procedures in
general.[37 ] For this reason, it may be prudent for physicians to inform obese patients and those
with COPD of the increased likelihood for readmission. Recent weight loss exceeding
10% of total body weight over 6 months is suggestive of a malignancy or significant
infection–these patients may need to be examined further for the possibility of other
pathology associated with these conditions, which could explain the higher rate of
readmission.[38 ]
Gupta et al[7 ] showed that preoperative albumin levels predict the incidence of postoperative complications
following cement augmentation for OVCF and that the likelihood of an adverse outcome
is proportional to the degree of hypoalbuminemia. To ensure the accuracy of results
in this study, serum albumin level was treated as a continuous variable instead of
a dichotomous variable. Similarly, our multivariate analysis included many of the
various elements that comprise the Charlson comorbidity index (CCI). This included
factors such as age, history of CHF, CVA, COPD, diabetes mellitus, and other comorbidities
for which information was provided within the ACS-NSQIP database. The CCI is used
to predict the risk of death within 1 year of hospitalization, based on comorbid conditions.
The literature suggests that ASA status and CCI are related and often in close agreement,
hence its relevance to this study.[21 ]
[39 ]
[40 ] While this study does not control for all possible confounders, we felt that both
CCI and preoperative serum albumin levels were important to incorporate in our analysis.
The ACS-NSIP database utilized in this study enabled us to analyze postprocedural
outcomes following cement augmentation for the treatment of OVCF. However, we are
unable to comment on how patients who are treated nonsurgically do in comparison to
those who undergo cement augmentation based on our data. To address this alternative
approach, we consider recent literature. Ong et al, in a 5-year study, showed that
both balloon kyphoplasty and vertebroplasty were associated with statistically lower
mortality rates in comparison to nonsurgical management of vertebral compression fracutres.[41 ] Similarly, Hirsch et al utilized a 10-year sample of Medicare data to conduct a
number needed to treat analysis. This study showed that at both 1-year and 5-year
follow-ups kyphoplasty patients did better. The adjusted number needed to treat to
save 1 life from nonsurgical management versus kyphoplasty was 14.8 at year 1 and
11.9 at year 5.[42 ] Finally, Hinde et al, in a 2020 systematic review involving over two million patients,
showed that patients who underwent vertebral augmentation procedures for treatment
of OVCF were 22% less likely to expire 10 years postintervention.[43 ] Thus, while some patients undergo complications from vertebral augmentation, for
most people suffering from OVCF, it remains a stronger alternative to nonoperative
management.
This study has multiple limitations. Our findings depend on the accuracy of deidentified
spreadsheets from the ACS-NSQIP database without the possibility of verification through
viewing original patient charts. Therefore, it is possible that patients may have
been improperly included or excluded due to potential errors within the ACS-NSQIP
database. While the NSQIP database is frequently utilized,[44 ]
[45 ] there was no obvious way to correct for this. Furthermore, the NSQIP database does
not provide information with regard to the specialty of the treating physician; it
is possible that the outcomes of augmentation procedures conducted by interventional
radiologists differ from those performed by surgeons or other subspecialists. It also
is unable to provide information regarding adherence (or lack thereof) by clinicians
to standardized care pathways such as SIR, SNIS, or the UCLA/Rand appropriateness
method. These guidelines help ensure the standard of care is followed when treating
patients.
From a more technical perspective, another challenge with the NSQIP database is its
use of the term “null” instead of “no” when identifying patients with CNS tumors.
While both terms are used throughout the database, the former term, in comparison
to the latter, does not provide the same degree of certainty that a patient does not
have a CNS tumor. In this study, both terms were considered acceptable for purposes
of meeting inclusion criteria. Finally, variables recorded in ACS-NSQIP changed from
year to year; periodically, some were added while others were removed. Although this
constraint was carefully tracked and accounted for, it necessitated the exclusion
of some variables from our regression analysis. Our analysis of 30-day readmission
rates is also limited by the database. ACS-NSQIP defines a 30-day readmission from
the date of the procedure and not the date of discharge.[46 ] Therefore, a sick patient discharged 3 weeks after the index procedure has only
9 days postdischarge to qualify for a 30-day readmission using this methodology. On
the contrary, a patient discharged 2 days following an uncomplicated procedure has
28 days to be readmitted to be considered a 30-day readmission. This record-keeping
limitation may have impacted our ability to fully assess the relationship between
30-day readmission and the comorbidities considered. Moreover, ACS-NSQIP does not
track readmissions that occur after the initial 30-day period following a procedure.
Other limitations include the decision to avoid distinguishing between different procedural
modalities. This study did not control for whether a patient underwent a multilevel
or a single-level procedure in treating an OVCF. It also did not distinguish between
cement augmentation techniques. In this database study, we relied on the clinical
judgment of the clinician to identify which procedure would be best for a given patient,
without controlling for the modality utilized. Future studies could compare other
factors that may be statistically associated with adverse outcomes in the setting
of OVCF, such as inpatient status prior to procedure or postoperative disposition.
It may also be valuable to investigate the correlation between the factors considered
in this study and other spine pathologies.
Conclusion
Data from this study may help clinicians identify comorbidities that can significantly
affect outcomes. By quantifying the risk associated with these comorbidities, clinicians
may be better equipped to guide the expectations of their patients and form treatment
plans. Of the 15 comorbidities considered in this study, four were statistically associated
with complications (preoperative functional status; ASA class > 2; history of coagulopathy;
history of CVA), two were associated with increased rates of mortality (chronic steroid
use; ASA class > 2), and four were associated with increased rates of readmission
(COPD; obesity; weight loss; ASA class >2). A history of CHF, DM, hypertension, ascites,
renal failure, and smoking were not found to be associated with these adverse outcomes.
