- Email: [email protected]
Little Insights from Big Data: Cerebrospinal Fluid Leak After Skull Base Surgery and the Limitations of Database Research
Original Article
Little Insights from Big Data: Cerebrospinal Fluid Leak After Skull Base Surgery and the Limitations of Database Research Avital Perry1, Panagiotis Kerezoudis1,3, Christopher S. Graffeo1, Lucas P. Carlstrom1, Maria Peris-Celda1, Fredric B. Meyer1, Mohamad Bydon1,3, Michael J. Link1,2
BACKGROUND: Cerebrospinal fluid (CSF) leak is a frustrating complication of skull base surgery. Published methodologies using national surgical databases to assess CSF leak have not accounted for variability between skull base operations.
-
OBJECTIVE: Our goal was to attempt the development of a novel framework for adapting big data techniques to skull base surgery and assess the reliability of corresponding data manipulations.
-
METHODS: A retrospective nested case-control analysis was performed using patients from the National Surgical Quality Improvement Program (NSQIP) registry, 2012e2015. Current Procedural Terminology and International Classification of Diseases, Ninth Revision codes identified possible skull base operations, which were systematically grouped by anatomic location. Meningioma, schwannoma, pituitary adenoma, and trigeminal neuralgia (TN) were included.
-
RESULTS: Of 2918 patients, 84 (2.9%) were readmitted/ reoperated on within 30 days for CSF leak. Operations involving the anterior fossa, both middle/posterior fossas in 1 approach, or the orbitocranial zygomatic approach were significantly associated with CSF leak, as were schwannomas and meningiomas in any location (8.5%, 3.1%, 10.2%, 4.1%, and 3.0%; all P < 0.0001). Multivariate analysis of only
middle/posterior fossa lesions identified schwannoma (odds ratio [OR], 2.7; 95% confidence interval [CI], 1.3e5.6; P [ 0.008), TN (OR, 5.4; 95% CI, 2e14.7; P [ 0.008), chronic obstructive pulmonary disease (OR, 3.9; 95% CI, 1.1e14; P [ 0.03), and increased operative time (OR, 4.0; 95% CI, 1.7e9.5; P [ 0.009) as significant CSF leak risk factors. CONCLUSIONS: Based on NSQIP data analyzed using a rational skull base/anatomic framework, risk factors for postoperative CSF leak include chronic obstructive pulmonary disease, operative time, anterior fossa meningioma, and middle/posterior fossa schwannoma or TN. Although databases such as NSQIP can be extensively manipulated to generate surrogate results that may provide limited insight, applications beyond their design should be approached carefully.
-
-
Key words Big data - Cerebrospinal fluid leak - Meningioma - NSQIP - Pituitary adenoma - Schwannoma - Skull base surgery - Trigeminal neuralgia -
Abbreviations and Acronyms BMI: Body mass index CI: Confidence interval COPD: Chronic obstructive pulmonary disease CPT: Current Procedural Terminology CSF: Cerebrospinal fluid ICP: Intracranial pressure NSQIP: National Surgical Quality Improvement Program
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
INTRODUCTION
C
erebrospinal fluid leak (CSF) leak is a potentially morbid complication of skull base surgery. Although it has been widely studied, previously reported rates of CSF leak have been variable. For example, vestibular schwannoma resections are posterior fossa surgeries believed to carry moderate risk of
OR: Odds ratio OZ: Orbitocranial zygomatic SSI: Surgical site infection TN: Trigeminal neuralgia From the Departments of 1Neurologic Surgery and 2OtolaryngologyeHead and Neck Surgery and 3Mayo Clinic Neuro-Informatics Laboratory, Mayo Clinic, Rochester, Minnesota, USA To whom correspondence should be addressed: Michael J. Link, M.D. [E-mail: [email protected]] Avital Perry and Panagiotis Kerezoudis contributed equally to this article. Citation: World Neurosurg. (2019) 127:e561-e569. https://doi.org/10.1016/j.wneu.2019.03.207 Journal homepage: www.journals.elsevier.com/world-neurosurgery Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2019 Elsevier Inc. All rights reserved.
www.journals.elsevier.com/world-neurosurgery
e561
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
postoperative CSF leak, particularly compared with higher-risk expanded endonasal operations, yet the published leak rates overlap significantly, with ranges of 3%e13.4% and 0.5%e15%, respectively.1-10 Rigorous interrogation of this question is challenging, because of the low incidence of CSF leak, and differences in reporting practices between centers. Correspondingly, techniques using big data (data characterized by large volume/sample size, variety in collected variables, and velocity regarding analytical processing) to identify risk factors for CSF leak and stratify preoperative risk is of great interest, particularly given the associations between CSF leak and subsequent adverse events, including increased length of hospital stay, hospital readmission, and meningitis.11,12 This is especially the case given that most large series reporting CSF leak after skull base surgery have been either single-institution or single-surgeon series, which, although well executed, frequently lack the external validity required to provide community practice or younger neurosurgeons with a reliable benchmark.3,13-15 Most importantly, preceding big data analyses have not identified a methodology integrating the critical differences between the approach techniques and anatomic disease locations that are inherent to skull base diseases.3,6,12 Often, many operations are pooled without discriminating between potentially meaningful subpopulations (e.g., expectations for a skull base meningioma vary significantly between an olfactory groove vs. cerebellopontine angle tumor). Correspondingly, our key objective was to attempt to develop the first rational framework for adapting big data techniques to skull base surgery, and in doing so, show the limitations of databases when applied to skull base diseases.
nerve injury, cardiac arrest, myocardial infarction, sepsis, or septic shock. Minor complications included deep vein thrombosis without pulmonary embolism, urinary tract infection, or superficial SSI.
METHODS Data Source and End Points We queried the National Surgical Quality Improvement Program (NSQIP; American College of Surgeons, Chicago, Illinois, USA), 2012e2015. The primary study outcome was CSF leak requiring readmission/reoperation within 30 days, identified using Current Procedural Terminology (CPT; American Medical Association, Chicago, Illinois, USA) codes (61618, 61619, 62100, 62120, 62121, 62273, 63707, and 63709), and International Classification of Diseases, Ninth and Tenth Revisions (World Health Organization, Geneva, Switzerland) edition diagnosis codes (349, 349.31, 349.81, 388.6, 388.61, 998.13, G96.0). Secondary outcomes and covariates of interest included age, race, gender, body mass index (BMI), functional status, American Society of Anesthesiologists classification, smoking status, diabetes, chronic corticosteroid use, index operative time, and index length of stay.16 Captured concomitant conditions included history of congestive heart failure, angina, myocardial infarction, or previous percutaneous coronary intervention; impaired sensorium, hemiplegia, paraplegia, or quadriplegia; cerebrovascular accident with or without residual neurologic deficit; chronic obstructive pulmonary disease (COPD) and/or ventilator dependency; chronic kidney disease and/or dialysis dependency; ascites or esophageal varices; and history of bleeding disorder. Major complications captured were deep surgical site infection (SSI), organ-space infection, pneumonia, reintubation, pulmonary embolism, failure to wean ventilatory support, renal failure/insufficiency, stroke, coma, cranial
e562
www.SCIENCEDIRECT.com
Inclusion and Exclusion Criteria Patients were identified using the CPT approach and procedure codes. Two authors (A.P. and C.S.G.) independently reviewed all possible approach CPTs to identify candidate codes designating skull base operations and sorted candidate codes by anatomic locations (e.g., anterior, middle, or posterior fossa, alone or in combination; Table 1). Disagreements were resolved via secondary review (M.J.L.). By convention, if a CPT code was ambiguous regarding which cranial fossa was accessed, the more inclusive designations would be selected. For example, although a code such as “60608: Resection or excision of neoplastic, vascular or infectious lesion of parasellar area, cavernous sinus, clivus or midline skull base; intradural, including dural repair, with or without graft” could be billed for an isolated cavernous sinus tumor (e.g., middle fossa), it could also be billed for a midclival or lower clival lesion (e.g., posterior fossa), and therefore it received a combined designation (“middle and/or posterior”). Only one code (“61592: Orbitocranial zygomatic approach”; OZ) was designated as potentially involving all 3 fossas. All patients with included codes were then sorted into mutually exclusive groups by anatomic location (Figure 1). We documented the placement of lumbar drainage system (CPT code 62272) as well as the use of skull base reconstruction techniques (CPT codes 20926, 61618, 61619) during the same odds ratio (OR) (Supplementary Table 1). International Classification of Diseases, Ninth and Tenth Revisions diagnosis codes were used to subgroup cases into 4 major diagnoses of interest: schwannoma, meningioma, pituitary tumor, or trigeminal neuralgia (TN; Supplementary Table 2). Exclusions included chemotherapy within 30 days, radiotherapy within 90 days, sepsis, disseminated cancer, emergent index operation, index surgeon other than neurosurgeon/otolaryngologist, or other previous
Table 1. Summary of Current Procedural Terminology Codes Identified as Candidate Codes for Skull Base Approaches, Sorted by Final Categorization Cranial Fossa
Code
Anterior
61330, 61580, 61581, 61582, 61583, 61584, 61585, 61586, 61600, 61601
Middle
61590, 69970, 61548
Posterior Anterior or middle
61305, 61518, 61519, 61520, 61521, 61524, 61526, 61575, 61576, 61595, 61596, 61597, 61615, 61616 Not applicable
Anterior or posterior
Not applicable
Middle or posterior
61598, 61605, 61606, 61607, 61608, 61450, 61458, 61460, 61530, 61591
Anterior or middle or posterior
61592
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
Figure 1. Graphic representation of National Surgical Quality Improvement Program (NSQIP) search results and subsequent application of study inclusion/exclusion
operation within 30 days. Given the small cohort sizes, subgroups identified with anterior-middle (n ¼ 2) or anteroposterior (n ¼ 5) combinations of cranial fossa codes were excluded.
Statistical Analysis Descriptive statistics included means with standard deviations (continuous) and frequencies with proportions (categorical). Comparisons were conducted using unpaired 2-tailed Student t test for (continuous), or Pearson c2/Fisher exact test (categorical). American Society of Anesthesiologists class was dichotomized (low [I/II] vs. high [III/IV]). Multivariable logistic regression models could be fitted for CSF leak only after middle, posterior, or middle-posterior fossa surgery, because they were the only anatomic subgroups including instances of all 4 diseases. Variables included in the model were chosen a priori based on preceding literature and clinical significance. Model adjustments included BMI, operative time, COPD, procedure type, and diagnosis. Operative time and BMI were modeled using restricted cubic splines with 3 knots, to allow for a more flexible relationship with the binary outcome of interest. Collinearity was examined using the variance inflation factor. Model discrimination was assessed using the c statistic. Statistical analysis was performed using open-source software (R Core Team [2015]. R: A language and environment for statistical computing [R Foundation for Statistical Computing, Vienna, Austria]; https://www.R-project.org/) and the rms package.17 STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines were used as appropriate (Supplementary Document 1). Statistical significance was determined using a threshold of a ¼ 0.05.
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
criteria, resulting in a definitive cohort of 2918 unique individual patients.
RESULTS Search Results, Baseline Demographics, and Clinical Characteristics A total of 6970 patients had 1 included CPT code (Table 1), of whom 2918 met inclusion/exclusion criteria (Figure 2). By our anatomic classification scheme, cases were coded as strictly anterior, middle, or posterior fossa operations in 130, 622, and 1,457 cases, respectively. An additional 611 patients were coded
Figure 2. Included patients were subgrouped by skull base anatomic location derived from Current Procedural Terminology code categorizations, as described in the Methods section (see also Table 1). Patients with anterior-middle or anteroposterior categorization were excluded, given the small n in these subgroups.
www.journals.elsevier.com/world-neurosurgery
e563
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
as middle-posterior fossa approaches, whereas 98 operations were coded as OZ operations potentially involving all 3 cranial fossas. Diagnoses included 856 schwannomas, 1000 meningiomas, 670 pituitary tumors, and 412 TNs. Overall, 84 patients (2.9%) were identified with readmissions/ reoperations within 30 days for CSF leak; these patients were compared with the remaining 2834 control patients. Median times to readmission or reoperation were 13 and 11 days, respectively. Baseline demographics and comorbidities were nonsignificant between the 2 cohorts (Table 2). However, a significantly higher proportion of patients with CSF leak had a history of COPD (6% vs. 1.4%; P ¼ 0.009). Operative Variables and Postoperative Outcomes Mean index operative time was significantly longer in patients who later had a readmission/reoperation for CSF leak (Table 3; 283 vs. 385 minutes; P < 0.001). By diagnosis, CSF leak rates were schwannoma 4.1%, meningioma 3%, pituitary tumor 1.2%, and TN 2.7% (P < 0.001). By anatomic approach group, CSF leak rates were markedly higher in both cohorts involving the anterior fossa, with patients who underwent anterior-middleposterior surgery (10.2%) and anterior (8.5%) operations 2-fold to 3-fold more likely to have a CSF leak than those with index operations categorized as middle-posterior (3.1%), posterior (2.5%), or middle (1.3%; overall P < 0.001) fossa procedures. Furthermore, the CSF leak group had higher rates of lumbar drain placement (8% vs. 3%; P ¼ 0.02) as well as skull base reconstruction techniques (47% vs. 15%; P < 0.001). CSF leak was associated with other complications, which occurred in 19% of the CSF leak cases, compared with 6.3% in the controls (Table 4; P ¼ 0.007). In particular, cases complicated by CSF leak were significantly more likely to develop sepsis/septic shock (7.1% vs. 1.4%; P ¼ 0.007), deep SSI (4.8% vs. 0.7%; P ¼ 0.004), or organspace SSI (3.6% vs. 0.5%). Surprisingly, index length of stay was not significantly different between the groups (5.11 vs. 6.40 days; P ¼ 0.08). Subgroup Analysis and Multivariable Regression Subgroup analysis of CSF leaks by disease within each anatomic region showed that, among posterior fossa CSF leaks, 81% occurred in patients with schwannoma (Table 5; P ¼ 0.002). Within the other anatomic regions, the differences in CSF leak rates between diseases were not significant. Multivariable analysis included those patients with middle, posterior, or middle-posterior disease processes (Table 6). Within these subgroups, COPD (OR, 3.98; 95% confidence interval [CI], 1.13e 13.98; P ¼ 0.03), increased operative time (OR, 4.03; 95% CI, 1.71e 9.47; P ¼ 0.009), and schwannoma (OR, 2.69; 95% CI, 1.30e5.57; P ¼ 0.008) or TN (OR, 5.42; 95% CI, 2e5.14.7; P ¼ 0.008) were associated with significantly higher odds of a postoperative CSF leak.
Table 2. Patient Demographics and Comorbidities No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Variable Age (years), mean (standard deviation)
54.4 (14.3)
52.0 (14.3)
0.14
Female sex
1820 (64.2)
50 (59.5)
0.44
White race
2023 (83.1)
61 (80.3)
0.61
Hispanic ethnicity
242 (9.63)
3 (4.00)
0.15
Body mass index (kg/m ), mean (standard deviation)
30.1 (6.96)
31.1 (6.86)
0.23
Body mass index >30 kg/m2 (obese)
1243 (44.2)
41 (48.8)
0.47
2
>0.99
Functional status Independent
2771 (98.1)
82 (98.8)
Partial dependent
51 (1.80)
1 (1.20)
Totally dependent
4 (0.14)
0 (0.00)
American Society of Anesthesiologists classification
0.92
Low
1382 (48.8)
40 (47.6)
High
1452 (51.2)
44 (52.4)
Smoker
367 (12.9)
14 (16.7)
0.41
326 (11.5)
10 (11.9)
>0.99
Comorbidities History of myocardial infarction
4 (0.14)
0 (0.00)
>0.99
1088 (38.4)
33 (39.3)
>0.99
Previous percutaneous coronary intervention
5 (2.16)
0 (0.00)
>0.99
Peripheral vascular disease
2 (0.86)
0 (0.00)
>0.99
Chronic obstructive pulmonary disease
40 (1.41)
5 (5.95)
0.009
Ventilator dependent
2 (0.07)
0 (0.00)
>0.99
Impaired sensorium
1 (0.43)
0 (0.00)
>0.99
Congestive heart failure Hypertension
Stroke
2 (0.86)
0 (0.00)
>0.99
On dialysis
1 (0.04)
0 (0.00)
>0.99
Bleeding diathesis
26 (0.92)
1 (1.19)
0.55
Need of chronic transfusion
1 (0.04)
0 (0.00)
>0.99
Ascites
0 (0.00)
0 (0.00)
N/A
Esophageal varices
0 (0.00)
0 (0.00)
N/A
DISCUSSION
Diabetes mellitus
326 (11.5)
10 (11.9)
>0.99
CSF leak is a known complication in skull base surgery, inciting a range of significant and potentially morbid sequelae. CSF leak has also proved an elusive end point to characterize in a precise, accurate, and generalizable fashion. Whereas single-institution or
On chronic steroids
199 (7.02)
7 (8.33)
0.81
e564
www.SCIENCEDIRECT.com
Values are number (%) except where indicated otherwise. Bold values denote statistical significance.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
Table 3. Summary of Operative Variables
Variable
Table 4. Postoperative Outcomes
No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Diagnoses, row Meningioma
0.012 970 (97)
30 (3)
Pituitary adenoma
642 (98.8)
8 (1.2)
Schwannoma
821 (95.9)
35 (4.1)
Trigeminal neuralgia
401 (97.3)
11 (2.7) <0.001
Procedures, row Anterior
119 (91.5)
11 (8.5)
Middle
614 (98.7)
8 (1.3)
Posterior
1421 (97.5)
36 (2.5)
Middle-posterior
592 (96.9)
19 (3.1)
Anterior-middle-posterior
88 (89.8)
10 (10.2)
Operative time (minutes), mean (standard deviation)
283 (172)
385 (180)
<0.001
Lumbar drain placement
88 (3.11)
7 (8.33)
0.02
Skull base reconstruction
421 (14.9)
40 (47.6)
<0.001
Values are number (%) except where indicated otherwise. Bold values denote statistical significance.
single-surgeon series reflect selected outcomes and a small number of patients, preceding big data investigations may have been limited in the granularity of their inclusion criteria by the nature of the database, particularly in the context of highly complicated and individualized skull base operations. We initially sought to design a novel framework for mapping CPT codes to skull base anatomic locations, through which we studied 4 prevalent diseases at significant risk of postoperative CSF leak in almost 3000 patients using NSQIP. However, the lack of anatomic granularity in CPT coding for skull base tumors made using a CPT codeebased data set difficult. Although some useful indirect information may be derived from our analysis and similar frameworks in the future, our overarching conclusion is that big data may not offer more than a little insight for the skull base surgeon. Regarding the study results, we noted that operations involving the anterior fossa (in particular, anterior fossa meningiomas) were at highest risk for CSF leak overall, alongside any operations that included a code for the OZ approach. Within the subset of operations limited to the middle and posterior fossas, we identified COPD, increased operative time, and the diagnoses of schwannoma or TN as significant risk factors for CSF leak. This situation fits with typical dogma, although the 2.9% incidence of CSF is lower than expected.1-10 There are several potential explanations for this discrepancy. Chiefly, NSQIP does not capture readmissions/reoperations that occur beyond 30 days after the index operation. In addition, NSQIP coding procedures have been shown to be subject to inaccuracies that may affect the reliability of the database, particularly for analysis of skull base tumor
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
Variable
No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Length of stay (days), mean (standard deviation)
5.11 (6.31)
6.40 (6.64)
0.08
Major complication
269 (9.49)
16 (19.0)
0.007
Sepsis/septic shock
39 (1.37)
6 (7.14)
<0.001
Stroke
39 (1.38)
1 (1.19)
>0.99
Organ-space SSI
19 (0.67)
4 (4.76)
0.004
Deep incisional SSI
14 (0.49)
3 (3.57)
0.012
Cardiac arrest
9 (0.32)
0 (0.00)
>0.99
On ventilation for >48 hours
61 (2.15)
2 (2.38)
0.25
Reintubation
57 (2.01)
3 (3.57)
0.25
Pneumonia
44 (1.55)
3 (3.57)
0.15
Pulmonary embolism
38 (1.34)
2 (2.38)
0.32
Bleeding requiring transfusion
123 (4.34)
5 (5.95)
0.42
Renal insufficiency
3 (0.11)
0 (0.00)
>0.99
Myocardial infarction
2 (0.07)
0 (0.00)
>0.99
Any minor complication
98 (3.46)
8 (9.52)
0.01
Urinary tract infection
46 (1.62)
0 (0.00)
0.642
Wound disruption
6 (0.21)
7 (8.33)
<0.001
Deep venous thrombosis
39 (1.38)
0 (0.00)
>0.99
Superficial SSI
11 (0.39)
2 (2.38)
0.052
Values are number (%) except where indicated otherwise. Bold values denote statistical significance. SSI, surgical site infection.
cases.18-20 Spurious coding and billing documentation practices embedded within the medical record by clinicians or support staff may have introduced 2 additional sources of bias: application of a skull base approach code to a simpler, lower-risk operation for fiduciary gain or inaccurate coding of unplanned take-back operations, to falsely improve quality metrics. Big data are not able to give us granular results in this instance. However, they do have the ability to zoom out and provide an abstract perspective on trends that, although perhaps inaccurate for any given patient, would not have become apparent without the aid of a large patient sample. An interesting and novel finding of the present analysis is that COPD may be an independent risk factor for CSF leak: a pairing of morbidity and complication sufficiently rare to escape notice in even large single-institution series, but one that may provide some insight into the pathophysiology of CSF leak. We propose that, within this subset of patients, coughing is an important and underappreciated risk factor for the development of CSF leak. In the COPD population at large, chronic cough has a prevalence of at least 55%, a percentage that would be expected to
www.journals.elsevier.com/world-neurosurgery
e565
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
Table 5. Incidence of Cerebrospinal Fluid Leak by Procedure Group
Diagnosis
No Cerebrospinal Fluid Leak, n (%)
Cerebrospinal Fluid Leak, n (%)
Anterior
n ¼ 119
n ¼ 11
Meningioma
103 (86.6)
11 (100)
Pituitary
15 (12.6)
0 (0.00)
Schwannoma
1 (0.84)
0 (0.00)
n ¼ 614
Middle Meningioma Pituitary
P Value
0.42
Table 6. Multivariable Logistic Regression Fitted for Occurrence of Cerebrospinal Fluid Leak after Surgery in the Middle and Posterior Cranial Fossae Odds Ratio
95% Confidence Interval
P Value
Body mass index
1.24
0.82e1.86
0.55
Presence of chronic obstructive pulmonary disease
3.98
1.13e13.98
0.03
Operative time
4.03
1.71e9.47
0.009
Variable
Diagnosis
n¼8 0.17
Meningioma
7 (87.5)
Schwannoma
2.69
1.30e5.57
0.008
5.42
2.0e14.7
0.009
2.19
0.74e6.50
0.16
3 (0.49)
0 (0.00)
Trigeminal neuralgia
9 (1.47)
1 (12.5)
Pituitary tumor
n ¼ 1421 714 (50.2)
n ¼ 36 7 (19.4)
Pituitary
10 (0.70)
0 (0.00)
Schwannoma
684 (48.1)
29 (80.6)
Trigeminal neuralgia Middle-posterior Meningioma Pituitary Schwannoma Trigeminal neuralgia Anterior-middle-posterior Meningioma
13 (0.91) n ¼ 592 78 (13.2)
3 (15.8)
8 (1.35)
0 (0.00) 6 (31.6)
379 (64.0)
74 (84.1)
0.002
0 (0.00)
127 (21.5) n ¼ 88
Bold values denote statistical significance.
n ¼ 19 0.53
10 (52.6) n ¼ 10 9 (90.0)
Pituitary
8 (9.09)
1 (10.0)
Schwannoma
6 (6.82)
0 (0.00)
>0.99
Bold values denote statistical significance.
increase dramatically in a hospitalized or a postanesthesia cohort.21,22 Mechanistically, coughing is a well-described cause of transient intracranial pressure (ICP) increases, and one can easily imagine how relentless ICP spikes would strain the dural repair, potentially precipitating a CSF leak.23 Secondary mechanisms in addition to cough may also contribute to COPD-associated CSF leak. Previous investigators have investigated the relationship between chronic lung disease and ICP, postulating that venous obstruction and secondary polycythemia caused by chronic hypoxia increase arterial and venous pressures, and consequently the ICP.24 Alternatively, COPD may be a marker for another, more insidious process driving CSF leak in these patients. One such example is obstructive sleep apnea, a common comorbidity in COPD, independent of obesity status.25 These patients show an additive chronic hypoxia and are therefore expected to have amplified ICP derangements.
e566
Reference
601 (97.9)
Schwannoma
Meningioma
Reference
0 (0.00)
Trigeminal neuralgia
Posterior
Reference
1 (0.16)
www.SCIENCEDIRECT.com
By contrast, where do big data perform poorly? They lack the ability to zoom in, whether on a particular observation or on a question that emerges from the analysis. Within our analysis, the nonsignificant difference in BMI is a curious wrinkle. In a preceding NSQIP study, Murphy et al.12 identified BMI as a predictor of 30-day postoperative complications after surgery for “benign cranial nerve neoplasms,” including CSF leak. Another recent study by Alattar et al.6 investigated almost 7000 cases of vestibular schwannoma surgery in the California Office of Statewide Health Planning and Development database and similarly identified BMI as a significant risk factor for CSF leak. Further still, in a large, single-institution series of 457 vestibular schwannoma cases operated on by the senior author (M.J.L.), Copeland et al.3 completed a risk-adjusted analysis that identified BMI, operative time, and translabyrinthine approach as independent risk factors for CSF leak. These results make a compelling case for the role of BMI as a key CSF leak risk factor, which accords with previously described BMI-induced pathophysiologic ICP changes, as well as anecdotal experience.26 That we did not reproduce this finding is likely attributable to above-average BMI in both groups (32 among CSF leak cases; 30 among controls), which is potentially an unintended consequence of our stringent inclusive/exclusion criteria. Without access to the medical records of these patients, that hypothesis will remain untested, and the question unanswered. Another example of the inability to drill down is our finding that operative time was significantly longer in the CSF leak group, by a mean 102 minutes. Numerous common and uncommon contributors to longer operative times can be imagined as motivating this finding, many of which would constitute major confounders (e.g., a larger tumor, a tumor that required more extensive bony removal to expose, an intraoperative complication or complex reconstruction that was not reflected in a specific billing code, and so on). This difficulty in measuring important confounders is reflected in the paradoxically higher rates of lumbar drain placement and use of skull base reconstruction methods in the CSF leak group. Both of these techniques have been reported to be
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
associated with decreased rates of CSF leak based on anecdotal evidence and previous reports.27-29
specialists, who undergo training and periodic auditing, but their submissions are dependent on the source data coded by the operating physician or their administrative team, allowing for several additional sources of bias, as mentioned earlier.
Strengths and Limitations The core strengths of our study lie in its novel methodology. We developed a new and more comprehensive approach to classifying patients derived from large national data sets by categorizing CPT codes based on a strictly anatomic scheme. This approach was further strengthened by our meticulous and redundant review process built around consensus decision making, which is a level of internal validity not typically seen in preceding studies derived from administrative databases.30 Notwithstanding, our study is subject to several major limitations. First, as we described in detail earlier, NSQIP and other national surgical registries are not designed to assess skull base patients or diseases. For instance, variables that are highly relevant to an analysis of CSF leak are absent from NSQIP (histopathology, tumor size and characteristics, extent of resection, and disease laterality, to name only a few of the most notable absences). Second, the originating centers and surgeons are deliberately obscured, and so there is no way to measure or account for differences in CSF leak rates between providers, which may not follow a normal distribution. A recent study by Babadjouni et al. showed that increased procedural volume is associated with lower 30-day and 90-day readmission rate after microsurgical resection of vestibular schwannoma.31 This important observation might also reflect not only the ability of centers with greater cumulative experience to manage postoperative complications of skull base surgery in the outpatient rather than inpatient setting but also the critical role of the learning curve in the repair process.31 Third, NSQIP captures data only within a 30-day window, excluding late leaks, which are a rare but appreciable subgroup.32 Fourth, lumbar drain insertion rates were underestimated, because bedside procedures are not reported in the registry. NSQIP variables are captured and entered by nurse
REFERENCES 1. Rodgers GK, Luxford WM. Factors affecting the development of cerebrospinal fluid leak and meningitis after translabyrinthine acoustic tumor surgery. Laryngoscope. 1993;103:959-962. 2. Bryce GE, Nedzelski JM, Rowed DW, Rappaport JM. Cerebrospinal fluid leaks and meningitis in acoustic neuroma surgery. Otolaryngol Head Neck Surg. 1991;104:81-87. 3. Copeland WR, Mallory GW, Neff BA, Driscoll CLW, Link MJ. Are there modifiable risk factors to prevent a cerebrospinal fluid leak following vestibular schwannoma surgery? J Neurosurg. 2015;122:312-316. 4. Bani A, Gilsbach JM. Incidence of cerebrospinal fluid leak after microsurgical removal of vestibular schwannomas. Acta Neurochir (Wien). 2002;144: 979-982 [discussion: 982]. 5. Selesnick SH, Liu JC, Jen A, Newman J. The incidence of cerebrospinal fluid leak after
CONCLUSIONS Within the limited scope of our model, we identified anterior fossa operations and complex approaches as significantly associated with higher CSF leak rates. In addition, among operations in the middle or posterior fossas, patients with schwannoma were at markedly higher risk than were patients with TN, who were in turn significantly higher risk than patients with meningioma or pituitary adenoma. With respect to risk factors, COPD and increased operative time independently had a significant association with CSF leak and show the inherent strengths and weaknesses of big data analyses. The COPD association is a particularly interesting and potentially meaningful discovery, because it seems to be underrecognized clinically and may not have come to light without big data power. By contrast, the observation of an association between increased operative time and risk of CSF leak is almost certainly confounded by several variables and difficult to interpret without more closely scrutinizing the cases. At its core, our analysis was designed as an attempt to transform data from a pragmatic coding system that is built to describe the labor performed in a way that it accurately describes the tumors on which that work was performed. Although we have produced a more sophisticated extraction and parsing system that certainly constitutes a marked improvement on preceding big data practices in the skull base literature, we are still essentially left with an unsatisfying solution. Our assessment is that the degree of granularity needed to fully answer clinically meaningful questions about complex cranial operations is simply not accessible via the big data model.
vestibular schwannoma surgery. Otol Neurotol. 2004;25:387-393. 6. Alattar AA, Hirshman BR, McCutcheon BA, et al. Risk factors for readmission with cerebrospinal fluid leakage within 30 days of vestibular schwannoma surgery. Neurosurgery. 2018;82: 630-637. 7. Shiley SG, Limonadi F, Delashaw JB, et al. Incidence, etiology, and management of cerebrospinal fluid leaks following trans-sphenoidal surgery. Laryngoscope. 2003;113:1283-1288. 8. Seiler RW, Mariani L. Sellar reconstruction with resorbable vicryl patches, gelatin foam, and fibrin glue in transsphenoidal surgery: a 10-year experience with 376 patients. J Neurosurg. 2000;93: 762-765. 9. Ciric I, Ragin A, Baumgartner C, Pierce D. Complications of transsphenoidal surgery: results of a national survey, review of the literature, and personal experience. Neurosurgery. 1997;40:225-236 [discussion: 236-237].
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
10. Black PM, Zervas NT, Candia GL. Incidence and management of complications of transsphenoidal operation for pituitary adenomas. Neurosurgery. 1987;20:920-924. 11. McCutcheon BA, Orosco RK, Chang DC, et al. Outcomes of isolated basilar skull fracture: readmission, meningitis, and cerebrospinal fluid leak. Otolaryngol Head Neck Surg. 2013;149:931-939. 12. Murphy ME, McCutcheon BA, Kerezoudis P, et al. Morbid obesity increases risk of morbidity and reoperation in resection of benign cranial nerve neoplasms. Clin Neurol Neurosurg. 2016;148:105-109. 13. Samii M, Matthies C. Management of 1000 vestibular schwannomas (acoustic neuromas): hearing function in 1000 tumor resections. Neurosurgery. 1997;40:248-260 [discussion: 260-262]. 14. Brackmann DE, Owens RM, Friedman RA, et al. Prognostic factors for hearing preservation in vestibular schwannoma surgery. Am J Otol. 2000; 21:417-424.
www.journals.elsevier.com/world-neurosurgery
e567
ORIGINAL ARTICLE AVITAL PERRY ET AL.
15. Becker SS, Jackler RK, Pitts LH. Cerebrospinal fluid leak after acoustic neuroma surgery: a comparison of the translabyrinthine, middle fossa, and retrosigmoid approaches. Otol Neurotol. 2003;24: 107-112. 16. Murphy M, Gilder H, McCutcheon BA, et al. Increased operative time for benign cranial nerve tumor resection correlates with increased morbidity postoperatively. J Neurol Surg B Skull Base. 2016;77:350-357. 17. Harrell FE Jr. rms: Regression Modeling Strategies. R package version 4.0-0. Cityscape; 2013. Available at: https://CRAN.R-project.org/package=rms. Accessed January 4, 2018. 18. Rolston JD, Han SJ, Chang EF. Systemic inaccuracies in the National Surgical Quality Improvement Program database: implications for accuracy and validity for neurosurgery outcomes research. J Clin Neurosci. 2017;37:44-47. 19. Bi WL, Mooney MA, Yoon S, et al. Variation in Coding Practices for Vestibular Schwannoma Surgery. J Neurol Surg B Skull Base. 2019;80:96-102. 20. Oravec CS, Motiwala M, Reed K, et al. In Reply: Big Data Research in Neurosurgery: A Critical Look at this Popular New Study Design. Neurosurgery. 2018;82:E188-E189. 21. Calverley PM. Cough in chronic obstructive pulmonary disease: is it important and what are the effects of treatment? Cough. 2013;9:17. 22. Kessler R, Partridge MR, Miravitlles M, et al. Symptom variability in patients with severe COPD:
e568
www.SCIENCEDIRECT.com
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
a pan-European cross-sectional study. Eur Respir J. 2011;37:264-272. 23. Bloomfield GL, Ridings PC, Blocher CR, Marmarou A, Sugerman HJ. A proposed relationship between increased intra-abdominal, intrathoracic, and intracranial pressure. Crit Care Med. 1997;25:496-503. 24. Carter CC, Fuller TJ. Increased intracranial pressure in chronic lung disease. Neurology. 1957;7: 169-174. 25. Owens RL, Malhotra A. Sleep-disordered breathing and COPD: the overlap syndrome. Respir Care. 2010;55:1333-1344 [discussion: 1344-1346]. 26. Radhakrishnan K, Ahlskog JE, Cross SA, Kurland LT, O’Fallon WM. Idiopathic intracranial hypertension (pseudotumor cerebri). Descriptive epidemiology in Rochester, Minn, 1976 to 1990. Arch Neurol. 1993;50:78-80. 27. Mehta GU, Oldfield EH. Prevention of intraoperative cerebrospinal fluid leaks by lumbar cerebrospinal fluid drainage during surgery for pituitary macroadenomas. J Neurosurg. 2012;116: 1299-1303. 28. Zanation AM, Carrau RL, Snyderman CH, et al. Nasoseptal flap reconstruction of high flow intraoperative cerebral spinal fluid leaks during endoscopic skull base surgery. Am J Rhinol Allergy. 2009;23:518-521. 29. Snyderman CH, Janecka IP, Sekhar LN, Sen CN, Eibling DE. Anterior cranial base reconstruction: role of galeal and pericranial flaps. Laryngoscope. 1990;100:607-614.
30. Faciszewski T, Jensen R, Berg RL. Procedural coding of spinal surgeries (CPT-4 versus ICD-9CM) and decisions regarding standards: a multicenter study. Spine. 2003;28:502-507. 31. Babadjouni R, Wen T, Donoho DA, et al. Increased hospital surgical volume reduces rate of 30- and 90-day readmission after acoustic neuroma surgery. Neurosurgery. 2019;84:726-732. 32. Perry A, Graffeo CS, Copeland WR 3rd, et al. Delayed cerebrospinal fluid rhinorrhea after gamma knife radiosurgery with or without preceding transsphenoidal resection for pituitary pathology. World Neurosurg. 2017;100: 201-207.
Conflict of interest statement: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating it are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors. Received 9 December 2018; accepted 20 March 2019 Citation: World Neurosurg. (2019) 127:e561-e569. https://doi.org/10.1016/j.wneu.2019.03.207 Journal homepage: www.journals.elsevier.com/worldneurosurgery Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2019 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
SUPPLEMENTARY DATA
Supplementary Table 1. Skull Base Reconstruction Techniques and Corresponding Current Procedural Terminology Codes Code
Description
20926
Tissue grafts, other (e.g., paratenon, fat, dermis)
61618
Secondary repair of dura for cerebrospinal fluid leak, anterior, middle, or posterior cranial fossa after surgery of the skull base; by free tissue graft (e.g., pericranium, fascia, tensor fascia lata, adipose tissue, homologous, or synthetic grafts)
61619
Secondary repair of dura for cerebrospinal fluid leak, anterior, middle, or posterior cranial fossa after surgery of the skull base; by local or regionalized vascularized pedicle flap or myocutaneous flap (including galea, temporalis, frontalis, or occipitalis muscle)
Supplementary Table 2. Pertinent Diagnostic Codes. List of International Classification of Diseases, Ninth Revision (ICD-9) and International Classification of Diseases, Tenth Revision (ICD-10) Diagnosis Codes Used to Identify the Diagnoses of Interest Diagnosis
ICD-9 Code
ICD-10 Code
Meningioma
225.2, 237.6, 192.1
C70.0, C70.9, D32.0, D32.9, D42.0
Pituitary tumor
227.3, 237.0, 253.9, 237, 255
D35.2, D35.3, D44.3, D44.4
Schwannoma
225.1, 237.72, 237.73, 192.0
Trigeminal neuralgia
350.1
C72.50, D33.3 G50.0
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
www.journals.elsevier.com/world-neurosurgery
e569
Little Insights from Big Data: Cerebrospinal Fluid Leak After Skull Base Surgery and the Limitations of Database Research Avital Perry1, Panagiotis Kerezoudis1,3, Christopher S. Graffeo1, Lucas P. Carlstrom1, Maria Peris-Celda1, Fredric B. Meyer1, Mohamad Bydon1,3, Michael J. Link1,2
BACKGROUND: Cerebrospinal fluid (CSF) leak is a frustrating complication of skull base surgery. Published methodologies using national surgical databases to assess CSF leak have not accounted for variability between skull base operations.
-
OBJECTIVE: Our goal was to attempt the development of a novel framework for adapting big data techniques to skull base surgery and assess the reliability of corresponding data manipulations.
-
METHODS: A retrospective nested case-control analysis was performed using patients from the National Surgical Quality Improvement Program (NSQIP) registry, 2012e2015. Current Procedural Terminology and International Classification of Diseases, Ninth Revision codes identified possible skull base operations, which were systematically grouped by anatomic location. Meningioma, schwannoma, pituitary adenoma, and trigeminal neuralgia (TN) were included.
-
RESULTS: Of 2918 patients, 84 (2.9%) were readmitted/ reoperated on within 30 days for CSF leak. Operations involving the anterior fossa, both middle/posterior fossas in 1 approach, or the orbitocranial zygomatic approach were significantly associated with CSF leak, as were schwannomas and meningiomas in any location (8.5%, 3.1%, 10.2%, 4.1%, and 3.0%; all P < 0.0001). Multivariate analysis of only
middle/posterior fossa lesions identified schwannoma (odds ratio [OR], 2.7; 95% confidence interval [CI], 1.3e5.6; P [ 0.008), TN (OR, 5.4; 95% CI, 2e14.7; P [ 0.008), chronic obstructive pulmonary disease (OR, 3.9; 95% CI, 1.1e14; P [ 0.03), and increased operative time (OR, 4.0; 95% CI, 1.7e9.5; P [ 0.009) as significant CSF leak risk factors. CONCLUSIONS: Based on NSQIP data analyzed using a rational skull base/anatomic framework, risk factors for postoperative CSF leak include chronic obstructive pulmonary disease, operative time, anterior fossa meningioma, and middle/posterior fossa schwannoma or TN. Although databases such as NSQIP can be extensively manipulated to generate surrogate results that may provide limited insight, applications beyond their design should be approached carefully.
-
-
Key words Big data - Cerebrospinal fluid leak - Meningioma - NSQIP - Pituitary adenoma - Schwannoma - Skull base surgery - Trigeminal neuralgia -
Abbreviations and Acronyms BMI: Body mass index CI: Confidence interval COPD: Chronic obstructive pulmonary disease CPT: Current Procedural Terminology CSF: Cerebrospinal fluid ICP: Intracranial pressure NSQIP: National Surgical Quality Improvement Program
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
INTRODUCTION
C
erebrospinal fluid leak (CSF) leak is a potentially morbid complication of skull base surgery. Although it has been widely studied, previously reported rates of CSF leak have been variable. For example, vestibular schwannoma resections are posterior fossa surgeries believed to carry moderate risk of
OR: Odds ratio OZ: Orbitocranial zygomatic SSI: Surgical site infection TN: Trigeminal neuralgia From the Departments of 1Neurologic Surgery and 2OtolaryngologyeHead and Neck Surgery and 3Mayo Clinic Neuro-Informatics Laboratory, Mayo Clinic, Rochester, Minnesota, USA To whom correspondence should be addressed: Michael J. Link, M.D. [E-mail: [email protected]] Avital Perry and Panagiotis Kerezoudis contributed equally to this article. Citation: World Neurosurg. (2019) 127:e561-e569. https://doi.org/10.1016/j.wneu.2019.03.207 Journal homepage: www.journals.elsevier.com/world-neurosurgery Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2019 Elsevier Inc. All rights reserved.
www.journals.elsevier.com/world-neurosurgery
e561
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
postoperative CSF leak, particularly compared with higher-risk expanded endonasal operations, yet the published leak rates overlap significantly, with ranges of 3%e13.4% and 0.5%e15%, respectively.1-10 Rigorous interrogation of this question is challenging, because of the low incidence of CSF leak, and differences in reporting practices between centers. Correspondingly, techniques using big data (data characterized by large volume/sample size, variety in collected variables, and velocity regarding analytical processing) to identify risk factors for CSF leak and stratify preoperative risk is of great interest, particularly given the associations between CSF leak and subsequent adverse events, including increased length of hospital stay, hospital readmission, and meningitis.11,12 This is especially the case given that most large series reporting CSF leak after skull base surgery have been either single-institution or single-surgeon series, which, although well executed, frequently lack the external validity required to provide community practice or younger neurosurgeons with a reliable benchmark.3,13-15 Most importantly, preceding big data analyses have not identified a methodology integrating the critical differences between the approach techniques and anatomic disease locations that are inherent to skull base diseases.3,6,12 Often, many operations are pooled without discriminating between potentially meaningful subpopulations (e.g., expectations for a skull base meningioma vary significantly between an olfactory groove vs. cerebellopontine angle tumor). Correspondingly, our key objective was to attempt to develop the first rational framework for adapting big data techniques to skull base surgery, and in doing so, show the limitations of databases when applied to skull base diseases.
nerve injury, cardiac arrest, myocardial infarction, sepsis, or septic shock. Minor complications included deep vein thrombosis without pulmonary embolism, urinary tract infection, or superficial SSI.
METHODS Data Source and End Points We queried the National Surgical Quality Improvement Program (NSQIP; American College of Surgeons, Chicago, Illinois, USA), 2012e2015. The primary study outcome was CSF leak requiring readmission/reoperation within 30 days, identified using Current Procedural Terminology (CPT; American Medical Association, Chicago, Illinois, USA) codes (61618, 61619, 62100, 62120, 62121, 62273, 63707, and 63709), and International Classification of Diseases, Ninth and Tenth Revisions (World Health Organization, Geneva, Switzerland) edition diagnosis codes (349, 349.31, 349.81, 388.6, 388.61, 998.13, G96.0). Secondary outcomes and covariates of interest included age, race, gender, body mass index (BMI), functional status, American Society of Anesthesiologists classification, smoking status, diabetes, chronic corticosteroid use, index operative time, and index length of stay.16 Captured concomitant conditions included history of congestive heart failure, angina, myocardial infarction, or previous percutaneous coronary intervention; impaired sensorium, hemiplegia, paraplegia, or quadriplegia; cerebrovascular accident with or without residual neurologic deficit; chronic obstructive pulmonary disease (COPD) and/or ventilator dependency; chronic kidney disease and/or dialysis dependency; ascites or esophageal varices; and history of bleeding disorder. Major complications captured were deep surgical site infection (SSI), organ-space infection, pneumonia, reintubation, pulmonary embolism, failure to wean ventilatory support, renal failure/insufficiency, stroke, coma, cranial
e562
www.SCIENCEDIRECT.com
Inclusion and Exclusion Criteria Patients were identified using the CPT approach and procedure codes. Two authors (A.P. and C.S.G.) independently reviewed all possible approach CPTs to identify candidate codes designating skull base operations and sorted candidate codes by anatomic locations (e.g., anterior, middle, or posterior fossa, alone or in combination; Table 1). Disagreements were resolved via secondary review (M.J.L.). By convention, if a CPT code was ambiguous regarding which cranial fossa was accessed, the more inclusive designations would be selected. For example, although a code such as “60608: Resection or excision of neoplastic, vascular or infectious lesion of parasellar area, cavernous sinus, clivus or midline skull base; intradural, including dural repair, with or without graft” could be billed for an isolated cavernous sinus tumor (e.g., middle fossa), it could also be billed for a midclival or lower clival lesion (e.g., posterior fossa), and therefore it received a combined designation (“middle and/or posterior”). Only one code (“61592: Orbitocranial zygomatic approach”; OZ) was designated as potentially involving all 3 fossas. All patients with included codes were then sorted into mutually exclusive groups by anatomic location (Figure 1). We documented the placement of lumbar drainage system (CPT code 62272) as well as the use of skull base reconstruction techniques (CPT codes 20926, 61618, 61619) during the same odds ratio (OR) (Supplementary Table 1). International Classification of Diseases, Ninth and Tenth Revisions diagnosis codes were used to subgroup cases into 4 major diagnoses of interest: schwannoma, meningioma, pituitary tumor, or trigeminal neuralgia (TN; Supplementary Table 2). Exclusions included chemotherapy within 30 days, radiotherapy within 90 days, sepsis, disseminated cancer, emergent index operation, index surgeon other than neurosurgeon/otolaryngologist, or other previous
Table 1. Summary of Current Procedural Terminology Codes Identified as Candidate Codes for Skull Base Approaches, Sorted by Final Categorization Cranial Fossa
Code
Anterior
61330, 61580, 61581, 61582, 61583, 61584, 61585, 61586, 61600, 61601
Middle
61590, 69970, 61548
Posterior Anterior or middle
61305, 61518, 61519, 61520, 61521, 61524, 61526, 61575, 61576, 61595, 61596, 61597, 61615, 61616 Not applicable
Anterior or posterior
Not applicable
Middle or posterior
61598, 61605, 61606, 61607, 61608, 61450, 61458, 61460, 61530, 61591
Anterior or middle or posterior
61592
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
Figure 1. Graphic representation of National Surgical Quality Improvement Program (NSQIP) search results and subsequent application of study inclusion/exclusion
operation within 30 days. Given the small cohort sizes, subgroups identified with anterior-middle (n ¼ 2) or anteroposterior (n ¼ 5) combinations of cranial fossa codes were excluded.
Statistical Analysis Descriptive statistics included means with standard deviations (continuous) and frequencies with proportions (categorical). Comparisons were conducted using unpaired 2-tailed Student t test for (continuous), or Pearson c2/Fisher exact test (categorical). American Society of Anesthesiologists class was dichotomized (low [I/II] vs. high [III/IV]). Multivariable logistic regression models could be fitted for CSF leak only after middle, posterior, or middle-posterior fossa surgery, because they were the only anatomic subgroups including instances of all 4 diseases. Variables included in the model were chosen a priori based on preceding literature and clinical significance. Model adjustments included BMI, operative time, COPD, procedure type, and diagnosis. Operative time and BMI were modeled using restricted cubic splines with 3 knots, to allow for a more flexible relationship with the binary outcome of interest. Collinearity was examined using the variance inflation factor. Model discrimination was assessed using the c statistic. Statistical analysis was performed using open-source software (R Core Team [2015]. R: A language and environment for statistical computing [R Foundation for Statistical Computing, Vienna, Austria]; https://www.R-project.org/) and the rms package.17 STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines were used as appropriate (Supplementary Document 1). Statistical significance was determined using a threshold of a ¼ 0.05.
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
criteria, resulting in a definitive cohort of 2918 unique individual patients.
RESULTS Search Results, Baseline Demographics, and Clinical Characteristics A total of 6970 patients had 1 included CPT code (Table 1), of whom 2918 met inclusion/exclusion criteria (Figure 2). By our anatomic classification scheme, cases were coded as strictly anterior, middle, or posterior fossa operations in 130, 622, and 1,457 cases, respectively. An additional 611 patients were coded
Figure 2. Included patients were subgrouped by skull base anatomic location derived from Current Procedural Terminology code categorizations, as described in the Methods section (see also Table 1). Patients with anterior-middle or anteroposterior categorization were excluded, given the small n in these subgroups.
www.journals.elsevier.com/world-neurosurgery
e563
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
as middle-posterior fossa approaches, whereas 98 operations were coded as OZ operations potentially involving all 3 cranial fossas. Diagnoses included 856 schwannomas, 1000 meningiomas, 670 pituitary tumors, and 412 TNs. Overall, 84 patients (2.9%) were identified with readmissions/ reoperations within 30 days for CSF leak; these patients were compared with the remaining 2834 control patients. Median times to readmission or reoperation were 13 and 11 days, respectively. Baseline demographics and comorbidities were nonsignificant between the 2 cohorts (Table 2). However, a significantly higher proportion of patients with CSF leak had a history of COPD (6% vs. 1.4%; P ¼ 0.009). Operative Variables and Postoperative Outcomes Mean index operative time was significantly longer in patients who later had a readmission/reoperation for CSF leak (Table 3; 283 vs. 385 minutes; P < 0.001). By diagnosis, CSF leak rates were schwannoma 4.1%, meningioma 3%, pituitary tumor 1.2%, and TN 2.7% (P < 0.001). By anatomic approach group, CSF leak rates were markedly higher in both cohorts involving the anterior fossa, with patients who underwent anterior-middleposterior surgery (10.2%) and anterior (8.5%) operations 2-fold to 3-fold more likely to have a CSF leak than those with index operations categorized as middle-posterior (3.1%), posterior (2.5%), or middle (1.3%; overall P < 0.001) fossa procedures. Furthermore, the CSF leak group had higher rates of lumbar drain placement (8% vs. 3%; P ¼ 0.02) as well as skull base reconstruction techniques (47% vs. 15%; P < 0.001). CSF leak was associated with other complications, which occurred in 19% of the CSF leak cases, compared with 6.3% in the controls (Table 4; P ¼ 0.007). In particular, cases complicated by CSF leak were significantly more likely to develop sepsis/septic shock (7.1% vs. 1.4%; P ¼ 0.007), deep SSI (4.8% vs. 0.7%; P ¼ 0.004), or organspace SSI (3.6% vs. 0.5%). Surprisingly, index length of stay was not significantly different between the groups (5.11 vs. 6.40 days; P ¼ 0.08). Subgroup Analysis and Multivariable Regression Subgroup analysis of CSF leaks by disease within each anatomic region showed that, among posterior fossa CSF leaks, 81% occurred in patients with schwannoma (Table 5; P ¼ 0.002). Within the other anatomic regions, the differences in CSF leak rates between diseases were not significant. Multivariable analysis included those patients with middle, posterior, or middle-posterior disease processes (Table 6). Within these subgroups, COPD (OR, 3.98; 95% confidence interval [CI], 1.13e 13.98; P ¼ 0.03), increased operative time (OR, 4.03; 95% CI, 1.71e 9.47; P ¼ 0.009), and schwannoma (OR, 2.69; 95% CI, 1.30e5.57; P ¼ 0.008) or TN (OR, 5.42; 95% CI, 2e5.14.7; P ¼ 0.008) were associated with significantly higher odds of a postoperative CSF leak.
Table 2. Patient Demographics and Comorbidities No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Variable Age (years), mean (standard deviation)
54.4 (14.3)
52.0 (14.3)
0.14
Female sex
1820 (64.2)
50 (59.5)
0.44
White race
2023 (83.1)
61 (80.3)
0.61
Hispanic ethnicity
242 (9.63)
3 (4.00)
0.15
Body mass index (kg/m ), mean (standard deviation)
30.1 (6.96)
31.1 (6.86)
0.23
Body mass index >30 kg/m2 (obese)
1243 (44.2)
41 (48.8)
0.47
2
>0.99
Functional status Independent
2771 (98.1)
82 (98.8)
Partial dependent
51 (1.80)
1 (1.20)
Totally dependent
4 (0.14)
0 (0.00)
American Society of Anesthesiologists classification
0.92
Low
1382 (48.8)
40 (47.6)
High
1452 (51.2)
44 (52.4)
Smoker
367 (12.9)
14 (16.7)
0.41
326 (11.5)
10 (11.9)
>0.99
Comorbidities History of myocardial infarction
4 (0.14)
0 (0.00)
>0.99
1088 (38.4)
33 (39.3)
>0.99
Previous percutaneous coronary intervention
5 (2.16)
0 (0.00)
>0.99
Peripheral vascular disease
2 (0.86)
0 (0.00)
>0.99
Chronic obstructive pulmonary disease
40 (1.41)
5 (5.95)
0.009
Ventilator dependent
2 (0.07)
0 (0.00)
>0.99
Impaired sensorium
1 (0.43)
0 (0.00)
>0.99
Congestive heart failure Hypertension
Stroke
2 (0.86)
0 (0.00)
>0.99
On dialysis
1 (0.04)
0 (0.00)
>0.99
Bleeding diathesis
26 (0.92)
1 (1.19)
0.55
Need of chronic transfusion
1 (0.04)
0 (0.00)
>0.99
Ascites
0 (0.00)
0 (0.00)
N/A
Esophageal varices
0 (0.00)
0 (0.00)
N/A
DISCUSSION
Diabetes mellitus
326 (11.5)
10 (11.9)
>0.99
CSF leak is a known complication in skull base surgery, inciting a range of significant and potentially morbid sequelae. CSF leak has also proved an elusive end point to characterize in a precise, accurate, and generalizable fashion. Whereas single-institution or
On chronic steroids
199 (7.02)
7 (8.33)
0.81
e564
www.SCIENCEDIRECT.com
Values are number (%) except where indicated otherwise. Bold values denote statistical significance.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
Table 3. Summary of Operative Variables
Variable
Table 4. Postoperative Outcomes
No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Diagnoses, row Meningioma
0.012 970 (97)
30 (3)
Pituitary adenoma
642 (98.8)
8 (1.2)
Schwannoma
821 (95.9)
35 (4.1)
Trigeminal neuralgia
401 (97.3)
11 (2.7) <0.001
Procedures, row Anterior
119 (91.5)
11 (8.5)
Middle
614 (98.7)
8 (1.3)
Posterior
1421 (97.5)
36 (2.5)
Middle-posterior
592 (96.9)
19 (3.1)
Anterior-middle-posterior
88 (89.8)
10 (10.2)
Operative time (minutes), mean (standard deviation)
283 (172)
385 (180)
<0.001
Lumbar drain placement
88 (3.11)
7 (8.33)
0.02
Skull base reconstruction
421 (14.9)
40 (47.6)
<0.001
Values are number (%) except where indicated otherwise. Bold values denote statistical significance.
single-surgeon series reflect selected outcomes and a small number of patients, preceding big data investigations may have been limited in the granularity of their inclusion criteria by the nature of the database, particularly in the context of highly complicated and individualized skull base operations. We initially sought to design a novel framework for mapping CPT codes to skull base anatomic locations, through which we studied 4 prevalent diseases at significant risk of postoperative CSF leak in almost 3000 patients using NSQIP. However, the lack of anatomic granularity in CPT coding for skull base tumors made using a CPT codeebased data set difficult. Although some useful indirect information may be derived from our analysis and similar frameworks in the future, our overarching conclusion is that big data may not offer more than a little insight for the skull base surgeon. Regarding the study results, we noted that operations involving the anterior fossa (in particular, anterior fossa meningiomas) were at highest risk for CSF leak overall, alongside any operations that included a code for the OZ approach. Within the subset of operations limited to the middle and posterior fossas, we identified COPD, increased operative time, and the diagnoses of schwannoma or TN as significant risk factors for CSF leak. This situation fits with typical dogma, although the 2.9% incidence of CSF is lower than expected.1-10 There are several potential explanations for this discrepancy. Chiefly, NSQIP does not capture readmissions/reoperations that occur beyond 30 days after the index operation. In addition, NSQIP coding procedures have been shown to be subject to inaccuracies that may affect the reliability of the database, particularly for analysis of skull base tumor
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
Variable
No Cerebrospinal Cerebrospinal Fluid Leak Fluid Leak (n [ 2834) (n [ 84) P Value
Length of stay (days), mean (standard deviation)
5.11 (6.31)
6.40 (6.64)
0.08
Major complication
269 (9.49)
16 (19.0)
0.007
Sepsis/septic shock
39 (1.37)
6 (7.14)
<0.001
Stroke
39 (1.38)
1 (1.19)
>0.99
Organ-space SSI
19 (0.67)
4 (4.76)
0.004
Deep incisional SSI
14 (0.49)
3 (3.57)
0.012
Cardiac arrest
9 (0.32)
0 (0.00)
>0.99
On ventilation for >48 hours
61 (2.15)
2 (2.38)
0.25
Reintubation
57 (2.01)
3 (3.57)
0.25
Pneumonia
44 (1.55)
3 (3.57)
0.15
Pulmonary embolism
38 (1.34)
2 (2.38)
0.32
Bleeding requiring transfusion
123 (4.34)
5 (5.95)
0.42
Renal insufficiency
3 (0.11)
0 (0.00)
>0.99
Myocardial infarction
2 (0.07)
0 (0.00)
>0.99
Any minor complication
98 (3.46)
8 (9.52)
0.01
Urinary tract infection
46 (1.62)
0 (0.00)
0.642
Wound disruption
6 (0.21)
7 (8.33)
<0.001
Deep venous thrombosis
39 (1.38)
0 (0.00)
>0.99
Superficial SSI
11 (0.39)
2 (2.38)
0.052
Values are number (%) except where indicated otherwise. Bold values denote statistical significance. SSI, surgical site infection.
cases.18-20 Spurious coding and billing documentation practices embedded within the medical record by clinicians or support staff may have introduced 2 additional sources of bias: application of a skull base approach code to a simpler, lower-risk operation for fiduciary gain or inaccurate coding of unplanned take-back operations, to falsely improve quality metrics. Big data are not able to give us granular results in this instance. However, they do have the ability to zoom out and provide an abstract perspective on trends that, although perhaps inaccurate for any given patient, would not have become apparent without the aid of a large patient sample. An interesting and novel finding of the present analysis is that COPD may be an independent risk factor for CSF leak: a pairing of morbidity and complication sufficiently rare to escape notice in even large single-institution series, but one that may provide some insight into the pathophysiology of CSF leak. We propose that, within this subset of patients, coughing is an important and underappreciated risk factor for the development of CSF leak. In the COPD population at large, chronic cough has a prevalence of at least 55%, a percentage that would be expected to
www.journals.elsevier.com/world-neurosurgery
e565
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
Table 5. Incidence of Cerebrospinal Fluid Leak by Procedure Group
Diagnosis
No Cerebrospinal Fluid Leak, n (%)
Cerebrospinal Fluid Leak, n (%)
Anterior
n ¼ 119
n ¼ 11
Meningioma
103 (86.6)
11 (100)
Pituitary
15 (12.6)
0 (0.00)
Schwannoma
1 (0.84)
0 (0.00)
n ¼ 614
Middle Meningioma Pituitary
P Value
0.42
Table 6. Multivariable Logistic Regression Fitted for Occurrence of Cerebrospinal Fluid Leak after Surgery in the Middle and Posterior Cranial Fossae Odds Ratio
95% Confidence Interval
P Value
Body mass index
1.24
0.82e1.86
0.55
Presence of chronic obstructive pulmonary disease
3.98
1.13e13.98
0.03
Operative time
4.03
1.71e9.47
0.009
Variable
Diagnosis
n¼8 0.17
Meningioma
7 (87.5)
Schwannoma
2.69
1.30e5.57
0.008
5.42
2.0e14.7
0.009
2.19
0.74e6.50
0.16
3 (0.49)
0 (0.00)
Trigeminal neuralgia
9 (1.47)
1 (12.5)
Pituitary tumor
n ¼ 1421 714 (50.2)
n ¼ 36 7 (19.4)
Pituitary
10 (0.70)
0 (0.00)
Schwannoma
684 (48.1)
29 (80.6)
Trigeminal neuralgia Middle-posterior Meningioma Pituitary Schwannoma Trigeminal neuralgia Anterior-middle-posterior Meningioma
13 (0.91) n ¼ 592 78 (13.2)
3 (15.8)
8 (1.35)
0 (0.00) 6 (31.6)
379 (64.0)
74 (84.1)
0.002
0 (0.00)
127 (21.5) n ¼ 88
Bold values denote statistical significance.
n ¼ 19 0.53
10 (52.6) n ¼ 10 9 (90.0)
Pituitary
8 (9.09)
1 (10.0)
Schwannoma
6 (6.82)
0 (0.00)
>0.99
Bold values denote statistical significance.
increase dramatically in a hospitalized or a postanesthesia cohort.21,22 Mechanistically, coughing is a well-described cause of transient intracranial pressure (ICP) increases, and one can easily imagine how relentless ICP spikes would strain the dural repair, potentially precipitating a CSF leak.23 Secondary mechanisms in addition to cough may also contribute to COPD-associated CSF leak. Previous investigators have investigated the relationship between chronic lung disease and ICP, postulating that venous obstruction and secondary polycythemia caused by chronic hypoxia increase arterial and venous pressures, and consequently the ICP.24 Alternatively, COPD may be a marker for another, more insidious process driving CSF leak in these patients. One such example is obstructive sleep apnea, a common comorbidity in COPD, independent of obesity status.25 These patients show an additive chronic hypoxia and are therefore expected to have amplified ICP derangements.
e566
Reference
601 (97.9)
Schwannoma
Meningioma
Reference
0 (0.00)
Trigeminal neuralgia
Posterior
Reference
1 (0.16)
www.SCIENCEDIRECT.com
By contrast, where do big data perform poorly? They lack the ability to zoom in, whether on a particular observation or on a question that emerges from the analysis. Within our analysis, the nonsignificant difference in BMI is a curious wrinkle. In a preceding NSQIP study, Murphy et al.12 identified BMI as a predictor of 30-day postoperative complications after surgery for “benign cranial nerve neoplasms,” including CSF leak. Another recent study by Alattar et al.6 investigated almost 7000 cases of vestibular schwannoma surgery in the California Office of Statewide Health Planning and Development database and similarly identified BMI as a significant risk factor for CSF leak. Further still, in a large, single-institution series of 457 vestibular schwannoma cases operated on by the senior author (M.J.L.), Copeland et al.3 completed a risk-adjusted analysis that identified BMI, operative time, and translabyrinthine approach as independent risk factors for CSF leak. These results make a compelling case for the role of BMI as a key CSF leak risk factor, which accords with previously described BMI-induced pathophysiologic ICP changes, as well as anecdotal experience.26 That we did not reproduce this finding is likely attributable to above-average BMI in both groups (32 among CSF leak cases; 30 among controls), which is potentially an unintended consequence of our stringent inclusive/exclusion criteria. Without access to the medical records of these patients, that hypothesis will remain untested, and the question unanswered. Another example of the inability to drill down is our finding that operative time was significantly longer in the CSF leak group, by a mean 102 minutes. Numerous common and uncommon contributors to longer operative times can be imagined as motivating this finding, many of which would constitute major confounders (e.g., a larger tumor, a tumor that required more extensive bony removal to expose, an intraoperative complication or complex reconstruction that was not reflected in a specific billing code, and so on). This difficulty in measuring important confounders is reflected in the paradoxically higher rates of lumbar drain placement and use of skull base reconstruction methods in the CSF leak group. Both of these techniques have been reported to be
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
associated with decreased rates of CSF leak based on anecdotal evidence and previous reports.27-29
specialists, who undergo training and periodic auditing, but their submissions are dependent on the source data coded by the operating physician or their administrative team, allowing for several additional sources of bias, as mentioned earlier.
Strengths and Limitations The core strengths of our study lie in its novel methodology. We developed a new and more comprehensive approach to classifying patients derived from large national data sets by categorizing CPT codes based on a strictly anatomic scheme. This approach was further strengthened by our meticulous and redundant review process built around consensus decision making, which is a level of internal validity not typically seen in preceding studies derived from administrative databases.30 Notwithstanding, our study is subject to several major limitations. First, as we described in detail earlier, NSQIP and other national surgical registries are not designed to assess skull base patients or diseases. For instance, variables that are highly relevant to an analysis of CSF leak are absent from NSQIP (histopathology, tumor size and characteristics, extent of resection, and disease laterality, to name only a few of the most notable absences). Second, the originating centers and surgeons are deliberately obscured, and so there is no way to measure or account for differences in CSF leak rates between providers, which may not follow a normal distribution. A recent study by Babadjouni et al. showed that increased procedural volume is associated with lower 30-day and 90-day readmission rate after microsurgical resection of vestibular schwannoma.31 This important observation might also reflect not only the ability of centers with greater cumulative experience to manage postoperative complications of skull base surgery in the outpatient rather than inpatient setting but also the critical role of the learning curve in the repair process.31 Third, NSQIP captures data only within a 30-day window, excluding late leaks, which are a rare but appreciable subgroup.32 Fourth, lumbar drain insertion rates were underestimated, because bedside procedures are not reported in the registry. NSQIP variables are captured and entered by nurse
REFERENCES 1. Rodgers GK, Luxford WM. Factors affecting the development of cerebrospinal fluid leak and meningitis after translabyrinthine acoustic tumor surgery. Laryngoscope. 1993;103:959-962. 2. Bryce GE, Nedzelski JM, Rowed DW, Rappaport JM. Cerebrospinal fluid leaks and meningitis in acoustic neuroma surgery. Otolaryngol Head Neck Surg. 1991;104:81-87. 3. Copeland WR, Mallory GW, Neff BA, Driscoll CLW, Link MJ. Are there modifiable risk factors to prevent a cerebrospinal fluid leak following vestibular schwannoma surgery? J Neurosurg. 2015;122:312-316. 4. Bani A, Gilsbach JM. Incidence of cerebrospinal fluid leak after microsurgical removal of vestibular schwannomas. Acta Neurochir (Wien). 2002;144: 979-982 [discussion: 982]. 5. Selesnick SH, Liu JC, Jen A, Newman J. The incidence of cerebrospinal fluid leak after
CONCLUSIONS Within the limited scope of our model, we identified anterior fossa operations and complex approaches as significantly associated with higher CSF leak rates. In addition, among operations in the middle or posterior fossas, patients with schwannoma were at markedly higher risk than were patients with TN, who were in turn significantly higher risk than patients with meningioma or pituitary adenoma. With respect to risk factors, COPD and increased operative time independently had a significant association with CSF leak and show the inherent strengths and weaknesses of big data analyses. The COPD association is a particularly interesting and potentially meaningful discovery, because it seems to be underrecognized clinically and may not have come to light without big data power. By contrast, the observation of an association between increased operative time and risk of CSF leak is almost certainly confounded by several variables and difficult to interpret without more closely scrutinizing the cases. At its core, our analysis was designed as an attempt to transform data from a pragmatic coding system that is built to describe the labor performed in a way that it accurately describes the tumors on which that work was performed. Although we have produced a more sophisticated extraction and parsing system that certainly constitutes a marked improvement on preceding big data practices in the skull base literature, we are still essentially left with an unsatisfying solution. Our assessment is that the degree of granularity needed to fully answer clinically meaningful questions about complex cranial operations is simply not accessible via the big data model.
vestibular schwannoma surgery. Otol Neurotol. 2004;25:387-393. 6. Alattar AA, Hirshman BR, McCutcheon BA, et al. Risk factors for readmission with cerebrospinal fluid leakage within 30 days of vestibular schwannoma surgery. Neurosurgery. 2018;82: 630-637. 7. Shiley SG, Limonadi F, Delashaw JB, et al. Incidence, etiology, and management of cerebrospinal fluid leaks following trans-sphenoidal surgery. Laryngoscope. 2003;113:1283-1288. 8. Seiler RW, Mariani L. Sellar reconstruction with resorbable vicryl patches, gelatin foam, and fibrin glue in transsphenoidal surgery: a 10-year experience with 376 patients. J Neurosurg. 2000;93: 762-765. 9. Ciric I, Ragin A, Baumgartner C, Pierce D. Complications of transsphenoidal surgery: results of a national survey, review of the literature, and personal experience. Neurosurgery. 1997;40:225-236 [discussion: 236-237].
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
10. Black PM, Zervas NT, Candia GL. Incidence and management of complications of transsphenoidal operation for pituitary adenomas. Neurosurgery. 1987;20:920-924. 11. McCutcheon BA, Orosco RK, Chang DC, et al. Outcomes of isolated basilar skull fracture: readmission, meningitis, and cerebrospinal fluid leak. Otolaryngol Head Neck Surg. 2013;149:931-939. 12. Murphy ME, McCutcheon BA, Kerezoudis P, et al. Morbid obesity increases risk of morbidity and reoperation in resection of benign cranial nerve neoplasms. Clin Neurol Neurosurg. 2016;148:105-109. 13. Samii M, Matthies C. Management of 1000 vestibular schwannomas (acoustic neuromas): hearing function in 1000 tumor resections. Neurosurgery. 1997;40:248-260 [discussion: 260-262]. 14. Brackmann DE, Owens RM, Friedman RA, et al. Prognostic factors for hearing preservation in vestibular schwannoma surgery. Am J Otol. 2000; 21:417-424.
www.journals.elsevier.com/world-neurosurgery
e567
ORIGINAL ARTICLE AVITAL PERRY ET AL.
15. Becker SS, Jackler RK, Pitts LH. Cerebrospinal fluid leak after acoustic neuroma surgery: a comparison of the translabyrinthine, middle fossa, and retrosigmoid approaches. Otol Neurotol. 2003;24: 107-112. 16. Murphy M, Gilder H, McCutcheon BA, et al. Increased operative time for benign cranial nerve tumor resection correlates with increased morbidity postoperatively. J Neurol Surg B Skull Base. 2016;77:350-357. 17. Harrell FE Jr. rms: Regression Modeling Strategies. R package version 4.0-0. Cityscape; 2013. Available at: https://CRAN.R-project.org/package=rms. Accessed January 4, 2018. 18. Rolston JD, Han SJ, Chang EF. Systemic inaccuracies in the National Surgical Quality Improvement Program database: implications for accuracy and validity for neurosurgery outcomes research. J Clin Neurosci. 2017;37:44-47. 19. Bi WL, Mooney MA, Yoon S, et al. Variation in Coding Practices for Vestibular Schwannoma Surgery. J Neurol Surg B Skull Base. 2019;80:96-102. 20. Oravec CS, Motiwala M, Reed K, et al. In Reply: Big Data Research in Neurosurgery: A Critical Look at this Popular New Study Design. Neurosurgery. 2018;82:E188-E189. 21. Calverley PM. Cough in chronic obstructive pulmonary disease: is it important and what are the effects of treatment? Cough. 2013;9:17. 22. Kessler R, Partridge MR, Miravitlles M, et al. Symptom variability in patients with severe COPD:
e568
www.SCIENCEDIRECT.com
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
a pan-European cross-sectional study. Eur Respir J. 2011;37:264-272. 23. Bloomfield GL, Ridings PC, Blocher CR, Marmarou A, Sugerman HJ. A proposed relationship between increased intra-abdominal, intrathoracic, and intracranial pressure. Crit Care Med. 1997;25:496-503. 24. Carter CC, Fuller TJ. Increased intracranial pressure in chronic lung disease. Neurology. 1957;7: 169-174. 25. Owens RL, Malhotra A. Sleep-disordered breathing and COPD: the overlap syndrome. Respir Care. 2010;55:1333-1344 [discussion: 1344-1346]. 26. Radhakrishnan K, Ahlskog JE, Cross SA, Kurland LT, O’Fallon WM. Idiopathic intracranial hypertension (pseudotumor cerebri). Descriptive epidemiology in Rochester, Minn, 1976 to 1990. Arch Neurol. 1993;50:78-80. 27. Mehta GU, Oldfield EH. Prevention of intraoperative cerebrospinal fluid leaks by lumbar cerebrospinal fluid drainage during surgery for pituitary macroadenomas. J Neurosurg. 2012;116: 1299-1303. 28. Zanation AM, Carrau RL, Snyderman CH, et al. Nasoseptal flap reconstruction of high flow intraoperative cerebral spinal fluid leaks during endoscopic skull base surgery. Am J Rhinol Allergy. 2009;23:518-521. 29. Snyderman CH, Janecka IP, Sekhar LN, Sen CN, Eibling DE. Anterior cranial base reconstruction: role of galeal and pericranial flaps. Laryngoscope. 1990;100:607-614.
30. Faciszewski T, Jensen R, Berg RL. Procedural coding of spinal surgeries (CPT-4 versus ICD-9CM) and decisions regarding standards: a multicenter study. Spine. 2003;28:502-507. 31. Babadjouni R, Wen T, Donoho DA, et al. Increased hospital surgical volume reduces rate of 30- and 90-day readmission after acoustic neuroma surgery. Neurosurgery. 2019;84:726-732. 32. Perry A, Graffeo CS, Copeland WR 3rd, et al. Delayed cerebrospinal fluid rhinorrhea after gamma knife radiosurgery with or without preceding transsphenoidal resection for pituitary pathology. World Neurosurg. 2017;100: 201-207.
Conflict of interest statement: The American College of Surgeons National Surgical Quality Improvement Program and the hospitals participating it are the source of the data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors. Received 9 December 2018; accepted 20 March 2019 Citation: World Neurosurg. (2019) 127:e561-e569. https://doi.org/10.1016/j.wneu.2019.03.207 Journal homepage: www.journals.elsevier.com/worldneurosurgery Available online: www.sciencedirect.com 1878-8750/$ - see front matter ª 2019 Elsevier Inc. All rights reserved.
WORLD NEUROSURGERY, https://doi.org/10.1016/j.wneu.2019.03.207
ORIGINAL ARTICLE AVITAL PERRY ET AL.
BIG DATA ANALYSIS OF CSF LEAK AFTER SKULL BASE SURGERY
SUPPLEMENTARY DATA
Supplementary Table 1. Skull Base Reconstruction Techniques and Corresponding Current Procedural Terminology Codes Code
Description
20926
Tissue grafts, other (e.g., paratenon, fat, dermis)
61618
Secondary repair of dura for cerebrospinal fluid leak, anterior, middle, or posterior cranial fossa after surgery of the skull base; by free tissue graft (e.g., pericranium, fascia, tensor fascia lata, adipose tissue, homologous, or synthetic grafts)
61619
Secondary repair of dura for cerebrospinal fluid leak, anterior, middle, or posterior cranial fossa after surgery of the skull base; by local or regionalized vascularized pedicle flap or myocutaneous flap (including galea, temporalis, frontalis, or occipitalis muscle)
Supplementary Table 2. Pertinent Diagnostic Codes. List of International Classification of Diseases, Ninth Revision (ICD-9) and International Classification of Diseases, Tenth Revision (ICD-10) Diagnosis Codes Used to Identify the Diagnoses of Interest Diagnosis
ICD-9 Code
ICD-10 Code
Meningioma
225.2, 237.6, 192.1
C70.0, C70.9, D32.0, D32.9, D42.0
Pituitary tumor
227.3, 237.0, 253.9, 237, 255
D35.2, D35.3, D44.3, D44.4
Schwannoma
225.1, 237.72, 237.73, 192.0
Trigeminal neuralgia
350.1
C72.50, D33.3 G50.0
WORLD NEUROSURGERY 127: e561-e569, JULY 2019
www.journals.elsevier.com/world-neurosurgery
e569