HYPOGLYCEMIA and neuroglycopenic symptoms

0021-972X/00/$03.00/0 Vol. 85, No. 9 The Journal of Clinical Endocrinology & Metabolism Printed in U.S.A. Copyright 2000 by The Endocrine Society Forty-Eight-Hour Fast: The Diagnostic Test for Insulinoma BOAZ HIRSHBERG, ANDREA LIVI, D. L. BARTLETT, S. K. LIBUTTI, H. R. ALEXANDER, J. L. DOPPMAN, M. C. SKARULIS, AND P. GORDEN Division of Intramural Research, National Institute of Diabetes, Digestive and Kidney Diseases (B.H., A.L., M.C.S.); Surgical Metabolism Section, National Cancer Institute (D.L.B., S.K.L., H.R.A.); and Department of Diagnostic Radiology (J.L.D.), National Institutes of Health, Bethesda, Maryland 20892 ABSTRACT Insulinoma causes fasting hypoglycemia due to inappropriate insulin secretion. Its diagnosis is based on demonstrating Whipple s triad during a supervised 72-h fast. For 75 yr, the 72-h fast has been the cornerstone for the diagnosis; however, it has never been critically assessed using newer assays for insulin, C peptide, and proinsulin. Thus, the aim of the current study is to assess the need for a full 72-h fast for the diagnosis of insulinoma. Patients with suspected hypoglycemia with documented glucose concentrations below 45 mg/dl were admitted to the NIH. Data obtained during the supervised fast of patients with pathologically proven insulinoma over a 30-yr period (1970 2000) were reviewed. We identified 127 patients with insulinoma. The average age of patients was 42.7 15.9 yr, with a predominance of females (62%). 107 patients had a benign tumor, 20 had malignant insulinoma, HYPOGLYCEMIA and neuroglycopenic symptoms that are corrected by the administration of carbohydrate are the hallmarks of the diagnosis of insulinoma (1 8). Fasting, therefore, is the major maneuver used in the diagnosis of insulinoma and has two important purposes. The first is to document hypoglycemia and its relationship to the patient s symptoms, and the second is to demonstrate inappropriate insulin concentrations in the face of hypoglycemia. The diagnosis of insulinoma has centered on the 72-h fast that was introduced long before it was possible to measure insulin or insulin-related components (1, 8, 9). Clearly, it is desirable to modify the diagnostic test as described in textbooks and protocols, to be conducted over a less protracted period if that can be justified by clinical experience. The purpose of the present study was 3-fold: 1) to determine the optimal duration of the diagnostic fast in the light of other modern biochemical laboratory measurements such as plasma insulin; 2) to determine the diagnostic value of direct measurement plasma proinsulin; and 3) to compare our large patient population with other large patient populations that have used similar criteria. and 15 patients had multiple endocrine neoplasia type 1. The fast was terminated due to hypoglycemia in 44 patients (42.5%) by 12 h, 85 patients (66.9%) by 24 h, and 120 (94.5%) by 48 h. Seven patients fasted beyond 48 h despite subtle neuroglycopenic symptoms and glucose and insulin concentrations diagnostic of insulinoma. Immunoreactive proinsulin was elevated at the beginning of the fast in 90% of 42 patients. Proinsulin in noninsulinoma, in contrast to insulinoma, patients is usually suppressible; therefore, samples taken in the suppressed state have the greatest diagnostic value. We conclude that with the current available insulin and proinsulin assays, the diagnosis of insulinoma can be made within 48 h. Thus, the 48-h fast should replace the 72-h fast in textbooks and hospital protocols as the new diagnostic standard. (J Clin Endocrinol Metab 85: 3222 3226, 2000) Subjects and Methods Patients We have identified 127 subjects with a pathological diagnosis of insulinoma and reviewed data collected during the diagnostic fast. These patients are part of a cohort recruited to a protocol approved by the institutional review board of the clinical center of the NIH. An additional 22 patients were studied over the same period, but did not meet criteria for diagnosis of insulinoma and are used for comparison. This protocol was intended to study the diagnosis, treatment, and natural history of diseases that cause hypoglycemia. Patients with suggestive history of hypoglycemia as well as documented glucose levels below 45 mg/dl were admitted to the Warren Grant Magnuson Clinical Center at the NIH. All patients had a complete history and physical examination, routine blood examination, and urine chemistries. Patients were fasted for up to 72 h under close medical supervision. Blood glucose levels, insulin, C peptide, and proinsulin were collected every 4 6 h. Serum samples were also screened for insulin antibodies. In addition, at the beginning and end of the fast, blood samples were taken for sulfonylurea screen, cortisol, and GH. The test was continued until the patient developed hypoglycemia, defined as plasma glucose below 40 mg/dl (2.22 mmol/l) accompanied by neuroglycopenic symptoms, such as confusion or blurred vision, or for a maximum of 72 h. Patients with documented hypoglycemia accompanied by inappropriate insulin and C peptide levels and negative sulfonylurea screen underwent localization studies and surgery when appropriate. Received May 4, 2000. Revision received June 1, 2000. Accepted June 5, 2000. Address all correspondence and requests for reprints to: Dr. Boaz Hirshberg, Division of Intramural Research, National Institute of Diabetes, Digestive and Kidney Diseases, National Institutes of Health, Building 10, Room 85-235B, Bethesda, Maryland 20892. Assays The assay procedure for measurement of insulin, C peptide, and proinsulin has been previously described (10). Results of determination of percent proinsulin and preliminary results of direct immunoreactive proinsulin have been presented (10). 3222

DIAGNOSTIC TEST FOR INSULINOMA 3223 Results Characteristics of insulinoma patients Over a 30-yr period, from 1970 2000, 127 patients were diagnosed with insulinoma. Surgical pathology confirmed the diagnosis in each case. The patients average age during their first hospitalization at the NIH was 42.7 15.9 yr; the youngest patient was 13 yr, and the oldest 81 was yr. There were 79 (62%) females and 48 (38%) males. The average delay between the initial symptoms and final diagnosis was 3.6 5.2 yr. One hundred and seven patients (84%) had benign insulinoma, 20 (16%) had malignant insulinoma, and 15 (12%) patients had multiple endocrine neoplasia type 1 (MEN1). For the entire patient group, the average weight was 77.6 22.2 kg, and the height was 167.8 9.9 cm. The average body mass index (BMI) was 27.7 7.4, and 62% of the patients had a BMI greater than 25. Diagnostic testing in insulinoma patients One hundred and nineteen patients completed a supervised fast. An additional 7 patients were constantly experiencing hypoglycemic episodes during the first 4 h after a meal (Fig. 1). Fifty-four patients (42.5%) became hypoglycemic within the first 12 h, and 85 patients (66.9%) became hypoglycemic within the first 24 h of the fast. The fast was concluded within 48 h in 120 (94.5%) of the patients. Seven patients in whom the fast was continued beyond 48 h are shown in Table 1. After reviewing the clinical as well as the laboratory data obtained during the fast, it is apparent that FIG. 1. Diagnostic fast in 127 patients with insulinoma. On the vertical axis is shown the percentage of patients reaching a diagnostic end point for the fast, and on the horizontal axis is shown the duration of the fast. each of these 7 patients achieved glucose and insulin concentrations that were consistent with the diagnosis of insulinoma by 48 h. The fast was continued because in the judgment of the physician the patient did not exhibit neuroglycopenic symptoms. Four of these patients (including the 2 who completed 72 h) had MEN1. These 4 patients with MEN1 had initially been evaluated for features of the syndrome that were unrelated to hypoglycemia, and a fast was performed due to a high index of suspicion. There was no association between the BMI and the length of the fast or the baseline insulin and proinsulin levels (results not shown). Measurements of proinsulin and insulin in insulinoma patients We have previously reported in 98 patients that the plasma percent proinsulin component was equal to or higher than 25% of the total immunoreactive insulin level in 87% of patients (10). In the present study we obtained direct measurements of plasma proinsulin in 42 patients from samples taken from the beginning of the fast [preliminary results were reported for a subset of 20 of these patients (10)]. In 38 of 42 patients (90%), immunoreactive proinsulin was equal to or greater than 0.2 ng/ml (22 pmol/l), which is the upper level of normal for this test. (Fig. 2) Samples from 50 patients taken at the end of the fast were available for measurement. There was a complete overlap in values from the basal to the suppressed state, but in the suppressed state 41 of 50 values (82%) were equal to or greater than 0.2 ng/ml (22 pmol/l). In 28 patients paired basal and suppressed samples were available for analysis (Fig. 3A). In general, immunoreactive proinsulin is poorly suppressible during the fast, and only 3 patients suppressed to 0.2 ng/ml. The calculated sensitivity of proinsulin for the diagnosis of insulinoma in the basal state is 89%, and that in the suppressed state is 90%. Measurement of proinsulin in noninsulinoma patients In a previous study we presented the relationship of proinsulin-like activity in plasma to the direct method of measuring proinsulin in plasma (10). In the present study we expanded on these results. To better understand the specificity of direct measurements of proinsulin in insulinoma patients, we present data for 21 patients who were studied TABLE 1. Characteristics of the seven patients fasted more then 48 h Patient no. Duration of fast (h) Duration (h) Fast could have been concluded a Glucose Insulin MEN1 Proinsulin b (mg/dl) ( U/mL) 1 50 36 39 11 N 17% 2 59 46 39 N 66% 3 62 38 50 24 Y 0.42 ng/ml 4 63 39 39 18 N 50% 5 64 44 40 10 Y 37% 6 72 30 48 11 Y 39% 7 72 30 42 31 Y 0.27 ng/ml a The data of the fast tests have been reevaluated. We have identified the first point during the fast test where the patient appeared to be symptomatic (according to the nursing and physician reports) along with low glucose levels associated with inappropriate high insulin levels. Four of these patients (including the two that fasted for 72 h) had MEN1 syndrome. All of them were initially evaluated for conditions that were unrelated to hypoglycemia, and a fast test was performed due to high index of suspicion. b Six of seven patients are considered to have an elevated value. See Ref. 10 for distinction in plasma proinsulin-like component (expressed as a percentage) and direct proinsulin measurements (expressed as nanograms per ml).

3224 HIRSHBERG ET AL. JCE&M 2000 Vol. 85 No. 9 FIG. 2. Direct immunoreactive proinsulin measurements in patients with and without insulinoma. Basal refers to the beginning of the fast, and suppressed refers to the end of the fast. Note that samples were available in 42 insulinoma patients (basal) and 50 suppressed patients. Twenty-eight of these samples have been previously reported and correlated with measurements of plasma proinsulin-like component (10). In the noninsulinoma patients, there were 16 samples obtained in the basal state, and 21 in the suppressed state. using the same protocol as the insulinoma patients, but did not meet the diagnostic criteria and hence are termed noninsulinoma patients. Sixteen patients had proinsulin measurements in the basal state, and 50% had values below 0.2 ng/ml. In the suppressed state, 95% of the patients had values below 0.2 ng/ml (Fig. 2). The calculated specificity is thus 50% in the basal and 95% in the suppressed state. In 16 patients paired basal and suppressed samples were available for measurement (Fig. 3B). All 16 subjects suppressed their proinsulin values, and all but 1 suppressed below 0.2 ng/ml (Figs. 2 and 3B). In 1 additional patient (not shown in Fig. 2) with MEN-1 whose blood glucose value was 33 mg/dl at 29 h of fasting, the basal proinsulin was 0.9 ng/ml and the suppressed value was 0.49 ng/ml, whereas the suppressed insulin value was 0.4 U/mL. We suspect that this patient may have an insulinoma and represents the rare group of patients who suppress insulin during the fast to a much greater extent than proinsulin. Discussion In the present study we verify that a 48-h fast provides sufficient information for the diagnosis of insulinoma. Thus, FIG. 3. Direct immunoreactive proinsulin measurements in paired samples taken at the beginning and the end of the diagnostic fast in insulinoma patients (A) and in noninsulinoma patients (B). Note the change in scale from A to B. by 48 h in all patients studied, hypoglycemia was observed in the presence of inappropriately elevated insulin concentrations. Approximately 95% of our patients had their fast terminated by 48 h, and the other 5% (seven patients) could have had their fast terminated on or before 48 h because they met the required glucose and insulin criteria. In these seven patients, the delay in the termination of the fast was due to inconsistency in the evaluation of neuroglycopenia. Careful review of the nursing as well as the physician notes revealed clear, although subtle, neuroglycopenia before 48 h, symptoms that corresponded with low glucose levels and inappropriately high insulin concentrations. Furthermore, in six of the seven patients with a fast longer than 48 h, the proinsulin levels were elevated, adding further supporting evidence for the diagnosis of insulinoma. Our data are in good agreement with studies of insulinoma patients from the Mayo Clinic and the Cleveland Clinic (6, 11). In a series of 51 patients reported by Service et al. (6) in

DIAGNOSTIC TEST FOR INSULINOMA 3225 1976, 92% developed hypoglycemia by 48 h. A previous series from the Mayo Clinic published in 1960 (2), reported that 59 of 79 patients (74%) developed hypoglycemia within 24 h of induction of the fast, and in only 2 patients was it necessary to continue the fast longer then 48 h. Both series used the classical Whipple s triad as the criteria for a positive fast, as insulin, C peptide, and proinsulin measurements were not available. Dizon et al. (11) reported a series of 59 patients seen at the Cleveland Clinic Foundation. Of the 49 patients who underwent a fast, only 1 patient did not achieve the glucose/insulin levels required by their protocol within first 48 h. Service (9) reported that several patients have been unnecessarily fasted for the full 72-h period, as subtle signs of neuroglycopenia had been overlooked. This is in agreement with our analysis of the seven patients in whom the fast test was extended beyond 48 h. A careful quantitative study has demonstrated a diminished responsiveness in insulinoma patients that is reversible (12). Proinsulin has been shown to be of diagnostic value in the diagnosis of insulinoma (10, 13, 14). We previously reported that 87% of our insulinoma patients had a plasma proinsulin component equal to or greater than 25% of the total immunoreactive insulin (10). In the present study we extend the number of patients previously reported in whom direct measurement of proinsulin is available. Ninety percent (38 of 42) had a value equal to or greater than the cut-off point of 0.2 ng/ml (22 pmol/l) at the start of the fast, and 82% (41 of 50) had values that met this criteria at the end of the fast. Our proinsulin measurements are also in clear agreement with the results reported from the Mayo Clinic (13). Direct measurement of proinsulin is now generally available and will replace the more laborious measurement of percent proinsulin component (for further discussion of the column method of measuring proinsulin-like component and direct measurement of proinsulin, see Ref. 10). Inappropriately elevated or nonsuppressed insulin values associated with hypoglycemia are the cornerstone of diagnosis of insulinoma. We now show that proinsulin values are poorly suppressible in these patients, whereas in noninsulinoma patients both insulin and proinsulin are readily suppressible. In fact, we have encountered 3 patients in whom insulin values, but not proinsulin values, are suppressed during the diagnostic fast. It is useful to have both a basal and a suppressed sample for proinsulin, as this markedly increases the specificity of the diagnosis of insulinoma. We believe that the 48-h fast and the measurement of proinsulin are sufficient for the diagnosis of insulinoma and that more prolonged fasting or other stimulation or suppression tests are unnecessary. The fast must be carried out in a supervised setting, and insulin should be determined every 6 h (1, 3, 9). This is because there may be some fluctuation in values during the fast, and the overall pattern is useful in the final analysis. Normal subjects may lower their glucose levels below 45 mg/dl in 48 h, but the corresponding insulin levels should be very low or undetectable (15). There are several other caveats that are important. Rarely, patients with insulinoma may suppress their insulin levels as their glucose levels fall. In fact, only three patients had insulin values below 10 U/mL at the end of their fast but still had elevated proinsulin concentrations (16, 17; see Ref. 10 for further discussion of insulin RIAs). In addition, patients with MEN1 may be confusing. This is due in part to the fact that these patients may present other manifestations of MEN1 and be asymptomatic with respect to glucose levels. These patients may be fasted due to a high index of suspicion, including random low glucose measurements, vague symptoms, or even direct measurements of high insulin levels during venous sampling, but still, in a way that is not clear, have minimal or any symptoms. Thus, the interpretation of fasting results in MEN1 patients without clear symptoms may pose a challenge. Furthermore, during surgery multiple tumors that secrete insulin or other hormones are the rule, but are rare in sporadic insulinomas (18). In our experience, that spans more then 30 yr, we have not convincingly seen islet hyperplasia or nesidioblastosis that could be etiologically related to the patients hypoglycemia. Thus, we only carry out pancreatic resection or enucleation when an insulinoma is clearly localized during surgery. We believe that former recommendations of distal blind pancreatectomy when localization is not possible are not appropriate. This is again based on the excellent sensitivity of preoperative calcium stimulation test and intraoperative ultrasound in localization of the tumor (19 23). It should be emphasized that invasive localization studies should be carried out only in patients in whom the biochemical diagnosis of insulinoma is confirmed. Recently, Service et al. (24) have described five patients with severe postprandial hypoglycemia in whom partial pancreatic resection guided by calcium stimulation was carried out. These patients did not have insulinoma, and the postprandial nature of the hypoglycemia differentiates these patients from patients with fasting hypoglycemia. We have not, to date, recognized patients with this syndrome. In our experience the most difficult differential diagnosis for insulinoma is factitious hypoglycemia, especially when oral hypoglycemic medications are used. In these cases levels of insulin and C peptide will be high as in insulinoma. This is farther complicated by the fact that current sulfonylurea assays fail to detect the new generation nonsulfonylurea hypoglycemic agents (25, 26). The 72-h fast became central to the diagnosis of insulinoma well before insulin and proinsulin determinations became available. Now that insulin and proinsulin measurements are widely available, all of the necessary information from a fast can be derived in the first 48 h. Thus, the 48-h fast should replace the 72-h fast in textbooks and hospital protocols and become a new standard. Certainly this will reduce cost, but it is imperative for both private insurers and clinical research centers to recognize that the fast must be carried out under supervised conditions. References 1. Service FJ. 1995 Hypoglycemic disorders. N Engl J Med. 332:1144 1152. 2. Breidahl HD, Priestley JT, Rynearson EH. 1955 Hyperinsulinism: surgical aspects, and results. Ann Surg. 142:698 706. 3. Service FJ. 1999 Classification of hypoglycemic disorders. Endocrinol Metab Clin North Am. 28:501 517. 4. Scholz DA, ReMine WH, Priestley JT. 1960 Hyperinsulinism: review of 95 cases of functioning pancreatic islet cell tumors. Proc Staff Meet Mayo Clinic. 35:545 550.

3226 HIRSHBERG ET AL. JCE&M 2000 Vol. 85 No. 9 5. Skillern PG, Rynearson EH. 1953 Endocrine review: medical aspects of hypoglycemia. J Clin Endocrinol Metab. 13:587 603. 6. Service FJ, McMahon MM, O Brien PC, Ballard DJ. 1991 Functioning insulinoma incidence, recurrence, and long-term survival of patients: a 60-year study. Mayo Clin Proc. 66:711 719. 7. Service FJ, Dale AJ, Elveback LR, Jiang NS. 1976 Insulinoma: clinical, and diagnostic features of 60 consecutive cases. Mayo Clin Proc. 51:417 429. 8. Comi RJ, Gorden P. 1995 Hypoglycemic disorders in the adult. In: Becker KL, ed. Principles and practice of endocrinology and metabolism. Philadelphia: Lippincott; 1342 1351. 9. Service FJ. 1999 Diagnostic approach to adults with hypoglycemic disorders. Endocrinol Metab Clin North Am. 28:519 532. 10. Gorden P, Skarulis MC, Roach P, et al. 1995 Plasma proinsulin-like component in insulinoma: a 25-year experience. J Clin Endocrinol Metab. 80:2884 2887. 11. Dizon AM, Kowalyk S, Hoogwerf BJ. 1999 Neuroglycopenic and other symptoms in patients with insulinomas. Am J Med. 106:307 310. 12. Mitrakou A, Fanelli C, Veneman T, Perriello G, Calderone S, Platanisiotis D. 1993 Reversibility of unawareness of hypoglycemia in patients with insulinomas. N Engl J Med. 329:834 839. 13. Kao PC, Taylor RL, Service FJ. 1994 Proinsulin by immunochemiluminometric assay for the diagnosis of insulinoma. J Clin Endocrinol Metab. 78:1048 1051. 14. Gorden P, Sherman B, Roth J. 1971 Proinsulin-like component of circulating insulin in the basal state, and in patients and hamsters with islet cell tumors. J Clin Invest. 50:2113 2122. 15. Cryer PE. 1999 Symptoms of hypoglycemia, thresholds for their occurrence, and hypoglycemia unawareness. Endocrinol Metab Clin North Am. 28:495 500. 16. Dons RF, Hodge J, Ginsberg BH, et al. 1985 Anomalous glucose and insulin responses in patients with insulinoma. Caveats for diagnosis. Arch Intern Med. 145:1861 1863. 17. Rayfield EJ, Pulini M, Golub A, Rubenstein AH, Horwitz DL. 1976 Nonautonomous function of a pancreatic insulinoma. J Clin Endocrinol Metab. 43:1307 1311. 18. Marx S, Spiegel AM, Skarulis MC, Doppman JL, Collins FS, Liotta LA. 1998 Multiple endocrine neoplasia type 1: clinical and genetic topics. Ann Intern Med. 129:484 94. 19. Grant CS. 1999 Surgical aspects of hyperinsulinemic hypoglycemia. Endocrinol Metab Clin North Am. 28:533 54. 20. Grant CS, van Heerden J, Charboneau JW, James EM, Reading CC. 1988 Insulinoma: the value of intraoperative ultrasonography. Arch Surg. 123:843 848. 21. Doherty GM, Doppman JL, Shawker TH, et al. 1991 Results of a prospective strategy to diagnose, localize, and resect insulinomas. Surgery. 110:989 996. 22. Doppman JL, Chang R, Fraker DL, et al. 1995 Localization of insulinomas to regions of the pancreas by intra-arterial stimulation with calcium. Ann Intern Med. 123:269 273. 23. Brown CK, Bartlett DL, Doppman JL, et al. 1997 Intraarterial calcium stimulation and intraoperative ultrasonography in the localization and resection of insulinomas. Surgery. 122:1189 1193. 24. Service FJ, Natt N, Thompson GB, et al. 1999 Noninsulinoma pancreatogenous hypoglycemia: a novel syndrome of hyperinsulinemic hypoglycemia in adults independent of mutations in Kir6.2 and SUR1 genes. J Clin Endocrinol Metab. 84:1582 1589. 25. Grunberger G, Weiner JL, Silverman R, Taylor S, Gorden P. 1988 Factitious hypoglycemia due to surreptitious administration of insulin. Diagnosis, treatment, and long-term follow-up. Ann Intern Med. 108:252 257. 26. Gordon MR, Flockhart D, Zawadzki JK, Taylor T, Ramey JN, Eastman RC. 1988 Hypoglycemia due to inadvertent dispensing of chlorpropamide. Am J Med. 85:271 272.