Outcomes in Pediatric Transsphenoidal Pituitary Surgery Stratified by the Use of Image Guidance: An Analysis of the Kids' Inpatient Database from 1997 to 2016

Abstract

Background  To evaluate the utilization of image guidance technology for pediatric transsphenoidal pituitary resection (TSPR) and analyze the complication rates, length of stay (LOS), and total cost for such surgeries as function of time and utilization of image guidance.

Methods  The Healthcare Cost and Utilization Project Kids' Inpatient Database (KID) was queried for all cases of TSPR between 1997 and 2016. Factors extracted included patient demographics, use of image guidance, LOS, total cost, and complications, including panhypopituitarism, diabetes insipidus (DI), cerebrospinal fluid (CSF) rhinorrhea, and cranial nerve (CN) III, IV, and VI palsies. Multivariate logistic regression was utilized to determine the odds ratio of developing panhypopituitarism, DI, CSF rhinorrhea, and CN palsies for image-guided versus nonimage-guided cases. A generalized linear model was used to determine the effect of image guidance on inflation-adjusted total cost and LOS.

Results  A total of 1,297 cases of TSPR were included in the KID over this time period. The majority were female, Caucasian, and older than 15 years. Utilization of image guidance has rapidly increased since 2006. Complication rates were comparable, but when controlling for other factors, the use of image guidance showed a lower risk of postoperative DI ( p  = 0.05). The use of image guidance also resulted in a shorter LOS by 2.84 days ( p  < 0.001) with no associated increase in total cost ( p  = 0.663).

Conclusion  The use of imaging guidance for pediatric TSPR has precipitously increased in recent years, as it is cost-effective, decreases LOS, and may lead to lower complication rates, such as DI.

Introduction

Pituitary surgery demonstrates a notable context for the interplay between surgery and technology. As improvements in technology have been made, advances in pituitary surgery have followed suit. Significant morbidity and mortality of the initial transfrontal approaches led many practitioners to develop the transsphenoidal approach in the early 20th century. ¹ This approach was later abandoned for nearly 35 years, until the introduction of the operating microscope led to its resurgence in 1965. ¹ This technique was unrivaled until the 1990s when the endoscope was introduced. ² Superior illumination, exposure, and visualization revolutionized pituitary surgery and the entire field of skull base surgery, given its greater safety and efficacy. ³ Most recently, stereotactic technology, or computer-guided surgery, has again transformed pituitary surgery as multiple studies have demonstrated a robust safety profile and lower rate of complications. ^{4 5 6 7 8} Image guidance has also proved to be economical. Chung et al looked into the Nationwide Inpatient Sample database and noted that image guidance for adult transsphenoidal pituitary resection (TSPR) was associated with a lower rate of cerebrospinal fluid (CSF) rhinorrhea, shorter length of stay (LOS), and lower cost. ⁹

Due to the relative infrequency of pituitary lesions necessitating surgical intervention in pediatrics, similar studies investigating safety, efficacy, and cost efficiency of TSPR in children, with or without mention of image guidance, are scarce at best. To date, there are a handful of case series and one population-based analysis investigating TSPR in the pediatric population. ^{10 11 12} Hanba et al investigated the Healthcare Cost and Utilization Project (HCUP) Kids' Inpatient Database (KID) and demonstrated a lower rate of complications, shorter hospital stay, and lower cost for transsphenoidal compared with transfrontal pituitary resection. ¹⁰ However, given the paucity of pediatric TSPR literature, there have been no studies investigating the use of image guidance in this population, to the best of our knowledge. Since this technology has become standard of care at many pediatric skull base centers, it is important to understand the ramifications of its use for the practicing clinician.

The objective of this study is to investigate the utilization and outcomes of image guidance in pediatric TSPR. In particular, emphasis was placed on analyzing the complication rates, LOS, and total cost for such surgeries as a function of time and utilization of image guidance. To accrue sufficient data for analysis, the KID was utilized as it represents a validated source for longitudinal and population-based studies.

Methods

The KID is a product of the HCUP which is administered by the Agency of Healthcare Research and Quality. It is the largest national pediatric database that collects inpatient data on health care costs, utilization, outcomes, procedures, and demographics. ¹³ It includes inpatient and discharge data from 5,118 hospitals over 44 of the 50 states in the United States. To evaluate long-term trends regarding TSPR, the latest seven editions of the database were utilized. Each edition is produced approximately every 3 years, and the data for this study were extracted from the 1997, 2000, 2003, 2006, 2009, 2012, and 2016 editions. All editions, with the exception of 2016, utilize the International Classification of Disease, Ninth Revision (ICD-9) codes for diagnoses and procedures. The most recent 2016 data utilized the International Classification of Disease, Tenth Revision (ICD-10) codes for diagnoses and procedures. Scaled data, such as number of patients, were transformed using HCUP-provided discharge weights to ensure these values represented national estimates. Furthermore, the total charge was selected to represent the total cost given that a HCUP-provided cost-to-charge ratio is not available for the KID for all years included in the study. As this study utilized a publicly available database which holds records in a deidentified nature, it was exempted from Institutional Review Board approval.

Our sample population from KID was obtained by searching for the following ICD-9 and ICD-10 procedure codes from the primary procedure identifier: 07.62 “Partial excision of pituitary gland, transsphenoidal approach”; 07.65 “Total excision of pituitary gland, transsphenoidal approach”; 0GT04ZZ “Endoscopic resection of pituitary gland”; 0GB04ZZ “Endoscopic excision of pituitary gland”; and 0GC04ZZ “Endoscopic extirpation of pituitary gland.” Once these data were aggregated, each record was then cross-referenced for the following ICD procedure code detailing computer guidance from the remaining procedure identifiers: 00.31 “Computer assisted surgery with CT/CTA”; 00.32 “Computer assisted surgery with MR/MRA”; 00.39 “Other computer assisted surgery”; 8E09XBG “Computer assisted procedure in head and neck region with CT”; 8E09XBH “Computer assisted procedure in head and neck region with MRI”; and 8E09XBZ “Computer assisted procedure in head and neck region, no qualifier.” To assess for surgical complications on the same admission, each record was queried for the following ICD 9 and 10 discharge diagnosis codes: 253.2 “Panhypopituitarism”; 349.81 “CSF rhinorrhea”; 253.5 “DI”; 378.51 “Oculomotor nerve palsy, partial”; 378.52 “Oculomotor nerve palsy, total”; 378.53 “Trochlear nerve palsy”; 378.54 “Abducens nerve palsy”; E23.0 “Panhypopituitarism”; G96.0 “CSF rhinorrhea”; E23.2 “DI”; H49.00 “CN III palsy”; H49.10 “CN IV palsy”; and H49.20 “CN VI palsy.”

Regression analysis was performed to determine whether independent variables including age, gender, race, and use of computer-assisted surgery were able to predict LOS, total charge, and surgical complication rates. Hospital care was aggregated across all years and transformed into 2016 U.S. dollar prices using the consumer price index. ¹⁴ Regression-based approaches to studying health services and health care costs have been well documented in the literature. ^{15 16 17} Cases were compiled into the statistical software SPSS version 24.0 (IBM Corporation, Armonk, New York, United States) for analysis. Multivariate logistic regression (MLogR) was utilized to determine the odds ratio of developing panhypopituitarism, diabetes insipidus (DI), CSF rhinorrhea, and cranial nerve (CN) palsies for image-guided versus nonimage-guided cases while controlling for age, gender, race, total number of procedures, and total number of diagnoses. A generalized linear model (GLM), with an underlying Gaussian distribution and an identity link function, was used to determine the effect of image guidance on total cost and LOS while holding other patient-related factors constant. The coefficients of the MLogR and GLM, which describe the change in the dependent variable for a unit increase in the independent variable, were tested for statistical significance using the Wald's test. Coefficients were considered statistically significant predictors if the p -value was less than or equal to 0.05.

Results

Querying the KID for the aforementioned ICD procedural codes returned a total of 1,297 cases of TSPR between 1997 and 2016. The majority of cases were performed in those who were female (62%), Caucasian (49%), and older than 15 years (66%) ( Table 1 ). The use of image guidance increased drastically between 1997 and 2016. Computer-assisted surgery was used in 0% of cases in 1997. While accounting for weighted discharges, computer-assisted surgery was used in 3.2% of cases starting in 2006 and steadily increasing to 34.5% by 2016 ( Fig. 1 ).

Table 1. Demographics, LOS, total cost, and complications for all TSPR cases in KID from 1997 to 2016.

	1997	2000	2003	2006	2009	2012	2016
Number of cases	88	165	208	175	303	230	128
With image guidance (%)	0.0	0.6	0.0	3.2	11.9	16.4	34.5
Without image guidance (%)	100.0	99.4	100.0	96.8	88.1	83.6	65.5
Age (%)
< 10 y	5.7	7.9	2.9	5.7	5.3	5.7	4.7
10–15 y	39.8	20.6	26.9	28.6	27.7	25.7	35.2
> 15 y	54.5	71.5	70.2	65.7	67.0	68.7	60.2
Race (%)
White	58.0	51.5	40.9	39.4	50.8	54.3	53.1
Black	5.7	9.1	8.7	5.1	11.6	10.9	15.6
Hispanic	8.0	15.2	13.9	16.0	18.2	15.7	13.3
Asian	6.8	4.8	1.4	3.4	2.3	3.0	2.3
Other	0.0	0.0	0.0	0.0	1.0	1.3	2.3
Unknown	21.6	19.4	35.1	36.0	16.2	14.8	13.3
Sex (%)
Male	30.7	33.3	32.7	41.4	32.0	38.3	39.1
Female	69.3	66.7	63.9	52.0	65.7	61.7	60.9
Average LOS (d)	4.7	4.1	5.9	4.5	4.6	4.7	5.6
SD of LOS (d)	2.9	2.1	16.2	3.3	4.2	4.0	6.5
Average total cost ($)	31,665.46	33,835.97	48,435.95	51,167.55	70,857.75	79,381.90	122,496.87
SD of total cost ($)	19,724.09	20,232.51	80,814.77	31,867.12	66,712.38	63,427.68	120,864.75
Complications (%)
Panhypopituitarism	6.8	6.1	6.3	10.3	9.6	10.0	21.9
CSF rhinorrhea	3.4	0.0	3.4	3.4	2.3	1.7	21.9
Diabetes insipidus	27.3	17.0	24.0	29.7	26.4	35.7	43.8
CN palsies	1.1	0.6	1.0	1.1	0.7	1.7	0.8

Abbreviations: CN, cranial nerve; CSF, cerebrospinal fluid; KID, Kids' Inpatient Database; LOS, length of stay; SD, standard deviation; TSPR, transsphenoidal pituitary resection.

Note : Average total costs have been transformed to 2016 U.S. dollar amounts using the consumer price index to account for inflation.

Fig. 1.

Increase in utilization of image guidance for pediatric transsphenoidal pituitary resection over time.

DI was found to be the most common surgical complication, occurring in 28.6% of all cases. When assessing surgical complication rates as a function of time, panhypopituitarism, CSF rhinorrhea, and DI all increased from 1997 to 2016, while CN palsies remained constant ( Table 1 ). When stratified by the use of image guidance, differences in raw complication rates appear negligible across the board; however, when utilizing MLogR to control for other variables, image-guided cases were found to have a lower risk of DI ( p  = 0.05) ( Table 2 ). Odds ratios ( p -values) of panhypopituitarism, CSF rhinorrhea, DI, and CN palsies for image-guided versus nonimage-guided cases are 1.367 (0.29), 1.012 (0.98), 0.621 (0.05), and 2.096 (0.31), respectively ( Table 3 ).

Table 2. Raw complication rates for all TSPR cases in 2016 separated by the use of image guidance.

Complication	Non-image guided (%)	Image guided (%)
Panhypopituitarism	20.24	25.00
CSF rhinorrhea	21.43	22.73
Diabetes insipidus	42.86	45.45
CN palsies	0.00	2.27

Abbreviations: CN, cranial nerve; CSF, cerebrospinal fluid; TSPR, transsphenoidal pituitary resection.

Table 3. Odds ratio and p -values for complications when image guidance was used compared with when it was not used .

Complication	Odds ratio	95.0% confidence interval	p -Value
Panhypopituitarism	1.367	0.764–2.446	0.29
CSF rhinorrhea	1.012	0.447–2.292	0.98
Diabetes insipidus	0.621	0.385–1.00	0.05
CN palsies	2.096	0.500–8.787	0.31

Abbreviations: CN, cranial nerve; CSF, cerebrospinal fluid.

For all TSPR, LOS was relatively constant over this time period with an average LOS (±standard deviation) of 4.7 ± 2.9 days in 1997 to 5.6 ± 6.5 days in 2016 ( Table 1 ). Regression analysis revealed that the use of image guidance resulted in a shorter LOS by 2.84 days compared with nonimage-guided cases ( p  < 0.001) ( Table 4 ). On the contrary, the total cost for admission steadily rose from $31,665 ± 19,724 in 1997 to $122,496 ± 120,864 in 2016 even when accounting for inflation ( Table 1 ). When further stratified, the use of image guidance did not demonstrate an increase in total charge for the hospitalization ( p  = 0.663) ( Table 4 ).

Table 4. Regression analysis to assess the effects of image guidance on length of stay and total hospital charge.

	Use of image guidance
	Marginal effects	p -Value	95% CI
Length of stay (d)	−2.84	<0.001	−4.19, −1.49
Total hospital charge ($)	3,480.60	0.663	−12,195, 19,156

Abbreviation: CI, confidence interval.

Note : While accounting for patient age, sex, race, and number of diagnoses/procedures performed in the hospital.

Discussion

Pituitary surgery has seen an incredible transformation guided by technological advancements since the first documented transnasal case in 1907. ¹ It began with headlamp illumination for visualization, but Norman Dott, a Scottish neurosurgeon, developed the lighted nasal speculum to provide better illumination in the narrow field. ¹ Gerard Guiot, a French neurosurgeon inspired by Dott's techniques, later introduced the X-ray film intensifier and fluoroscopy for intraoperative image guidance. ¹ Guiot is also credited as being the first neurosurgeon to use an endoscope for transsphenoidal surgery, which was performed in 1961, despite him later abandoning it due to poor visualization. ¹ Four years later, Jules Hardy, a fellow under Guiot, made a paradigm shift in the transsphenoidal approach when he began using a microscope for resection in 1965. ¹ This allowed, for the first time, to differentiate between tumor and normal pituitary gland. Thirty years later, the endoscope was reintroduced into the approach after significant improvements in illumination and image quality were made, largely replacing the microscope as standard of care. Finally, frameless stereotactic technology utilizing computed tomography (CT) or magnetic resonance imaging was then introduced in the 1990s. ¹⁸ Since then its utilization has grown and has become the standard of care at many skull base centers around the nation given its added safety and accuracy. ^{4 5 6 7 8} For example, it has been shown that there is a discrepancy between the surgeon's estimation of sella margins without image guidance and the accuracy of frameless stereotaxy—4.5 ± 3mm and 1.3 ± 0.6 mm, respectively. ^{19 20}

A large meta-analysis recently demonstrated that the use of image guidance in selected populations decreases the risk of major and total complications in endoscopic sinus surgery. ²¹ As there are much fewer cases for TSPR than sinus surgery, the volume of data regarding image guidance use is substantially less. The largest study investigating its use in TSPR was a population-based analysis conducted with the Nationwide Inpatient Sample, an adult equivalent of the KID. It demonstrated that the use of image guidance is associated with a lower rate of CSF rhinorrhea, shorter LOS, and lower cost. ⁹ The same had not yet been studied in the pediatric population. This is a consequence of the infrequency of pediatric pituitary lesions, low necessity for surgery, and that the traditional transcranial approach is still more frequently performed than in adults. ^{10 12}

Pediatric TSPR data are mostly limited to two larger case series and one population-based analysis. Jane et al studied 22 pediatric craniopharyngioma patients who underwent transsphenoidal resection. They saw a gross total resection rate of 68% and vision improvement in 64%, despite a 67% development of panhypopituitarism and 56% rate of DI. ¹² Zhan et al looked at 56 pediatric patients with pituitary adenomas undergoing TSPR and noted a total resection rate of 87.5%. ¹¹ Postoperatively, 7.1% developed panhypopituitarism, 8.9% developed DI, and 3.5% developed a CSF leak. ¹¹ Finally, Hanba et al also utilized the KID to assess total cost, LOS, and complication rates of TSPR. While they only assessed the 2009 and 2012 datasets, they found the risks of panhypopituitarism, DI, CSF rhinorrhea, and visual changes to be 10.2, 30.4, 2.3, and 6.4%, respectively. ¹⁰ Not only were these complication rates lower than their transfrontal counterparts but transsphenoidal patients also had a shorter LOS and total cost. ¹⁰ With the support of these data, it is not surprising that pediatric skull base centers have followed suit with widespread adoption of TSPR; however, data on image guidance are still lacking in this population.

This is the first study that demonstrates the increasing utilization of image guidance in pediatric TSPR cases. While introduced in the adult community in the 1990s, image guidance did not gain traction in the pediatric realm until 2006. Since then, its utilization has increased dramatically, likely in part due to the comfort and reassurance, it provides by improving surgical accuracy and spatial resolution. Postec et al demonstrated a mean accuracy of 1.1 mm when used for pediatric sinus surgery cases, which has been corroborated in other studies. ^{22 23 24}

Given the long-term population-based analysis of this study, it has also provided a unique insight into the evaluation of complication rates. Of note, it is the largest study to evaluate TSPR complication rates in the pediatric population. Overall, DI was seen to be the most common complication with rates similar to prior reports. ^{10 12} As a function of time, panhypopituitarism, CSF rhinorrhea, and DI displayed an increase in frequency from 1997 to 2016. This may simply be due to an increase in reporting or coding but could also be a product of more aggressive tumor resection or choosing to operate on more advanced disease. Finally, when stratified by the use of image guidance, raw complication rates appear similar between the two groups. For example, when isolating the 2016 data to control for possible coding disparities between years, all complications show image-guided rates to be within 5% of nonguided rates; however, when analyzing the entire dataset using MLogR to control for other factors, the risk of DI is lower with the use of image guidance ( p  = 0.05). The fact that this is on the cusp of significance and the other complications were not suggests that either the study was underpowered to detect a significance in the other complications or there is truly no difference in surgical complications with the use of image guidance. In general, as a large proportion of image guidance is CT based, good soft tissue delineation between tumor and normal pituitary gland remains difficult. Therefore, it may not be surprising that image guidance does not decrease the risk of injury to normal pituitary tissue and surrounding structures and thus certain complication rates. One shortcoming of this study is that the majority of computer-assisted surgery codes were of the “unspecified” or “no qualifier” type, thus not allowing direct comparison between CT and magnetic resonance–guided complication rates.

Regarding LOS, while it appears to have overall remained constant from 1997 to 2016 for all cases of TSPR, regression analysis demonstrated that addition of image guidance decreased LOS by 2.84 days ( p  < 0.001). This could be attributed to either increased comfort of early postoperative discharge or the decreased rate of DI, as Hanba et al demonstrated that DI was associated with greater hospital charges and LOS after TSPR. ¹⁰ Corresponding total charge rose precipitously, notably greater than the rate of inflation, from $31,665 in 1997 to $122,496 in 2016. When further stratified, the use of image guidance was not associated with an increase or decrease in total charge for the hospitalization ( p  = 0.663). The theoretically anticipated decrease in total charge from the decreased rate of DI could be offset by an increase in operating room time and cost attributed to image guidance registration and repeated location confirmation intraoperatively. These two metrics further support the use of image guidance in pediatric TSPR cases.

This is the largest study investigating pediatric TSPR, and the only known study evaluating trends and outcomes for image guidance within that population. Despite this, there are limitations of this study. First, since the datasets utilized encompass both ICD-9 and ICD-10 codes, there is a possibility of coding discrepancies between editions and years. While small in number, there are likely some coding inaccuracies due to human error that cannot be delineated in this dataset. Second, the KID does not include any data regarding tumor size, Knosp's classification, suprasellar extension, revision status, pathology, or methodology of skull base repair. These could be confounding factors as image guidance may have been used more frequently in extensive or revision cases necessitating more operating room time and overall cost, as well as contribute to higher complication rates and LOS. Third, while the ICD-10 pituitary procedure codes are specific for use of an endoscope, the ICD-9 procedure codes are not. As such, even despite the declining use of a purely microscopic approach to TSPR, the KID did not allow for assurance of a purely endoscopic approach across all years. Fourth, given a lack of admission ICD diagnosis codes in the database, it is unable to determine if any CN palsies present in discharge ICD diagnosis codes were present preoperatively. Finally, additional outcomes not available in the KID may be useful for further evaluation of image guidance for pediatric TSPR. Other factors such as rate of gross total resection, tumor recurrence, and improvement in surgical complications may be insightful. Further investigation is necessary to elucidate the effects of TSPR and image guidance on these metrics. Despite these shortcomings of the database, the population-based, multicenter, longitudinal approach to this study makes it unique, generalizable, and provides external validity and statistical power to a needed body of literature.

Conclusion

Originally developed for adult transsphenoidal pituitary surgery, the use of image guidance has multiplied in the pediatric skull base community for similar reasons related to safety, accuracy, and reassurance. Its utilization has rapidly increased since 1997, as image guidance may reduce complications, such as DI, the most common pediatric complication of TSPR. Finally, the use of image guidance also portends a shorter hospital stay without increasing the overall cost of the admission. These data shed light on an understudied area of clinical practice and reinforce the use of image guidance in pediatric transsphenoidal cases.

Funding Statement

Funding None.