Keywords
postoperative complications - neurosurgery - neurological deficit
Introduction
Neurosurgery is associated with high-rates of mortality and morbidity, due to the
complexity of brain structures. The complications can range from mild postoperative
nausea and vomiting to devastating neurological deterioration. Many times the complications
are analyzed in a specific group of patients. A large number of studies have been
dedicated to reporting specific complications in specific subgroups of surgeries.[1]
[2]
[3]
[4]
[5]
[6]
[7] However, it is important to have an overview of all the complications occurring
in a neurosurgical setup, which will help in planning and execution of measures required
for effective management of neurosurgical patients. Such data are currently lacking
in India. This multicenter collaborative effort was envisioned to collect observational
data regarding postoperative complications within 24 hours in cranial surgeries among
the Indian population.
Aim
To describe the postoperative neurological complications occurring within the first
24 hours after cranial surgery and to identify the predictive factors for those complications.
Methods
Three institutes participated in the study. They are National Institute of Mental
Health and Neuro Sciences, Bangalore; All India Institute of Medical Sciences, New
Delhi; and Postgraduate Institute of Medical Education and Research, Chandigarh. Data
was collected after obtaining the ethics committee approval from the respective institutes.
All the three institutes are tertiary care academic institutes. All three institutes
have DM Neuroanesthesia training program and all possess advanced multimodality neuromonitoring
facilities. The study was of a prospective, observational, multicentre design, with
data collected over a period of two months or 100 cases, whichever is earlier, from
each participating institute. The data was collected from all the institutes in the
same year, was collected in a paper format and entered into a predesigned Microsoft
excel sheet. The excel sheet was designed by the investigators and distributed among
all the three centers to maintain uniformity in the data capture. The relevant explanations
for the parameters were given in the excel sheet itself. Data confidentiality was
assured by excluding any patient identifiers from the worksheet.
All the patients aged 18 years and above of both sexes undergoing elective or emergency
craniotomies during the designated month were included in the study. The data included
multiple variables collected preoperatively, intraoperatively, and at one hour and
24 hours, postoperatively. In this study, we intended to describe the postoperative
neurological complications that occurred within the first 24 hours after surgery and
finding out the predictive factors for those complications. The postoperative neurological
complications assessed were: 1. Neurological deficit (ND) defined as new focal neurological
motor deficit in the immediate postoperative period (24 hours) relative to preoperative
status. 2. Deterioration of sensorium (SD) defined as reduction in Glasgow coma score
(GCS) by 2 or more points (within 24 hours postoperatively) compared with preoperative
GCS. 3. Postoperative seizures (SZs) defined as any seizure activity detected within
24 hours in the postoperative period.
Statistical Analysis
Data was compiled in a Microsoft Excel worksheet and analysis was conducted on R software
version 3.5.2.[8] In this study, the analysis was restricted to only neurological complications, that
is, postoperative new NDs, postoperative SD (fall in GCS more than 2 points compared
with baseline), and postoperative seizures, in the first 24 hours after craniotomy.
All possible variables (both preoperative and intraoperative) associated with the
above neurological complications were analyzed. They were initially tested using Chi-square/Fisher
exact test or Mann–Whitney U test. The predictors which were found statistically significant
at p < 0.2 for the respective complications were entered into a multiple logistic regression
model (using glm function of R). Due to high-multicollinearity between multiple variables on account
of low-event rates of some factors, the model estimates and confidence intervals were
found to be unstable (large standard errors and coefficients). Hence, penalized logistic
regression using Firth method was used to run the same models (using package logistf for R).[9] This method reduces variability of estimates. The regression estimates were more
stable. To improve generalizability of the results, bootstrapping of the dataset was
done to obtain 1000 datasets, and variability of coefficients (95% confidence intervals)
was inferred, while correcting for bias induced with repeated resampling (using package
boot for R).[10] Results of both modelling procedures (Firth and bootstrapped Firth) are presented
in the form of odds ratios. The nominal data are presented as percentages, and interval/ordinal
scale data presented as median and interquartile range. Alpha error of 5% was taken
as significant.
Results
Data was collected from 279 cases (institution 1 = 110, institution 2 = 100, institution
3 = 69). The total number of neurological complications was 53 (19%). There were 28
patients with new postoperative ND (10.04%), 24 patients had SD (8.6%), and 17 patients
had seizures (6.1%). There were few patients who had two complications together. However,
no patient had all the three complications. The demographic details are given in [Table 1]. The univariate tests of association of putative predictors with the outcome variables
are provided in [Supplementary Table S1] (ND), [S2] (SD) and [S3] (SZ) (available in the online version).
Table 1
Demographic variables
|
Variable
|
Descriptive
|
|
Abbreviations: GCS, Glasgow coma scale’ IQR, interquartile range; TBI, traumatic brain
injury.
Note: Age and weight as means ± standard deviation (SD), GCS as median (IQR) and sex,
diagnosis, procedure type, comorbidities, previous surgery as percentages.
|
|
Age (yrs)
|
39 ± 17
|
|
Weight (kg)
|
57 ± 15
|
|
Sex (F/M) (%)
|
37.6/62.4
|
|
Diagnosis (infectious/infratentorial/tumor/supratentorial/tumor/TBI/vascular) (%)
|
1.1/15.8/55.9/10.4/16.8
|
|
Procedure type (elective/emergency) (%)
|
64.9/35.1
|
|
Comorbidities (cardiovascular/endocrine/neurological/none/respiratory) (%)
|
15.4/6.5/9.3/63.4/5.4
|
|
Previous Surgery (no/yes) (%)
|
87.5/12.5
|
|
GCS (IQR)
|
15 (15–15)
|
New Neurological Deficits
The factors which were found significant (at p < 0.2) on univariate analysis were included in the multivariate model. These included
diagnosis, preoperative comorbidities, intraoperative opioid used, intraoperative
bradycardia, hypoxia, hypercapnia, potassium level change, coagulopathy, institution,
and duration of anesthesia. Of these, intraoperative potassium level change and coagulopathy
were excluded due to low-event rate for the dependent variable. The model was found
to be significantly better than a model of no effect (p < 0.001).
The model coefficients are shown in [Table 2]. Factors found significant after bootstrapping were institution and diagnosis. NDs
were significantly less in institution 2 compared with institution 1. Diagnosis of
traumatic brain injury (TBI) (compared with supratentorial tumor diagnosis) was associated
with very low risk of NDs postoperatively. Vascular diagnosis was associated with
a higher chance of ND.
Table 2
Coefficients for variables entered into the multiple regression model for postoperative
new NDs
|
Factors
|
Firth OR (95% CL)
|
p-Value
|
Boot OR (95% CL)
|
p-Value
|
|
Abbreviations: IO, intraoperative; NDs, neurological deficits; NS, not significant.
Note: Reference levels: for institution–institution 1, for diagnosis–supratentorial
tumor, for comorbidity–no comorbidity, for opioid–fentanyl. p < 0.05 is statistical level of significance.
|
|
Intercept
|
0.03 (0–0.22)
|
< 0.05
|
0.03 (0–0.83)
|
< 0.05
|
|
Institution 2
|
0.15 (0.04–0.47)
|
< 0.05
|
0.15 (0.06–0.71)
|
< 0.05
|
|
Institution 3
|
0.12 (0–1.77)
|
NS
|
0.12 (0.01–2.45)
|
NS
|
|
Diagnosis infratentorial tumor
|
0.42 (0.07–2.06)
|
NS
|
0.42 (0.03–3.89)
|
NS
|
|
Diagnosis TBI
|
0.08 (0–0.8)
|
< 0.05
|
0.08 (0.01–0.31)
|
< 0.05
|
|
Diagnosis vascular
|
8.35 (2.26–35.84)
|
< 0.05
|
8.36 (1.19–25)
|
< 0.05
|
|
Comorbidity cardiovascular
|
1.83 (0.5–6.17)
|
NS
|
1.83 (0.25–8.84)
|
NS
|
|
Comorbidity endocrine
|
0.13 (0–1.63)
|
NS
|
0.13 (0.01–1.89)
|
NS
|
|
Comorbidity neurological
|
4.25 (0.92–17.3)
|
NS
|
4.25 (0.49–17.73)
|
NS
|
|
Comorbidity respiratory
|
2.62 (0.41–13.86)
|
NS
|
2.62 (0.12–16.83)
|
NS
|
|
Opioid (morphine)
|
0.41 (0.02–10.15)
|
NS
|
0.41 (0.06–6.26)
|
NS
|
|
IO bradycardia present
|
5.85 (1.3–26.31)
|
< 0.05
|
5.85 (0.31–49.5)
|
NS
|
|
IO hypoxia present
|
4.44 (0.58–26.85)
|
NS
|
4.44 (0.04–29.7)
|
NS
|
|
IO hypercarbia present
|
5.62 (0.89–45.65)
|
NS
|
5.62 (0.23–35.48)
|
NS
|
|
Anesthesia duration
|
1.01 (1–1.01)
|
NS
|
1.01 (1–1.01)
|
NS
|
Postoperative Deterioration of Sensorium
The factors which were found significant on univariate analysis (at p < 0.2) were included into the multivariate model. They are diagnosis, emergency or
elective nature of surgery, maintenance anesthetic used, inhalational agent used,
antiepileptic use, intraoperative brain swelling, intraoperative hypertension, sodium
level change, coagulopathy, hypothermia, and duration of anesthesia. Sodium level
change and coagulopathy were excluded from final analysis due to low-event rate for
the dependent variable. The model was found to be significantly better than a model
of no effect (p < 0.001).
The model coefficients are shown in [Table 3]. The duration of anesthesia was found to be significantly predictive of SD (OR/CI
= 1.01 / 1–1.02).
Table 3
Coefficients for variables entered into the multiple regression model for postoperative
SD
|
Factors
|
Firth OR (95% CL)
|
p-Value
|
Boot OR (95% CL)
|
p-Value
|
|
Abbreviations: IO, Intraoperative; NS, Not significant; SD, sensorium deterioration;
TBI, traumatic brain injury.
Note: Reference levels: for diagnosis–supratentorial tumor, for procedure status–elective,
for inhalational agent–isoflurane. p < 0.05 is statistical level of significance.
|
|
Intercept
|
0 (0–0.01)
|
< 0.05
|
0 (0–0.03)
|
< 0.05
|
|
Diagnosis infratentorial tumor
|
0.58 (0.104–2.56)
|
NS
|
0.58 (0.13–2.83)
|
NS
|
|
TBI
|
1.22 (0.08–13.03)
|
NS
|
1.22 (0.17–20.09)
|
NS
|
|
Vascular
|
2.28 (0.61–8.21)
|
NS
|
2.27 (0.21–10.7)
|
NS
|
|
Procedure status - emergency
|
2.52 (0.68–9.61)
|
NS
|
2.51 (0.31–14.59)
|
NS
|
|
N2O used
|
1.58 (0.31–12.22)
|
NS
|
1.57 (0.14–8)
|
NS
|
|
Inhalational agent–desflurane
|
1.56 (0.22–8.28)
|
NS
|
1.55 (0.11–15.33)
|
NS
|
|
Inhalational agent–sevoflurane
|
2.13 (0.66–7.15)
|
NS
|
2.14 (0.55–9.87)
|
NS
|
|
IO antiepileptic used
|
0.89 (0.22–3.22)
|
NS
|
0.89 (0.12–6.96)
|
NS
|
|
Brain swelling present
|
5.48 (1.67–17.95)
|
< 0.05
|
5.47 (0.86–21.76)
|
NS
|
|
Hypertension present
|
2.62 (0.66–9.16)
|
NS
|
2.61 (0.21–8.5)
|
NS
|
|
Hypothermia present
|
1.99 (0.31–10.52)
|
NS
|
1.99 (0.08–20.09)
|
NS
|
|
Duration of anesthesia
|
1.01 (1.01–1.02)
|
< 0.05
|
1.01 (1–1.02)
|
< 0.05
|
Postoperative Seizures
The factors which were found significant on univariate analysis (at p < 0.2) were included into the multivariate model. They are emergency or elective
nature of surgery, preoperative neurological deficits, nitrous oxide use, antiepileptic
use, steroid use, opioid used, intraoperative hypertension, arrhythmia, hypothermia,
institute factor, anesthesia duration, and total GCS score. The final model was significantly
better than a null model (p = 0.004). [Table 4] shows the coefficients for variables in the final model.
Table 4
Coefficients for variables entered into the multiple regression model for postoperative
seizures
|
Factors
|
Firth OR (95% CL)
|
p-Value
|
Boot OR (95% CL)
|
p-Value
|
|
Abbreviations: GCS, Glasgow coma scale; IO, intraoperative; NS, not significant; SZs,
seizures.
Note: Reference levels: for institution–institution 1; for procedure status–elective;
for opioid–fentanyl. p < 0.05 is statistical level of significance.
|
|
Intercept
|
0.09 (0–1.83)
|
NS
|
0.09 (0–2.28)
|
NS
|
|
Institution 2
|
2.42 (0.46–17.55)
|
NS
|
2.42 (0.31–20.57)
|
NS
|
|
Institution 3
|
0 (0–0.08)
|
< 0.05
|
0 (0–21.16)
|
NS
|
|
Procedure status–emergency
|
1.35 (0.3–5.25)
|
NS
|
1.35 (0.31–6.16)
|
NS
|
|
Preoperative Neurological deficit present
|
2.23 (0.7–7.3)
|
NS
|
2.23 (0.47–8.86)
|
NS
|
|
Opioid–morphine
|
156.17 (5.3–2.9 *104)
|
< 0.05
|
156.18 (0.02–1.26*107)
|
NS
|
|
N2O used
|
0.38 (0.05–2.45)
|
NS
|
0.38 (0.03–8.42)
|
NS
|
|
Antiepileptic used
|
2.41 (0.51–9.83)
|
NS
|
2.41 (0.16–12.85)
|
NS
|
|
Steroid used
|
0.22 (0–2.26)
|
NS
|
0.22 (0.06–1.29)
|
NS
|
|
Hypertension
|
1.31 (0.21–5.97)
|
NS
|
1.31 (0.26–6.89)
|
NS
|
|
IO arrythmia present
|
2.02 (0.41–8.12)
|
NS
|
2.02 (0.29–8.04)
|
NS
|
|
Hypothermia
|
6.23 (1.02–34)
|
< 0.05
|
6.23 (0.68–39.49)
|
NS
|
|
Preop GCS
|
0.96 (0.77–1.2)
|
NS
|
0.96 (0.77–1.3)
|
NS
|
Discussion
This study assessed the postoperative neurological complications within 24 hours after
a cranial surgery. The incidence of all neurological complications was 19%. The incidence
of postoperative NDs was 10%, SD was 8.6%, and seizures was 6.1%. This study also
tried to assess the independent predictors for the above neurological complications.
Many studies have assessed complications in specific groups of patients. In our study,
we have assessed complications in all types of cranial surgeries within 24 hours.
It is important that all the neurological setups should be prepared to handle the
possible complications. There is a need to assess whether these neurological complications
can be prevented. Toward this end, we tried to assess the independent predictors of
these neurological complications. With an understanding of these predictors, one should
be able to decrease these complications and improve the outcomes of the patients.
Pooling of data from multiple institutions should help in applicability of the results
across multiple institutions.
Postoperative New Neurological Deficits
Postoperative ND is a matter of concern and new or worsened ND following neurosurgery
is known to influence patient outcome following neurosurgery. In a retrospective study,
Rehman et al[11] reported that development of a postoperative ND following glioblastoma resection
significantly affected survival. The authors observed that development of a permanent
ND postoperatively had ominous prognosis with reduced survival time. However, patients
who had temporary deficits had improved survival.
In this study, the initial univariate tests specified the putative variables which
were associated with the incidence of postoperative NDs at a p < 0.2. Diagnosis of the patient was found to be predictive of ND. Patients with vascular
lesions (aneurysms and arteriovenous malformations) were approximately eight times
more susceptible to ND. TBI showed 92% less chance of ND compared with supratentorial
tumors. The increased deficits with vascular lesions are easily explainable, as these
patients are prone to develop vasospasm/delayed cerebral ischemia and thus the ND.
However, in TBI patients, the incidence of postoperative deficits was less. This might
be due to the following reasons: 1.Many patients might be having deficits preoperatively
2. The deficits are truly less frequent postoperatively, as there is an improvement
in the deficits after the operation rather than worsening of the deficits. 3. Another
reason could be that these patients are sedated and ventilated postoperatively; hence,
it was difficult to assess the NDs. It also depends on the availability of the dedicated
neuro-ICU and availability of the ICU beds.
Institution 2 reported 85% less chance of NDs compared with Institution 1, which may
be explained by the differences in hospital policy and expertise of treating surgeons.
Surgeon’s expertise and attitude toward treatment may play an important role. The
hospital policy regarding choice of patients for surgery may also play a role. Some
hospitals are very aggressive in their treatment approach, and they may accept even
poor grade patients for surgery. Such institutions may have higher incidence of postoperative
NDs. In one study, age has been shown to be a risk factor for postoperative ND.[12] However, in our study, age was not seen as a significant factor affecting the NDs.
Postoperative stroke rates vary significantly in various surgical populations. It
could be 0.1 to 10%.[13] The high-rates are seen in cardiac surgery, vascular surgery and neurological surgeries.
In a review of glioma surgery, the incidence of new NDs was 0 to 20%.[14] Studies with newer advancements in surgical techniques have reported lesser incidence
of complications. In another study of meningiomas, an incidence of 14.8% has been
reported. In surgery for vestibular schwannoma, the incidence of new NDs was as high
as 31%.[15] Compared with these studies, the incidence of postoperative NDs in the current study
was relatively low. It may be because our study was limited to only 24 hours.
Postoperative Deterioration of Sensorium
In this study, postoperative SD was observed in 8.6% of patients. Various studies
have quoted different incidences of postoperative SD. One Indian study has quoted
an incidence of 11%.[16] In the final analysis of the results, anesthesia duration and intraoperative brain
swelling were found to be independently predictive of SD. This may be explained by
residual sedation of anesthetics due to storage in fat compartments over prolonged
exposure. Also, it can be related to the experience of neurosurgeon. A less experienced
surgeon takes longer time than the experienced surgeon. Intraoperative brain swelling
also can theoretically cause SD due to decrease in the cortical cerebral blood flow.
However, the effect did not survive bootstrapping.
Postoperative Seizures
In this study, seizures were observed postoperatively in 6% of patients. Various studies
have reported an incidence of postoperative seizures ranging from 1 to 12%.[11]
[13]
[17] One Indian study has documented an incidence of 6.3%, which is almost similar to
our study. Dorzi et al[18]analyzed the risk factors for seizures following resection of primary brain tumors.
The independent risk factors reported in their study were presence of preoperative
seizures and small tumor size. Preoperative seizure history is a well-known risk factor
for postcraniotomy seizures.[19] Another interesting finding in this study was association of small tumor size and
postoperative seizures. The authors explained this association based on the requirement
of more brain tissue dissection or manipulation for surgical access in small lesions.
In the regression analysis, the Firth model has shown institution 3, usage of morphine
and presence of hypothermia as predictive factors. However, the confidence intervals
of morphine use were seen to be impossibly large and none of the factors were significant
with bootstrapping.
Strengths of Study
Multicentre nature of the data provides a pragmatic view of the topic, with variation
in practices and outcomes.
Limitations of Study
In spite of multicentre nature of the study, the event rate for the outcomes was relatively
low. Finding the causative factors for the neurological complications became difficult.
This study was not designed to assess complication differences between the institutions
but to look for differences in the management strategies. Other limitations are as
follow: We have not included and evaluated the facilities available in the institutions,
for example, dedicated neuro-ICU, navigation, awake craniotomy, intraoperative neuromonitoring,
etc., which can have a bearing on the outcomes. Types of lesions operated at various
centers also have not been taken into account.
Conclusion
Overall postoperative neurological complications are 19%. The incidences of postoperative
NDs, SD, and postoperative seizures were 10, 8.6, and 6.1%, respectively. There is
a large variation in the institutional reporting of the complications. Further well-designed
studies with larger sample sizes and better models are required to overcome the limitation
of low-event rate for prediction and prognostication of postoperative neurological
complications following surgery.
