The Economic Implications of Three Biochemical Screening Algorithms for Pheochromocytoma

ANNA M. SAWKA, AMIRAM GAFNI, LEHANA THABANE, AND WILLIAM F. YOUNG, JR. Division of Endocrinology, Metabolism, Nutrition, and Internal Medicine (W.F.Y.), Mayo Clinic, Rochester, Minnesota 55905; Department of Internal Medicine and Division of Endocrinology (A.M.S.), St. Joseph’s Healthcare, Hamilton, Ontario, Canada L8N 4A6; Department of Internal Medicine and Division of Endocrinology (A.M.S.), McMaster University, Hamilton, Ontario, Canada L8N 3Z5; Centre for Evaluation of Medicines (L.T.), St. Joseph’s Healthcare, Hamilton, Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 Ontario, Canada L8N 1G6; and Department of Clinical Epidemiology and Biostatistics (L.T., A.G.), McMaster University, Hamilton, Ontario, Canada L8N 3Z5

Pheochromocytoma is a rare, life-threatening condition. Using offs would undergo 24-h urinary measurements (total meta- a modeling technique, we studied the economic implications of nephrines and fractionated catecholamines) and be imaged if detection strategies for pheochromocytoma (third-party payer positive. We determined that, if 100,000 hypertensive patients perspective). The diagnostic efficacy of biochemical tests was (including 500 patients with pheochromocytoma) were tested, based on Mayo Clinic Rochester data. In all hypothetical algo- algorithm A (measurement of fractionated plasma metaneph- rithms, positive biochemical tests were followed by abdominal rines alone) would detect 489 pheochromocytoma patients at computerized tomography and, if negative, metaiodobenzylgua- a cost of 56.6 million dollars, whereas B (24-h urinary mea- nidine scintigraphy. surements) would detect 457 pheochromocytoma patients for In each hypothetical algorithm, imaging would be indi- 39.5 million dollars, and C (combination of measurements of cated after positive biochemical testing as follows: algorithm fractionated plasma metanephrines and urines) would detect A, fractionated plasma metanephrine measurements above 478 patients for 28.6 million dollars. None of the screening the laboratory reference range; or algorithm B, abnormal strategies for pheochromocytoma described are affordable if measurements of 24-h urinary total metanephrines or cat- implemented on a routine basis in extremely low-risk pa- echolamines. In algorithm C, subjects with fractions of plasma tients. However, algorithm C may be the least costly, and at a metanephrine at or above 0.5 nmol/liter or normetanephrine reasonable level of sensitivity, for subjects in whom the sus- at or above 1.80 nmol/liter would undergo imaging, whereas picion of disease is moderate. (J Clin Endocrinol Metab 89: those with values between the reference range and these cut- 2859–2866, 2004)

ATECHOLAMINE-SECRETING TUMORS are rare ical screening strategies (and subsequent imaging) for de- C neoplasms of chromaffin cells (estimated incidence, tection of pheochromocytoma. 1.55–8 per million persons per year) that arise from the adrenal medulla (pheochromocytoma) or paraganglia (para- Subjects and Methods ganglioma) (1–5). Catecholamine-secreting tumors are some- Subjects who underwent biochemical tests times sought as part of an evaluation for secondary causes of We reviewed the medical records of 416 outpatients (including 47 hypertension, unexplained spells, or incidental adrenal patients with histologically confirmed pheochromocytoma or paragan- masses, or in patients with rare genetic predispositions to glioma) who had concurrent measurements of fractionated plasma pheochromocytoma. There is no standardized approach to metanephrines, 24-h urinary total metanephrines, and 24-h urinary catecholamines between January 1, 1999, and November 29, 2001, to esti- biochemical screening for catecholamine-secreting tumors mate the diagnostic efficacy of biochemical tests (updated published between or within institutions. series) (Fig. 1) (11). Indications for testing in patients without pheo- Although recent reports have suggested that measure- chromocytoma included hypertension in 148 patients, spells with or ments of fractionated plasma metanephrines may be a con- without sustained or paroxysmal hypertension in 126, adrenal mass(es) on an imaging study in 57, and high-risk group (including patients with venient biochemical test for pheochromocytoma, an optimal high-risk familial syndromes, pheochromocytomas, or paragangliomas strategy for dealing with mildly elevated (borderline) ele- cured surgically previously) in 38 subjects (Fig. 1). Of the 47 pheochro- vations of these measurements is needed (6–10). The eco- mocytoma patients, 30 had an adrenal pheochromocytoma, 17 had at nomic implications of different biochemical testing strategies least one extra-adrenal pheochromocytoma, 17 had malignant pheochromocytoma, and 6 had a genetic syndrome predisposing to pheo- and subsequent imaging of positive screens have not been chromocytoma (those with genetic syndrome screened before Novem- explored. Our aim was to explore, using a modeling tech- ber 27, 2001) (Fig. 1). After November 27, 2001, subjects with a known nique, the economic implications of three proposed biochem- genetic predisposition to pheochromocytoma were excluded from the study to prevent excessive representation of subjects with rare familial syndromes who are prone to being seen in quaternary care centers (36 Abbreviations: CI, Confidence interval; CT, computerized tomogra- subjects, including 4 subjects with pheochromocytoma excluded). An- phy; MIBG, metaiodobenzylguanidine. other 47 subjects were excluded because of an abnormal spectral curve, indicating drug interference in measurement of 24-h urinary total JCEM is published monthly by The Endocrine Society (http://www. metanephrines. endo-society.org), the foremost professional society serving the en- All catecholamine-producing tumors were histologically confirmed. docrine community. In terms of extra-adrenal paragangliomas, only catecholamine-secreting

2859 2860 J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening

FIG. 1. Description of subjects from whom diagnostic efficacy data of biochemical tests were obtained. “Syndromic pheochromocytoma” refers to patients that have a genetic disorder that in- creases their risk to harbor a catecholamine- secreting tumor (e.g. familial paraganglioma, neurofibromatosis type 1, von Hippel-Lindau disease, Carney triad, and multiple endocrine neoplasia type 2). “Hx pheochromocytoma” refers to patients that had prior resection of a catecholamine-secreting tumor. Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 paragangliomas were included. All subjects without pheochromocy- TABLE 1. Charges to the third-party payer (in U.S. dollars) toma were assigned a different clinical diagnosis by their treating phy- sician at the completion of their evaluation. The Institutional Review Variables Costs Board of the Mayo Foundation approved the study, and signed consent was verified for all patients whose medical records were reviewed. Venipuncture $ 20.50 There was no sponsor involvement, nor funding for the study. (added to plasma metanephrine cost) Fractionated plasma metanephrines $ 105.00 Biochemical assays (not including venipuncture) 24-h urinary total metanephrines $ 110.00 ($121.00)a a Liquid chromatography with electrochemical detection was used for 24-h urinary fractionated catecholamines $ 170.00 ($187.00) a measurement of fractionated plasma metanephrines (reported as meta- 24-h urinary creatinine $ 32.50 ($35.75) nephrine and normetanephrine fractions) and 24-h urinary cat- CT scan of abdomen $1460.00 echolamines (reported as norepinephrine, epinephrine, and dopamine (with and without iv contrast) 123 fractions), whereas urinary total metanephrines were measured by spec- [ I]MIBG scan $1875.00 trophotometry (12–15). All biochemical assays were performed at the (with and without SPECT) Mayo Medical Center. In the case of multiple measurements of the same SPECT, Single-proton emission CT. metabolite for the same patient, only the first concurrent measurement a Costs inflated by 10% for urinary measurements to adjust for of plasma and urinary analytes was included in the study. Fractionated need for repeat collections because of drug interference or incomplete plasma metanephrine measurements were performed via venipuncture collections. in the sitting posture, either fasting or nonfasting.

Interpretation of biochemical tests egy. The outcome of interest was the number of patients with pheochromocytoma expected to be detected by each strategy. The costs of For fractionated plasma metanephrines, the upper limits of the 95% false-positive biochemical tests were reflected only in the costs of sub- reference range established by Mayo Medical Laboratories were 0.5 sequent imaging and not in potential costs of needless surgery or its nmol/liter (98 pg/ml) for the metanephrine fraction and 0.9 nmol/liter possible complications. (165 pg/ml) for the normetanephrine fraction, and measurements at or A decision analysis model of screening of hypertensive patients solely above either of these levels were considered positive in algorithm A. A by measurement of fractionated plasma metanephrines was developed, urinary total metanephrine content at or above 6.6 ␮mol/24 h (Ն1.3 with all positive screens (defined by the metanephrine or normetaneph- mg/24 h) was considered positive (15). For urinary catecholamines, rine fractions being above the reference range), followed by imaging values approximately twice that of the upper limit of the 95% reference (algorithm A; see Fig. 3A). In algorithm B, only subjects with positive range were considered positive, specifically the following: 24-h urinary 24-h urinary total metanephrines or catecholamines would undergo content of norepinephrine at or above 1005 nmol (Ն170 ␮g), epinephrine further imaging (see Fig. 3B). In algorithm C, those patients with a at or above 191 nmol (Ն35 ␮g), or dopamine at or above 4571 nmol (Ն700 plasma metanephrine fraction at or above 0.5 nmol/liter (98 pg/ml) or ␮g) (11, 16). A positive 24-h urinary total metanephrine or catecholamine a plasma normetanephrine fraction at or above 1.80 nmol/liter (330 result was defined by positivity of either the urinary total metanephrines pg/ml) or those with indeterminate plasma measurements but positive or any catecholamine fraction in algorithms B or C. In algorithm C, urinary testing, would undergo imaging (see Fig. 3C). measurements of plasma metanephrine and normetanephrine within The imaging protocol for patients with positive biochemical screens the 95% reference range were considered negative. In contrast, a bio- in both strategies began with computerized tomography (CT) (with and chemical test was considered positive in algorithm C, if either the plasma without iv contrast) of the abdomen, and then if negative, [131I]- or normetanephrine fraction was greater than or equal to twice the upper [123I]metaiodobenzylguanidine (MIBG) scintigraphy (efficacy for 131I limit of normal (1.80 nmol/liter; 330 pg/ml) or the plasma metanephrine and costs for 123I shown). The sensitivity of CT imaging of the abdomen fraction was above 0.5 nmol/liter (98 pg/ml). In algorithm C, measure- was assumed to be 98% with a specificity of 70% (17). The sensitivity of ments of plasma normetanephrine or metanephrine fractions outside MIBG scintigraphy in detecting benign sporadic pheochromocytoma these definitions were considered indeterminate, and urinary testing was assumed to be 87.4% with a specificity of 98.9% (18). would be indicated. The cutoffs for positivity for 24-h urinary total The analysis was performed from a third-party payer perspective, metanephrine and fractionated catecholamine measurements were the with the term “costs” referring to charges to the third-party payer. All same for algorithm C as described above for algorithm B. New positivity costs were reported in 2002 U.S. dollars. Costs of biochemical testing cutoffs for measurements of fractionated plasma metanephrines in were obtained from the Mayo Medical Laboratories, and the costs of algorithm C have been established (see the supplemental data published venipuncture and imaging investigations were obtained from the Mayo on The Endocrine Society’s Journals Online web site at http://jcem. Clinic Rochester Business Office (Table 1). The cost of venipuncture was endojournals.org). included with plasma measurements. The cost of a 24-h urinary creatinine was included with urinary tests. The cost of urinary diagnostic Analyses and assumptions testing was artificially inflated by 10% to account for the need for repeat urinary measures in those with an incomplete collection or drug inter- A modeling technique was used to develop decision trees, incorpo- ference. Surgical costs were excluded. For the purpose of the decision rating the diagnostic efficacy data of biochemical tests. The horizon analysis model, the prevalence of pheochromocytoma in the hyperten- (endpoint) of the analyses was diagnosis or exclusion of pheochromo- sive population that would typically be screened was estimated at 0.5% cytoma, for hypothetical hypertensive patients subjected to each strat- (19). Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 2861

For sensitivities, specificities, and proportions pertaining to biochem- subjects with indeterminate plasma results, but positive 24-h ical testing algorithms, 95% confidence intervals (CI) were calculated urinary total metanephrines and catecholamines. Thus, us- using Wilson’s method (20). If combinations of biochemical tests were ing algorithm C, 91.5% (95% CI, 80.1–96.6%) of patients with performed in algorithms, the overall levels of sensitivity and specificity for combinations of tests were calculated (incorporating all biochemical pheochromocytoma would be expected to have a positive tests), using actual patient data of those with positive, negative, and fractionated plasma metanephrine measurement (43 of 47), indeterminate first-level biochemical tests. Sensitivity analyses were and 84.3% (95% CI, 80.2–87.6%) of patients without pheo- performed by determining the expense per case of pheochromocytoma chromocytoma would be expected to have a negative mea- detected as well as the expense per case of pheochromocytoma ruled out in patients without disease for variable levels of pretest probability of surement (311 of 369) (Fig. 2). Those with indeterminate disease. We also performed a sensitivity analysis, examining the total plasma results who would undergo confirmatory urinary cost of each algorithm, if all patients with positive or negative CT testing, including an estimated 6.4% of patients with pheo- imaging would undergo MIBG scintigraphy. We performed a third chromocytoma (3 of 47; 95% CI, 2.2–17.2%) and 11.7% of sensitivity analysis, examining the variable total cost of each algorithm, subjects without pheochromocytoma (43 of 369; 95% CI, 8.8–

using variable imaging costs (for CT and MIBG). Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 15.3%). The sensitivity of urinary measures in the pheochro- Results mocytoma patients with indeterminate plasma measures Estimates of diagnostic efficacy of biochemical tests would be estimated to be 66.7% (2 of 3; 95% CI, 20.8–93.9%), with a specificity of 100.0% (43 of 43; 95% CI, 91.8–100.0%). Based on the data collected from 416 outpatients (includ- The biochemical testing strategy in algorithm C was esti- ing 47 patients with histologically confirmed pheochromo- mated to have an overall sensitivity of 95.7% (45 of 47; 95% cytoma or paraganglioma), who underwent concurrent CI, 85.8–98.8%) with a specificity of 95.9% (354 of 369; 95% measurements of fractionated plasma metanephrines, CI, 93.4–97.5%), when both levels of testing were included 24-h urinary total metanephrines, and 24-h urinary cat- (Fig. 2). echolamines, we estimated the diagnostic efficacy of the three biochemical screening strategies. In algorithm A, mea- Cost implications of screening strategies surement of fractionated plasma metanephrines alone (using for pheochromocytoma the upper limit of the 95% reference ranges for the cutoffs for positivity) was estimated to have a sensitivity of 97.9% (46 of For the purpose of the cost analysis, in all three screening 47; 95% CI, 88.9–99.6%) and a specificity of 84.3% (311 of 369; algorithms, a 0.5% prevalence of pheochromocytoma was 95% CI, 80.2–87.6%) (Fig. 2). In algorithm B, measurement of assumed in a target hypertensive population (19), so 500 24-h urinary total metanephrines and catecholamines, was patients with pheochromocytoma would be expected to be in estimated to have a sensitivity of 91.5% (43 of 47; 95% CI, a sample of 100,000 hypertensive subjects. In algorithm A, 80.1–96.6%) and a specificity of 98.4% (363 of 369; 95% CI, fractionated plasma metanephrine measurements above the 96.5–99.3%) (Fig. 2). In algorithm C, subjects with positive reference range would be followed by imaging (Fig. 3A). If biochemical testing included subjects with positive fraction- 100,000 subjects with hypertension would be screened using ated plasma metanephrine measurements (plasma meta- algorithm A, and all those with positive biochemical screens nephrine fraction of Ն0.5 nmol/liter or a plasma normeta- imaged, 489 of 500 subjects with pheochromocytoma (overall nephrine fraction of Ն1.80 nmol/liter) as well as those sensitivity, 97.8%) would be expected to be detected, and

FIG. 2. Sensitivity and specificity of biochemical tests in screening algorithms for detection of pheochromocytoma. 2862 J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021

FIG. 3. Algorithms for screening for pheochromocytoma in 100,000 hypertensive subjects. A, Fractionated plasma metanephrine measurement followed by imaging in patients with either plasma measurement above the 95% reference range. B, Measurement of 24-h urinary total metanephrines and fractionated catecholamines followed by imaging in those with abnormal results. C, Fractionated plasma metanephrine measurement followed by 24-h urinary total metanephrines and catecholamine measurement in those with indeterminate plasma measurements. Imaging is performed in those with plasma metanephrine fractions over 0.5 nmol/liter or normetanephrine fractions over 1.80 nmol/liter or positive 24-h urinary total metanephrines or catecholamines. Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 2863 Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021

FIG.3.Continued

TABLE 2. Comparison of costs and effectiveness of algorithms for 500 with pheochromocytoma) would be screened using diagnosis of pheochromocytoma algorithm B (24-h urinary total metanephrines and cat-

Overall Overall Cost echolamines followed by imaging in those with positive val- Algorithm sensitivity (%)a specificity (%)a (millions)b ues), 457 subjects with pheochromocytoma (91.4%) would be Ac 97.8 95.2 56.6 expected to be detected, and 99,010 of 99,500 (99.5%) of Bd 91.4 99.5 39.4 subjects without pheochromocytoma would be expected to Ce 95.6 98.7 28.6 be reassured with a negative diagnosis (Fig. 3B). Further- a Overall sensitivity and specificity refer to algorithm A, B, or C, more, 1592 subjects without pheochromocytoma would un- in which positive biochemistry is followed by CT scanning and, if 123 dergo CT scanning of the abdomen, and 478 of these subjects negative, [ I]MIBG scanning. 123 131 b Cost (millions) refers to estimated cost of biochemical testing and would undergo [ I]- or [ I]MIBG scanning. The total cost imaging for 100,000 hypertensive patients screened using algorithm per 100,000 subjects screened would be estimated at 39.5 A, B, or C, assuming a prevalence of pheochromocytoma of 0.5%. million dollars (Table 2). c Algorithm A refers to traditional interpretation of fractionated If algorithm C (measurement of fractionated plasma meta- plasma metanephrine measurements, followed by imaging for positive test results (i.e. imaging if either the metanephrine or normeta- nephrines in all and 24-h urinary total metanephrines and nephrine fraction is above the reference range). catecholamines in those with indeterminate plasma values, d Algorithm B consists of imaging if 24-h urinary total metaneph- followed by imaging in all with positive biochemical tests) rines or catecholamines are abnormal. were followed in 100,000 subjects with hypertension, 478 of e Algorithm C consists of imaging in those with high fractionated plasma metanephrine measurements or those with indeterminate 500 subjects with pheochromocytoma would be expected to fractionated plasma metanephrine measurements and abnormal 24-h be detected (95.6%), and 98,245 of 99,500 subjects without urinary total metanephrines or catecholamines. pheochromocytoma would be reassured with a negative diagnosis (overall specificity of 98.7%) (Fig. 3C). Furthermore, 94,694 of 99,500 of subjects without pheochromocytoma 4079 subjects without pheochromocytoma would undergo would be reassured with a negative diagnosis (overall spec- CT scanning, and 1224 of these subjects would undergo 131 ificity, 95.2%). However, 15,621 subjects without pheochro- [ I]MIBG scintigraphy. The cost of algorithm C for 100,000 mocytoma would undergo CT scanning of the abdomen, and subjects screened would be 28.6 million dollars (Table 2). 10,935 of these subjects would undergo [123I]- or [131I]MIBG Sensitivity analyses were performed examining the ex- scanning. The total cost per 100,000 subjects screened would pense per pheochromocytoma case detected and the expense be estimated at 56.6 million dollars (Table 2). per patient with true negative test results in whom the di- If the same 100,000 subjects with hypertension (including agnosis was ruled out for varying levels of pretest probability 2864 J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening of pheochromocytoma (Fig. 4). Using the assumptions of our disease of 0.5% to approximately $1,800 per case detected at model, for algorithm A, the cost per pheochromocytoma a pretest probability of disease of 75%. For algorithm B, the detected would be expected to range from a high of approx- cost per pheochromocytoma detected would be expected to imately $115,700 per case detected at a pretest probability of range from a high of approximately $86,400 per case detected at a pretest probability of disease of 0.5% to approximately $2,000 per case detected at a pretest probability of disease of 75%. The least expensive algorithm, particularly at the lowest levels of pretest probability of disease, was algorithm C, for which the cost per pheochromocytoma detected would be expected to range from a high of approximately $59,800 per case detected at a pretest probability of disease of 0.5% to approximately $1,800 per case detected at a pretest probability of disease of 75%. Of note, the cost of testing for pheo- Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 chromocytoma detected of algorithms A and B approached that of C when the pretest level of suspicion of pheochromocytoma was over 5–10% (Fig. 4A). The cost per true negative patient with the diagnosis of pheochromocytoma ruled out by testing was generally lower across all levels of pretest probability of disease in algorithm C, compared with the other algorithms (Fig. 4B). We also examined the total cost of each algorithm, if the imaging strategy were modified such that all patients with either positive or negative CT findings were subjected to MIBG scintigraphy (an imaging approach used by some clinicians outside Mayo Clinic). The total cost was calculated for 100,000 hypothetical hypertensive patients, including 500 patients with pheochromocytoma. In such a situation, 16,111 patients would be subjected to MIBG scintigraphy in algorithm A, at a total cost of biochemistry and imaging of 66.3 million dollars. Using algorithm B, 2050 patients would be subjected to MIBG, at a total cost of biochemistry and imaging of 41.2 million dollars. Furthermore, using algorithm C, 4558 patients would be expected to undergo MIBG scanning, at a total cost of biochemistry and imaging of 31.8 million dollars. Algorithm C would thus still be the least costly using such an imaging algorithm. We also examined the total cost of imaging 100,000 hypothetical hypertensive patients, including 500 patients with pheochromocytoma, by varying the hypothetical charges for imaging (Table 3). We found that algorithm C was the least expensive of the three algorithms, when the costs of CT and MIBG were 50 or 75% of that charged at Mayo Clinic Roch- ester. Algorithms A and C were equivalent in total costs, when imaging charges were a quarter of that charged by Mayo Clinic; and algorithm A was slightly less expensive FIG. 4. Expenses for detecting and ruling out pheochromocytoma at than the others, when imaging charges were 10% (or less) of varying levels of pretest suspicion of disease for each of the algorithms that charged by Mayo Clinic. Thus, the relative overall cost described. A, Expense (in U.S. dollars) per case of pheochromocytoma of each algorithm was variable with imaging charges. detected using each algorithm at varying levels of pretest probability of disease. B, Expense per true negative case ruled out using each algorithm at varying levels of pretest probability of disease. In both Discussion A and B, the algorithms shown are described as follows: algorithm A, fractionated plasma metanephrine measurement followed by imaging We examined three screening strategies for detection of in patients with either plasma measurement above the 95% reference range; algorithm B, measurement of 24-h urinary total metaneph- pheochromocytoma (incorporating measurements of frac- rines and fractionated catecholamines followed by imaging in those tionated plasma metanephrines or urinary total metaneph- with abnormal results; algorithm C, fractionated plasma metanephrines and catecholamines or both) and found that none of the rine measurement followed by 24-h urinary total metanephrines and strategies was dominant (most effective and least costly). It catecholamine measurement in those with indeterminate plasma is not known how highly patients and clinicians value the measurements. Imaging is performed in those with plasma metanephrine fractions over 0.5 nmol/liter or normetanephrine fractions changes in sensitivity of detection of pheochromocytoma, over 1.80 nmol/liter or positive 24-h urinary total metanephrines or relative to costs. However, the least costly strategy for de- catecholamines. tection of sporadic pheochromocytoma in hypertensive sub- Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 2865

TABLE 3. Sensitivity analysis comparing costs of each algorithm it is unclear how many people with false-positive biochem- using variable hypothetical imaging charges (CI and MIBG istry and imaging would undergo needless surgery. How- scintigraphy) ever, the rate of surgery for false-positive biochemistry and Imaging charges Charges imaging would be expected to be highest in algorithm A Algorithm (% Mayo charges)a (millions)b (fractionated plasma metanephrine measurements), lowest Ac 10 17.0 in B (24-h urinary measurements), and intermediate in C Bd 10 34.9 (combination testing). Thus, had surgical costs been included e C 10 17.8 in a model, we would expect that A (fractionated plasma A 25 23.6 B 25 35.6 metanephrines) would appear to be even more costly than C 25 23.6 the other two strategies. Algorithm B (24-h urinary measure- A 50 34.6 ments) would be expected to have the least number of need- B 50 36.9 less surgeries in patients with false-positive biochemical test-

C 50 22.6 Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 A 75 45.6 ing and imaging. B 75 38.2 Others have suggested that screening exclusively by mea- C 75 25.6 surement of fractionated plasma metanephrines (at currently a Percentage of Mayo charges for CT and MIBG scintigraphy. accepted cutoffs) could result in cost savings because of less b Cost (millions) refers to estimated cost of biochemical testing and costs incurred in multiple biochemical tests (21). In contrast, imaging for 100,000 hypertensive patients screened using algorithm we have found that such a strategy (algorithm A) would be A, B, or C, assuming a prevalence of pheochromocytoma of 0.5%. expected to be the most costly because of an excessively high c Algorithm A refers to traditional interpretation of fractionated plasma metanephrine measurements, followed by imaging for posi- rate of expensive imaging procedures in subjects with mildly tive test results (i.e. imaging if either the metanephrine or normeta- elevated levels of plasma normetanephrine. It is important nephrine fraction is above the reference range). that imaging costs be incorporated in any future analyses of d Algorithm B consists of imaging if 24-h urinary total metaneph- cost effectiveness of biochemical tests for detection of rines or catecholamines are abnormal. pheochromocytoma. e Algorithm C consists of imaging in those with high fractionated plasma metanephrine measurements or those with indeterminate Our study is limited in external generalizability given that fractionated plasma metanephrine measurements and abnormal 24-h it has been performed in a tertiary care center, the costs to urinary total metanephrines or catecholamines. third-party payers may vary among institutions, and the preferences of clinicians and patients for biochemical tests jects, in whom the risk of pheochromocytoma is extremely and imaging may be variable. Moreover, spectrophotometric low, is algorithm C (measurement of fractionated plasma urinary metanephrine measurements have been replaced by metanephrines in all subjects and 24-h urinary total meta- HPLC assays in many laboratories, and the sensitivities and nephrines and catecholamines in those with indeterminate specificities of urinary measures in our study may not be plasma values followed by imaging in all subjects with pos- generalizable to that of newer assays. Furthermore, our study itive biochemical tests). Algorithm C is approximately half as is limited by the fact that the diagnostic efficacy of MIBG expensive as algorithm A and a quarter less expensive than scintigraphy and CT in the quoted settings were extrapolated algorithm B (24-h urinary measurements). Algorithm C was from the literature and not prospectively determined in pa- found to have an overall sensitivity of detection pheochro- tients subject to the presented biochemical testing algo- mocytoma of 95.6% and a specificity of 95.2% (including rithms. We also studied a relatively small sample size of biochemical testing and imaging). In contrast, algorithm A, patients, particularly a limited number of patients with pheo- a 95% reference range-based approach to interpretation of chromocytoma with indeterminate fractionated plasma fractionated plasma metanephrine measurements, is the metanephrine measurements. We also studied a limited most effective at detecting pheochromocytoma, with an es- number of patients with adrenal incidentaloma, and our timated overall sensitivity of 97.9%, but is also the most findings may not be directly generalizable to all such pa- costly. Nonetheless, none of the strategies are affordable for tients, particularly if a mass characteristic of pheochromo- widespread implementation as a screening measure in low- cytoma is seen on imaging studies. risk, hypertensive patients, given the rarity of pheochromo- It is important to consider the broader policy implications cytoma and the high prevalence of essential hypertension. of the costs of the algorithms presented for a population- Thus, it is important to identify a high-risk subgroup of based screening program. It has been previously suggested hypertensive patients in deciding to test for pheochromocy- that the diagnosis of pheochromocytoma should be consid- toma. Of note, with increasing suspicion of disease, the ex- ered in many Americans with hypertension (6). It is esti- pense of testing by measuring fractionated plasma meta- mated that there are currently 12 million Americans with nephrines alone, approaches that of combined plasma and known uncontrolled hypertension in the United States (22). urinary testing when the pretest probability of disease ex- If all of these patients were screened once using algorithm A ceeds approximately 5–10% (which may include patients (using the Mayo imaging algorithm), the total cost of diag- who have a suspicious adrenal mass on imaging or those nostic testing would be 6.8 billion dollars, which is more than with genetic predisposition to disease). In such situations, half the annual direct medical expenditure for hypertension fractionated plasma metanephrine measurements may be in the United States (10 billion dollars) (23). Even if these 12 indicated, given the potential superior sensitivity of these million Americans would be screened with the least costly measurements. strategy, algorithm C (with the Mayo imaging algorithm), the In our analysis, we excluded the costs of surgery, because total cost for diagnostic testing would be estimated at 3.4 2866 J Clin Endocrinol Metab, June 2004, 89(6):2859–2866 Sawka et al. • Cost Effectiveness of Pheochromocytoma Screening billion dollars (approximately one-third of the U.S. yearly 6. Lenders JWM, Pacak K, Walther MM, Linehan WM, Mannelli M, Friberg P, Keiser HR, Goldstein DS, Eisenhofer G 2002 Biochemical diagnosis of pheo- hypertension expenditure). chromocytoma: which is the best test? JAMA 287:1427–1434 Missing a pheochromocytoma may have devastating con- 7. Pacak K, Linehan WM, Eisenhofer G, McClellan MW, Goldstein DS 2001 sequences for the affected individual, yet the efficacy of a NIH Conference: recent advances in genetics, diagnosis, localization, and treatment of pheochromocytoma. Ann Intern Med 134:315–329 screening program must be balanced with its costs to society. 8. Lenders JWM, Keiser HR, Goldstein DS, Willemsen JJ, Friberg P, Jacobs MC, Further research should be directed toward determining Kloppenborg PW, Thien T, Eisenhofer G 1995 Plasma metanephrines in the characteristics of hypertensive patients who are best served diagnosis of pheochromocytoma. Ann Intern Med 123:101–109 9. Raber W, Raffesberg W, Bischoff M, Scheuba C, Niederle B, Gasic S, by screening for pheochromocytoma. Other mechanisms of Waldhausl W, Roden M 2000 Diagnostic efficacy of unconjugated plasma improving cost effectiveness of screening could include im- metanephrines for the detection of pheochromocytoma. Arch Intern Med proving efficacy of biochemical tests (or combinations of 160:2957–2963 10. Eisenhofer G, Lenders JWM, Linehan WM, Walther MM, Goldstein DS, tests), improving specificity of imaging techniques, or low- Keiser HR 1999 Plasma normetanephrine for detecting pheochromocytoma in ering costs of biochemical testing and imaging modalities. von Hippel-Lindau disease and multiple endocrine neoplasia 2. N Engl J Med Furthermore, if less costly biochemical tests, albeit with 340:1872–1879 Downloaded from https://academic.oup.com/jcem/article/89/6/2859/2870332 by guest on 24 September 2021 11. Sawka AM, Jaeschke R, Singh RJ, Young Jr WF 2003 A comparison of lower sensitivity, may be available, it is important to consider biochemical tests for pheochromocytoma: measurement of fractionated whether such tests may be preferred for screening of low-risk plasma metanephrines compared to the combination of 24-hour urinary meta- individuals. Although tremendous progress has been made nephrines and catecholamines. J Clin Endocrinol Metab 88:2553–2558 12. Lenders JWM, Eisenhofer G, Armando I, Keiser HR, Goldstein DS, Kopin over the years in detecting pheochromocytoma, widespread IJ 1993 Determination of metanephrines in plasma by liquid chromatography screening of hypertensive subjects is not currently afford- with electrochemical detection. Clin Chem 39:97–103 able, and attention should be focused toward optimal re- 13. Weincove C 1991 ACP Broadsheet No. 127: measurement of catecholamines and their metabolites in urine. J Clin Pathol 44:269–275 source use in detecting this rare condition. 14. Crout JR, Pisano JJ, Sjoerdsma A 1961 Urinary excretion of catecholamines and their metabolites in pheochromocytoma. Am Heart J 61:375–381 Acknowledgments 15. Pisano JJ 1960 A simple analysis for normetanephrine and metanephrine in the urine. Clin Chim Acta 5:406–414 16. Kudva YC, Sawka AM, Young Jr WF 2003 Clinical review 164: The laboratory Received July 1, 2003. Accepted February 18, 2004. diagnosis of adrenal pheochromocytoma: the Mayo Clinic experience. J Clin Address all correspondence and requests for reprints to: Dr. William Endocrinol Metab 88:4533–4539 F. Young, Jr., Division of Endocrinology, Metabolism, Nutrition, and 17. Bravo EL, Gifford Jr RW, Manger WM 1993 Adrenal medullary tumors: Internal Medicine, Mayo Clinic, 200 First Street SW, Rochester, Minne- pheochromocytoma. In: Mazzaferri EL, Samaan NA, eds. Endocrine tumors. sota 55905. Boston: Blackwell Science; 426–447 18. Shapiro B, Copp JE, Sisson JC, Eyre PL, Wallis J, Beierwaltes WH 1985 Iodine-131 metaiodobenzylguanidine for the locating of suspected pheochro- References mocytoma: experience in 400 cases. J Nucl Med 26:576–585 1. Stenstrom G, Svardsudd K 1986 Pheochromocytoma in Sweden 1958–1981. 19. Manger WM, Gifford Jr RW 1996 Clinical and experimental pheochromocy- An analysis of the National Cancer Registry Data. Acta Med Scand 220:225–232 toma. 2nd ed. Cambridge, MA: Blackwell Science; 2–5 2. Beard CM, Sheps SG, Kurland LT, Carney JA, Lie JT 1983 Occurrence of 20. Altman DG, Machin D, Bryant TN, Gardner MJ 2000 Statistics with confi- pheochromocytoma in Rochester, Minnesota, 1950 through 1979. Mayo Clin dence. 2nd ed. London: British Medical Journal; 105–110 Proc 58:802–804 21. Eisenhofer G, Walther M, Keiser HR, Lenders JWM, Friberg P, Pacak K 2000 3. Hartley L, Perry-Keene D 1985 Phaeochromocytoma in Queensland—1970– Plasma metanephrines: a novel and cost-effective test for pheochromocytoma. 83.AustNZJSurg 55:471–475 Braz J Med Biol Res 33:1157–1169 4. Anderson GS, Tofdahl DB, Lund JO, Strandgaard S, Nielsen PE 1988 The 22. Hyman DJ, Pavlik VN 2001 Characteristics of patients with uncontrolled incidence rate of phaeochromocytoma and Conn’s syndrome in Denmark, hypertension in the United States. N Engl J Med 345:479–486 1977–1981. J Hum Hypertens 2:187–189 23. Ramsey SD, Neil N, Sullivan S, Perfetto E 1999 An economic evaluation of 5. Fernandez-Calvet L, Garcia-Mayor RV 1994 Incidence of pheochromocytoma the JNC hypertension guidelines using data from a randomized controlled in South Galicia, Spain. J Intern Med 236:675–677 trial. J Am Board Fam Pract 12:105–114

JCEM is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the endocrine community.