Surgery (2017) -

ELSEVIER

Contents lists available at ScienceDirect

Surgery journal homepage: www.elsevier.com/locate/ymsy

SURGERY

+

3

.

American Association of Endocrine Surgeons

Less is more: cost-effectiveness analysis of surveillance strategies for small, nonfunctional, radiographically benign adrenal incidentalomas

Kathryn Chomsky-Higgins, MD, MS a,*, Carolyn Seib, MD, MAS ª, Holly Rochefort, MD ª, Jessica Gosnell, MD ª, Wen T. Shen, MD, MA ª, Jamess G. Kahn, MD, MPH b, Quan-Yang Duh, MD ª, and Insoo Suh, MD ª

ª UCSF Department of Surgery, University of California, San Francisco, San Francisco, CA

b UCSF Institute for Health Policy Studies, University of California, San Francisco, San Francisco, CA

ARTICLE INFO

Article history: Accepted 5 July 2017

Background. Guidelines for management of small adrenal incidentalomas are mutually inconsistent. No cost-effectiveness analysis has been performed to evaluate rigorously the relative merits of these strategies. Methods. We constructed a decision-analytic model to evaluate surveillance strategies for <4cm, non- functional, benign-appearing adrenal incidentalomas. We evaluated 4 surveillance strategies: none, one- time, annual for 2 years, and annual for 5 years. Threshold and sensitivity analyses assessed robustness of the model. Costs were represented in 2016 US dollars and health outcomes in quality-adjusted life-years. Results. No surveillance has an expected net cost of 158 more and adds 0.2 quality-adjusted life-years for an incremental cost- effectiveness ratio of $778/quality-adjusted life-years. The strategies involving more surveillance were dominated by the no surveillance and one-time surveillance strategies less effective and more expen- sive. Above a 0.7% prevalence of adrenocortical carcinoma, one-time surveillance was the most effective strategy. The results were robust to all sensitivity analyses of disease prevalence, sensitivity, and speci- ficity of diagnostic assays and imaging as well as health state utility.

Conclusion. For patients with a < 4cm, nonfunctional, benign-appearing mass, one-time follow-up eval- uation involving a noncontrast computed tomography and biochemical evaluation is cost-effective. Strategies requiring more surveillance accrue more cost without incremental benefit. (Surgery 2017;160:XXX-XXX.) 2017 Elsevier Inc. All rights reserved.

Increased use of cross-sectional imaging has led to an increas- ing prevalence of incidentally detected adrenal masses, so-called adrenal incidentalomas. Data on the natural history of such masses are lacking. Although it has been reasonably well established that masses <1 cm are unlikely to be pathologic, and those that are 4 cm or larger warrant resection, management of lesions that fall between these diameters remains a topic of controversy and inquiry.1-4

The optimal evaluation and follow-up of benign-appearing, nonfunctional adrenal incidentalomas is unknown. Official recom- mendations conflict; some recommend limited or no surveillance and others suggest more extensive monitoring.1,2 We hypoth- esized that a formal cost-utility analysis incorporating a cost- effectiveness analysis involving quality-adjusted life-years (QALYs) using population-based data could inform the current debate on the optimal management algorithm for adrenal incidentalomas,

perhaps suggesting a decreased surveillance for low risk lesions. We constructed a decision analytic model to test this hypothesis.

Methods

Model structure

A decision analytic model using Markov disease states was created using TreeAge Pro (TreeAge Software, Inc., Williamstown, MA). We compared 4 strategies for surveillance of a patient with an adrenal incidentaloma: no surveillance, single surveillance at 12 months, annual surveillance for 2 years, and annual surveillance for 5 years. Surveillance was defined as a noncontrast abdominal computed to- mography (CT), 1 mg dexamethasone suppression test, and plasma levels of renin, aldosterone, and metanephrine.

Model structure (Fig 1) and inputs were based on review of the literature on surgical management of adrenal masses in the United States and Europe. A health care system perspective was used. Costs of operation, complications, laboratory testing, office visits, and imaging were included. Non-healthcare-related costs were omitted. The time horizon was remaining life expectancy, with age-specific mortality determined by life tables from the Centers for Disease

Departmental funding was provided for the article.

Presented at the Annual Meeting of the American Association of Endocrine Sur- geons, April 2-4, 2017, Orlando FL.

* Reprint requests: Kathryn Chomsky-Higgins, MD, MS, Department of Endocrine Surgery, University of California, San Francisco, 1600 Divisadero Street, Box 1674, San Francisco, CA 94143-1674.

E-mail: kate.chomsky-higgins@ucsf.edu; kchomsky@alamedahealthsystem.org.

1-4cm Adrenal Incidentaloma: < 10HU on Noncontrast CT, Nonfunctional

Fig 1. Model structure, simplified. The clinical scenario is a 1 to 4cm incidentally discovered adrenal mass that was <10 HU on noncontrast CT and ruled out for primary aldosteronism, Cushing's syndrome, and pheochromocytoma on biochemical workup. We considered a strategy of no surveillance after the initial workup versus various surveillance regimens. In the no surveillance arm, we first established the proportion of patients in the population with actual disease. Major branch points included whether the disease became clinically apparent, whether the patient was then a candidate for adrenalectomy, and whether the operation involved a complication. In the surveil- lance arm, we considered the additional elements of the sensitivity, specificity, and risks of surveillance testing. The surveillance part of the model was replicated for three different surveillance strategies (1-time, 2-time yearly, and 5-time yearly surveillance) and compared against the no surveillance arm.

No Surveillance

Surveillance

Disease

No Disease

Disease

No Disease

Disease becomes clinically apparent

True positive

False negative

False positive

True negative

Disease becomes clinically apparent

Adrenalectomy +/- complication

Adrenalectomy +/- complication

Adrenalectomy +/- complication

Death

Normal lifespan

Normal Lifespan

Death

Death

Normal lifespan

Control and Prevention.5 The model cycled annually. Discounting of 3% was applied for both costs and utilities.

The base case considered a cohort of 40-year-old patients with incidentally discovered, 1 to 4cm adrenal masses and an initial normal workup, including noncontrast CT showing a mass with <10 Hounsfield Units (HU) and no suspicious features as well as evaluation for pheochromocytoma, hypercortisolism, and hy- peraldosteronism. The model addresses subsequent management. All patients were considered candidates for laparoscopic adrenal- ectomy except for those diagnosed with malignant disease when the mass became evident clinically (i.e., became symptomatic, either by mass effect or hyperfunction) as opposed to being discovered at the time of surveillance testing when the patient was asymptom- atic. Patients with disease discovered during surveillance were presumed to undergo operative intervention and to be cured, that is, return to normal life expectancy. Even in the case of adrenocor- tical carcinoma (ACC), patients with disease identified during surveillance underwent operation and were exposed to the risks of laparoscopic adrenalectomy (Table I) rather than the greater risks of adrenalectomy after clinical discovery of the disease (Table II). Conversely, patients with cancer metastatic to the adrenal gland from a nonadrenal primary malignancy were not assumed to be cured. Life expectancy after adrenal metastasectomy was available in the literature and was assigned to corresponding patients.13

Transition probabilities

Estimates of disease prevalence in the population referred to an endocrinologist or endocrine surgeon for monitoring of an incidentally discovered adrenal mass after initial workup were

found on review of the literature, as were the sensitivity and specificity of biochemical testing and imaging used for surveil- lance (Tables I and II).6 To account for disease-specific mortality associated with untreated disease, disease-specific hazard ratios or survival data were applied to those branches of the model pertaining to each pathology.7-9,15 Rates of complication and mor- tality for laparoscopic adrenalectomy were those applied in a previous cost-effectiveness analysis, using data derived originally from the Healthcare Cost and Utilization Project Nationwide Inpa- tient Sample.3,16 These data were stratified by age.

Our model required that a distinction be made between (1) the annual risk that a mass would progress to a point of becoming evident clinically, and (2) the annual risk that a mass would pro- gress to a point of becoming evident via surveillance mechanisms as opposed to the risk of simply growing or becoming functional. Values for these variables in our population of interest were not available. We instead assumed in the base case that the risk of true disease becoming apparent by surveillance was 4 times that of the risk of it becoming clinically evident in a given year, and then varied this ratio in sensitivity analysis by using very broad ranges of values.

Costs

Surveillance costs were ascertained from the 2016 Medicare reimbursement schedule.17 Treatment costs were adjusted to 2016 using the Consumer Price Index for urban consumers. Costs for complicated and uncomplicated laparoscopic adrenalectomy derived from the data of the Healthcare Cost and Utilization Project Na- tionwide Inpatient Sample and stratified by age were sourced

Chomsky-Higgins et al/Surgery (2017)-

Table I. Probability inputs, part 1
Input*ValueSensitivity analysis rangeReference
Prevalence values
Benign adenoma0.971N/A6
Subclinical Cushing's syndrome0.0060.001-0.0106
Aldosterone producing adenoma0.0010-0.0066
Pheochromocytoma0.0020.001-0.0046
Adrenocortical carcinoma0.0190-0.0386
Cancer metastatic to adrenal0.0010-0.0026
Accuracy of diagnostic testing
Benign adenoma (CT)0.750.68-0.896
Subclinical Cushing's syndrome (1 mg DXT)0.90±20%6
Aldosterone producing adenoma (renin/ aldosterone)0.950.90-16
Pheochromocytoma (plasma metanephrines)0.890.80-0.976
Adrenocortical carcinoma (CT)0.67+20%6
Cancer metastatic to adrenal (CT)0.95+20%6
Hazard ratios
Subclinical Cushing's syndrome1.441.2-1.737
Aldosterone producing adenoma1.441.2-1.738
Pheochromocytoma14.43.2-64.48
Risk of resectability (if not discovered by surveillance testing)
Adrenocortical carcinoma0.88+20%9
Cancer metastatic to adrenal0.5+20%10
Mortality risk, laparoscopic adrenalectomy
Age ≤400.003+20%3
Age 41-600.001
Age 61-700.005
Age >700.029
Complication risk, laparoscopic adrenalectomy
Age ≤400.062±20%3
Age 41-600.079
Age 61-700.085
Age >700.219

* Sourced from literature review. Two pathologies, aldosteronoma and ACC, were not found among the 1,410 patients followed in the Cawood review. For aldosteronoma, 0.001 was used as baseline prevalence and prevalence in the pop- ulation prior to initial evaluation was used as the upper limit for sensitivity analysis. For ACC, estimated prevalence in the population prior to initial evaluation was used, as estimates in the literature suggest a prevalence of ACC as high as 14%.11 We elected to err on the side of opposition to our hypothesis by assuming this higher ACC prevalence.

Table II. Probability inputs, part 2
InputValueSensitivity analysis rangeReference
Risk of death after lap adrenalectomy for ACC identified clinically, by years after adrenalectomy
<20.403±20%12
2 to <50.284
5 to 100.254
>100.060
Risk of death from unresectable ACC, by year after diagnosis
10.5±20%12
51.0
Risk of death after adrenal metastasectomy, by years after adrenalectomy
00.365±20%13
10.606
20.365
30.500
Risk of death with unresectable cancer metastatic to adrenal, metastatic colon cancer used as proxy
00.601±20%14
10.506
20.426
30.327
40.25
Annual risk of progression to being clinically evident (i.e., symptomatic)
Subclinical Cushing's syndrome0.20.1-0.9(Expert opinion)
Aldosterone producing adenoma0.20.1-0.9
Pheochromocytoma0.20.1-0.9
Adrenocortical carcinoma0.250.1-0.9
Cancer metastatic to adrenal0.250.1-0.9
Annual risk of progression to being evident by surveillance mechanisms
Subclinical Cushing's syndrome0.80.1-0.9(Expert opinion)
Aldosterone producing adenoma0.80.1-0.9
Pheochromocytoma0.80.1-0.9
Adrenocortical carcinoma0.80.1-0.9
Cancer metastatic to adrenal0.80.1-0.9

confidence intervals from the literature or +20% intervals when con- fidence intervals were unavailable. A probabilistic sensitivity analysis assessed aggregate parameter uncertainty, using triangular distributions around inputs found to contribute most to model variability.

from literature (Table III).3,16 Costs varied by age due to greater duration of stay and greater complication rates for older patients.

Health state utilities

Effectiveness was measured in QALYs. Per convention, the utility of one, fully healthy life year equals one QALY, and death is equiv- alent to zero QALYs. The National Institutes of Health EQ-5D Index Scores provided QALY values for “one year living with surveil- lance” and “utility of life after surgery for malignant disease” based on the value for “unspecified neoplasm”; utility for one year living with a long-term operative complication was approximated from the score for “abdominal hernia.”18 An effectiveness toll, a one- time QALY decrement, was taken for each CT the patient underwent (Table IV).19

Analysis

Base-case and sensitivity analyses were performed. Incremen- tal cost-effectiveness ratios were calculated where a strategy added benefit and costs. One-way sensitivity analyses examined the effects of uncertainty around base case inputs, using 95%

Table III. Cost inputs
Input*ValueSensitivity analysis rangeReference
Annual cost of surveillance (2016 US$)
Serum aldosterone$55.51±20%17
Plasma renin activity$29.96
Electrolyte panel$9.55
Serum metanephrines$23.07
Serum cortisol, total$22.21
CT scan, adrenal protocol$264.24
Office visit, established patient$43.68
Total$448.22
Cost lap adrenalectomy and hospitalization, without complication
Age ≤40$9,441±20%3
Age 41-60$10,473
Age 61-70$10,573
Age >70$11,440
Cost lap adrenalectomy and hospitalization, with complication
Age =40$15,553±20%3
Age 41-60$18,009
Age 61-70$16,435
Age >70$18,779

* Sourced from literature review. Costs varied by age due to longer duration of stay and higher complication rates for older patients.

Table IV. Utility inputs
Input*Value (QALYs)Sensitivity analysis rangeReference
Utility of surveillance One year of otherwise healthy life undergoing surveillance0.8270.778-118
Disutility of CT QALY toll for each CT, accounting for lifetime attributable risk of radiation0.004 × [0.5 ×life expectancy]±20%19
Utility "alive" Utility of one year of healthy life1
(Standard)
Utility "dead"
Utility of one year dead Disutility of uncomplicated adrenalectomy0(Standard)
QALY toll for undergoing surgery0.08±20%20
Disutility of complicated adrenalectomy
QALY toll for undergoing operation with complication0.16±20%20
Utility of life with malignancy Utility of one of year life with ACC or cancer metastatic to adrenal, not amenable to adrenalectomy0.8270.778-118

* Sourced from literature review.

Results

Base case

The approach of No surveillance would cost 158 more and provide 0.2 more QALYs for an incremental cost-effectiveness ratio of $778/QALY. Base case outcomes are presented in Table V and il- lustrated on the cost-effectiveness plane in Fig 2. The 2 strategies involving more surveillance were dominated by the no surveil- lance and one-time surveillance strategies, both strategies were less effective and more costly due to false positive test results, expo- sure to radiation, and unnecessary operative interventions. The number needed to treat with a second round of surveillance versus only one round to prevent a single death from ACC was 1,224.

Sensitivity analysis

One-way sensitivity analyses are diagrammed in Fig 3. Vari- ables with the most influence on results were the utility value in QALYs accrued by a patient during a year of surveillance and the

Fig 2. Base case results presented on cost-effectiveness plane. Increasing effective- ness is represented on the x axis, and increasing cost on the y axis. Two of our proposed strategies fell into the left upper quadrant. These are known as "domi- nated" strategies, in that they were both less effective and more costly. No strategies fell into the right lower quadrant, representing less costly, more effective strate- gies. These would be known as dominant strategies. The other 2 strategies, no surveillance and one-time surveillance, fell into the right upper quadrant. They were less costly and more effective than the more extensive surveillance strategies. Though one-time surveillance is slightly more costly, it is also more effective. Its ICER with respect to no surveillance is $778/QALY, clearly below our willingness-to-pay thresh- old of $150,000/QALY. ICER, incremental cost-effectiveness ratio.

5x

2x

Dominated strategies

At 1x surveillance strategy: ICER = $778/QALY

1x

Incremental Effectiveness

No surveillance

Incremental Cost

prevalence of ACC; however, only prevalence of ACC suggested a change in the optimal surveillance strategy. Above a threshold of 0.7% prevalence of ACC, one-time surveillance was the most effec- tive strategy at a willingness-to-pay threshold of $150,000. Below this prevalence, no surveillance was the preferred strategy. Results were robust to all the other sensitivity analyses disease preva- lence, sensitivity and specificity of diagnostic assays and imaging, and health state utility.

We also looked at the effect of age on the predictions by the model as illustrated in Fig 4. Once again, more extensive (2-time and 5-time surveillance) were dominated by the no surveillance and one-time surveillance strategies in all age groups. One-time sur- veillance was cost-effective for patients <60 years old; for patients >60 years old, this strategy also was dominated by the no surveil- lance strategy.

A probabilistic sensitivity analysis (second order Monte Carlo anal- ysis) was performed to examine parameter uncertainty in the model. A cost-effectiveness scatterplot illustrates the analysis in Fig 5. In 96% of 10,000 iterations, the one-time surveillance strategy was found to be cost-effective at a willingness-to-pay threshold of $150,000.

Discussion

Current guidelines of professional medical societies differ in their recommendations regarding the management of an adrenal

Table V. Base case outcomes
Surveillance strategy*Cost (2016 US$)Incremental cost (2016 US$)Effectiveness (QALYs)Incremental effectiveness (QALYs)Incremental cost-effectiveness ($/QALY)
No surveillance262n/a26.22n/a0
One-time surveillance42015826.420.2778
Two-time annual surveillance7,0816,66225.96-0.46Dominatedț
One-time per year for 5 years9,9069,48725.77-0.66Dominatedț

* Outcomes for the base case as predicted by our model. Please note that global discounting impacts the calculations in such a way that direct calculation of the ICER from the values here will return results slightly different compared with values produced by the software.

+ By the no-surveillance and one-time surveillance strategies.

Fig 3. Tornado diagram. This tornado diagram illustrates a compilation of one-way sensitivity analyses on model inputs. The x axis is presented in Net Health Benefits, a composite value representing cost, effectiveness, and willingness-to-pay. Prevalence of adrenocortical carcinoma exerted the most influence and was the only factor with a threshold indicated by the solid vertical line. Further analysis showed this threshold to be 0.7%, greater than which the optimal strategy was one-time surveillance, and less than which it was no surveillance. SCS, subclinical Cushing's syndrome; metastasis, disease metastatic to adrenal from another primary site; pheo, pheochromocytoma; aldo, aldosteronoma.

Prevalence ACC

Utility of surveillance

Prevalence metastasis

Mortality hazard ratio, pheochromocytoma

Progression ACC to clinically apparent, annual risk

Prevalence pheochromocytoma

Progression metastasis to clinically apparent, annual risk

Progression pheo/ aldo/ SCS to clinically apparent, annual risk

Prevalence subclinical Cushing’s syndrome

Prevalence aldosteronoma

Base case 26.42

26.3

26.35

26.4

26.45

26.5

26.55

26.6

Net Health Benefits

incidentaloma. The guidelines of the American Association of Clin- ical Endocrinologists and American Association of Endocrine Surgeons recommend resection of incidentally discovered adrenal masses that are hormonally active or of diameter >4 cm. For those masses that do not fulfill the criteria for resection, radiographic eval- uation is recommended at 3 to 6 months after initial diagnosis, then annually for 1 to 2 years, and hormonal evaluation is recommended

annually for 5 years.1 Guidelines published in 2016 by the Europe- an Society of Endocrinology and the European Network for the Study of Adrenal Tumors suggest much less aggressive surveillance.2 These recommendations ascribe more importance to the characteristics on cross sectional imaging than size in the determination of the risk of malignancy.21,22 Specifically, based on a high negative predictive value of noncontrast CT for ACC, they recommend that for

Fig 4. Effect of age on model predictions. The predictions of our model of cost-effectiveness are plotted here by decade of age on the cost-effectiveness plane. Each decade of age is represented by a line connecting data points representing each potential surveillance strategy. Again, the once annual surveillance for 2 years and 5 years were dominated in all age groups. One-time surveillance was cost-effective for those <60 years old. For those >60 years old, this too was dominate by the no surveillance and one-time surveillance strategies. The trend shows that for younger age groups, it is more effective and minimally costly to perform one-time surveillance. For older age groups, >60 years of age, no surveillance may be the only cost-effective strategy.

5x

Incremental Cost

2x

No surveillance

1x

Incremental Effectiveness

Age:

80 70

60 50

40

30

20

Fig 5. Monte Carlo analysis. The results of a Monte Carlo analysis run with 10,000 iterations demonstrate consistently that the 2-time yearly and 5-time yearly surveillance strategies were dominated by the no surveillance and one-time surveillance strategies. One-time surveillance was found to be cost-effective in 96% of iterations.

10,500

10,000

Five-time surveillance, annual

9,500

9,000

8,500

8,000

7,500

7,000

Two-time surveillance, annual

6,500

Cost, $US

6,000

5,500

5,000

4,500

4,000

3,500

3,000

2,500

2,000

1,500

1,000

No surveillance

One-time surveillance

500

0

25.50

25.60

25.70

25.80

25.90

26.00

26.10

26.20

26.30

26.40

26.50

26.60

Effectiveness, QALYs

patients with an initial noncontrast CT showing the mass as <10 HU and biochemical workup indicating a nonfunctional lesion, no further surveillance is indicated.

Our study suggests that an optimal strategy might be closer to the new European guidelines, though we would still advocate for 1 time. Repeat, noncontrast CT, and biochemical analysis after the initial workup, particularly for patients <60 years old. This result is based largely on the understood prevalence of ACC in our pop- ulation of interest, that is, patients with small adrenal masses that are completely benign-appearing and nonfunctional at time of de- tection. The estimate of prevalence used in our model errs toward overestimation intentionally. A lesser prevalence of ACC is associ- ated with no surveillance as the most cost-effective strategy. If we assume a greater prevalence of ACC as used in our model and compare 1-time versus 2-time surveillance, the number needed to treat to avoid one death due to ACC is 1,224. The true prevalence among patients who have undergone initial testing, however, is more likely <1 in 1,000, given that in a review of 1,400 patients combin- ing the results of multiple studies, no such mass developed into an ACC.6 This observation would suggest a number needed to treat of 23,256.

There are some striking findings related to cost as predicted by this model. First, there is a large cost difference between the 1-time and 2-time surveillance strategies. We investigated this thor- oughly by interrogating the Markov analyses within the model and showed that the large increase is attributable to the costs of un- necessary operative interventions (and associated complications) that result from false-positive surveillance testing on a population with an extremely low prevalence of disease. The assumptions made in the model (e.g., all who have positive testing undergo opera- tion) may inflate this numerical difference somewhat, but the underlying pattern to the results and the conclusion drawn from these results are stable. Second, there is only a small cost differ- ence between the no-surveillance and one-time surveillance strategies. It is important to appreciate that the model accounts for the difference between not only the input costs, but also the down- stream effects of each strategy. Again, interrogation of the Markov analyses shows that one-time surveillance does a good job of iden- tifying the few patients who either were the victims of false- negative initial testing or developed disease in the time between

initial and subsequent workup. There are costs associated with op- erative intervention and treatment of these patients, but there are also similar costs associated with patients in the observation (no- surveillance) branch of the model who have disease that becomes clinically important. In both scenarios, these costs are spread across the population.

The present study has several limitations. Most notably, reli- able information regarding the natural history of adrenal incidentaloma is scarce. Selection bias clouds interpretation of data from operative and oncologic series, which tend to report an in- flated incidence. True incidence of each of the pathologies addressed here is not available. Subclinical Cushing’s syndrome presents a par- ticular challenge, given the lack of consistent definition, unclear risk associated with asymptomatic autonomous cortisol secretion, and absence of clear indication for operative intervention.2,23 Our model would support an individualized approach to these patients with clinical management by an endocrinologist consistent with previ- ous recommendations.2 Second, we combined all surveillance (both CT and laboratory testing) into a composite value in our model. Others have found that the risk of development of hypercortisolism and pheochromocytoma is small but most relevant at 36 to 42 and 48 to 54 months after initial discovery of an incidental adrenal mass.24 We did perform a variation of the model that considered how the continuation of biochemical surveillance yearly for 5 years in all surveillance groups (even after discontinuation of imaging sur- veillance) would impact our results. The impact was extremely small and did not change our conclusions. Still, continued biochemical sur- veillance after discontinuation of imaging could be further investigated particularly in selected patient groups. Third, esti- mates for operative complications and mortality are sourced from the best data available but may also overstate risk. This overesti- mate would decrease the number of QALYs accrued for treated disease and overestimate the number that failed to accrue due to false positive results. In addition, these estimates are associated with age. We would have preferred to use estimates associated with frailty, had such estimates been available, because this approach would reflect the operative risk more accurately. Finally, as in any model, several assumptions were made, including: all patients with pos- itive testing undergo operation, all patients with ACC found by surveillance testing are cured with adrenalectomy, and all patients

are candidates for laparoscopic adrenalectomy. Clearly, these as- sumptions may not always represent real-world behavior of disease. Additional research on the long-term natural history of adrenal incidentaloma would add greatly to our ability to predict the safest and most effective course of management for these patients.

To our knowledge, this is the first cost-utility analysis evaluat- ing surveillance strategies for small adrenal incidentalomas. For those patients with a < 4cm, nonfunctional, benign-appearing mass, a one- time follow-up evaluation involving noncontrast CT and biochemical evaluation is cost effective, and additional surveillance beyond this is not cost effective. While allowing for individualized manage- ment, we suggest that one-time surveillance may be a safe and potentially superior alternative to more extensive regimens for pa- tients in this population.

References

1. Zeiger MA, Thompson GB, Duh Q-Y, et al. The American Association of Clinical Endocrinologists and American Association of Endocrine Surgeons medical guidelines for the management of adrenal incidentalomas. Endocr Pract 2009;15(suppl 1):1-20.

2. Fassnacht M, Arlt W, Bancos I, et al. Management of adrenal incidentalomas: European Society of Endocrinology clinical practice guideline in collaboration with the European Network for the Study of Adrenal Tumors. Eur J Endocrinol 2016;175:G1-34.

3. Wang TS, Cheung K, Roman SA, Sosa JA. A cost-effectiveness analysis of adrenalectomy for nonfunctional adrenal incidentalomas: is there a size threshold for resection? Surgery 2012;152:1125-32.

4. Young WF. Clinical practice. The incidentally discovered adrenal mass. N Engl J Med 2007;356:601-10.

5. Centers for Disease Control and Prevention. CDC national vital statistics report. Volume 64, number 4; 2016.

6. Cawood TJ, Hunt PJ, O’Shea D, Cole D, Soule S. Recommended evaluation of adrenal incidentalomas is costly, has high false-positive rates and confers a risk of fatal cancer that is similar to the risk of the adrenal lesion becoming malignant; time for a rethink? Eur J Endocrinol 2009; 161:513-27.

7. Willey JZ, Moon YP, Kahn E, et al. Population attributable risks of hypertension and diabetes for cardiovascular disease and stroke in the northern Manhattan study. J Am Heart Assoc 2014;3:e001106.

8. Stolk RF, Bakx C, Mulder J, Timmers HJLM, Lenders JWM. Is the excess cardiovascular morbidity in pheochromocytoma related to blood pressure or to catecholamines? J Clin Endocrinol Metab 2013;98:1100-6.

9. Kebebew E, Reiff E, Duh Q-Y, Clark OH, McMillan A. Extent of disease at presentation and outcome for adrenocortical carcinoma: have we made progress? World J Surg 2006;30:872-8.

10. Bradley CT, Strong VE. Surgical management of adrenal metastases. J Surg Oncol 2014;109:31-5.

11. Mansmann G, Lau J, Balk E, Rothberg M, Miyachi Y, Bornstein SR. The clinically inapparent adrenal mass: update in diagnosis and management. Endocr Rev 2004;25:309-40.

12. Tran TB, Postlewait LM, Maithel SK, et al. Actual 10-year survivors following resection of adrenocortical carcinoma. J Surg Oncol 2016;114:971-6.

13. Lo CY, van Heerden JA, Soreide JA, et al. Adrenalectomy for metastatic disease to the adrenal glands. Br J Surg 1996;83:528-31.

14. Edge SB, Compton CC. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann Surg Oncol 2010;17:1471-4.

15. Sidhu S, Sywak M, Robinson B, Delbridge L. Adrenocortical cancer: recent clinical and molecular advances. Curr Opin Oncol 2004;16:13-8.

16. Park HS, Roman SA, Sosa JA. Outcomes from 3144 adrenalectomies in the United States: which matters more, surgeon volume or specialty? Arch Surg 2009;144:1060-7.

17. U.S. Centers for Medicare & Medicaid Services. Medicare physician fee schedule; 2017. https://www.cms.gov/apps/physician-fee-schedule/search/search- criteria.aspx. Accessed January 24, 2017.

18. Sullivan PW, Ghushchyan V. Preference-based EQ-5D index scores for chronic conditions in the United States. Med Decis Making 2006;26:410-20.

19. Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated lifetime attributable risk of cancer. Arch Intern Med 2009; 169:2078-86.

20. Dovirak O, Mao J, Taylor K, Chang P, Wagner AA. How to quantify recovery after laparoscopic adrenalectomy: an assessment of patient-reported health- related quality of life. Surg Laparosc Endosc Percutan Tech 2016;26:290-4.

21. Duh Q-Y. Adrenal gland: new guidelines for adrenal incidentalomas. Nat Rev Endocrinol 2016;12:561-2.

22. Dinnes J, Bancos I, Ferrante di Ruffano L, et al. Management of endocrine disease: imaging for the diagnosis of malignancy in incidentally discovered adrenal masses: a systematic review and meta-analysis. Eur J Endocrinol 2016; 175:R51-64.

23. Starker LF, Kunstman JW, Carling T. Subclinical Cushing syndrome: a review. Surg Clin North Am 2014;94:657-68.

24. Loh HH, Yee A, Loh HS, Sukor N, Kamaruddin NA. The natural progression and outcomes of adrenal incidentaloma: a systematic review and meta-analysis. Minerva Endocrinol 2017;42:77-87.

Discussion

Dr Kyle A. Zanocco (Los Angeles, CA): Congratulations on a well- executed study. I just have one question.

Could you provide what the overall baseline probability esti- mate was for a patient developing an indication for surgery during the surveillance period? And also, was a sensitivity analysis done on that probability? And it would be interesting to know what the threshold probability of that number would be.

Dr Kathryn H. Chomsky Higgins: Any patient with any type of pathology, what was the basis for them developing an indication for surgery? So one of the challenges of the study is determining the risk of becoming detectable on surveillance versus clinical de- tection. And largely what we had to do in this case, because there’s not good evidence for the differentiation between those 2, was use expert opinion and then do very wide confidence intervals.

So we began at a baseline of looking at a 20% risk of progres- sion per year to becoming clinically evident, and then an 80% progression to becoming evident on surveillance. And then, as I said, we used wide confidence intervals in our sensitivity analyses.

Dr Rebecca S. Sippel R. (Madison, WI): I really enjoyed your talk. What I’m looking at and trying to understand is, you were saying that it’s most cost effective to do one follow-up, which I think we agree.

I think the question is, what’s the right timing of that? It really depends on what you’re trying to capture with it, and I can’t tell

from the analysis. It seems like what was driving it was a very small chance that this was an ACC. And if that’s the case, then we have to do that follow up earlier or it’s not going to be helpful.

But the reality is probably most of the problem in long-term follow-up is they develop subclinical hypercortisolism, which may not be captured at 1 year, but instead that follow-up might be better at 3 to 5 years. And can you answer if we’re going to do 1 follow- up, when should it be? Is it an immediate 1-year follow-up, or are we better off at 3 to 5 years?

Dr Kathryn H. Chomsky Higgins: I agree with both of those points. And I think in terms of the timing, I think one of the points of data that’s not clear is exactly precisely when we can expect cancer to grow to a point where we would be concerned about it, at what rate and at what volumetric rate it would grow when we would find it. And so we didn’t really look at-because we didn’t have that data, we couldn’t say exactly when the best actual timing was.

In terms of subclinical Cushing’s, there are certainly data that suggest that it’s more likely to develop 3 to 5 years out, or in a couple of different time ranges at 3 to 5 years out. And I think that it’s perfectly reasonable to do a more focused analysis for that pathology, perhaps not involving a CT, because it would be less relevant, but doing a biochemical work up at that time.

Chomsky-Higgins et al/Surgery (2017) -

Dr Carrie C. Lubitz (Boston, MA): A couple questions. First of all, I believe in your base case you said in the beginning, these were patients with Houndsfield units with less than 10.

Was the prevalence of ACC adjusted for that? The pretest prob- ability would probably be much less.

My second question is, and I know it’s only an 8-minute pre- sentation, you didn’t go into what data you used for the utilities. So, I think that’s very relevant in these studies where the length of life and quality of life, the length of life is not going to be much dif- ferent, but the quality adjustment for that is extremely important for the input parameters. Did you do a sensitivity analysis on that, as well?

Dr Kathryn H. Chomsky Higgins: No, we did not.

We used one prevalence for ACC for less than 10 Houndsfield units and we didn’t adjust for various level of Houndsfield units. In terms of utilities, that is a real challenge because there’s not a good set of utilities for what it is to live with surveillance for adrenal incidentaloma.

We found in our literature review a very nice paper that tried to its goal was to standardize utilities in the US population, and we used that, and we tried to use approximations, living for example with a non-specific neoplasm. And we, of course, did do sensitiv- ity analyses in all of those. And we tried again to use relatively wide confidence intervals because we didn’t have a good match up of that data.

Dr Carrie C. Lubitz (Boston, MA): One follow-up. Was there a consequence of delayed diagnosis, at least in terms of the utility for the patient or disutility for the patient?

Dr Kathryn H. Chomsky Higgins: No, that didn’t have an effect on the model.

Dr Michael J. Campbell (Sacramento, CA): Great study. I really enjoyed it.

I have a specific question, because what comes up in my clinic, unfortunately, is not the patient with the CT with the Houndsfield unit less than 10 as Dr. Lubitz pointed out. It’s actually the patient with wash out suggestive of an adenoma or an MRI that drops out, suggestive of an adenoma. Knowing that the data with that is

somewhat less suggestive of a conclusive adenoma, were you able to adjust your analysis to see whether she should follow those pa- tients with further studies? And how does that change how often you should follow them?

Dr Kathryn H. Chomsky Higgins: In this study, we specifically used a sensitivity for non-contrast CT. And so we certainly could adjust the model and incorporate a different sensitivity for any of those types of imaging studies and get a result from that, but we didn’t do that as part of this study. I think probably the other im- portant point that your point brings up, is this is really just a model and you have to look at exactly what we put into it and adjust for your clinical situation.

Dr David F. Schneider (Madison, WI): Great study. Really im- portant question, and I applaud you for taking it on.

I have a follow-up question about the utilities, as well. So you showed that you did the sensitivity analysis for the utility of being under surveillance. So I’m wondering about your point estimate or your estimate for utility of not being under surveillance, because I think that’s equally important.

Dr Kathryn H. Chomsky Higgins: Utility of not being under surveillance?

Dr David F. Schneider (Madison, WI): So that was your other arm, right?

You had 3 different sort of schemes, right? No surveillance, at one time point and then more aggressive. So I’m just wondering, I think that those patients that are not being sur- veilled, there’s probably some different health state preference for being in that state as well because they knew they had something there.

Dr Kathryn H. Chomsky Higgins: So not being surveilled but knowing they have a disease, we treated those as a normal health state.

No, I’m sorry, we treated those as a nonspecified neoplasm, because it was the worry of having that. And then there was an ad- ditional disutility actually for undergoing the surveillance, so you actually lost something by undergoing a round of surveillance each time.