Insulinoma: how reliable is the biochemical evidence? NK Chammas 1, JD Teale 2 and JD Quin 1 Original Article Abstract Addresses 1 The Royal Sussex County Hospital Eastern Road, Brighton, UK 2 SAS Peptide Hormone Unit, Clinical Laboratory, Royal Surrey County Hospital, Guildford GU2 7XX, UK Correspondence Dr JD Teale E-mail: dteale@royalsurrey.nhs.uk Background and methods We report a case of insulinoma in which the diagnosis was very challenging as some of the biochemical data were consistently equivocal. In order to assess the relative reliability of the analytical tests, retrospective biochemical data on 45 other cases of histologically confirmed insulinoma were evaluated, enabling the most secure diagnostic process to be identified. Results The data showed that insulin concentrations alone, although measurable, were equivocal in 17% of cases. The addition of C-peptide values clarified the diagnosis in about 50% of the borderline cases, whilst ketone (b-hydroxybutyrate) concentrations were low during the prevailing hypoglycaemia in all cases. Conclusion The combination of these three tests is suggested as the most effective method for the biochemical diagnosis of hypoglycaemia due to insulinoma. Introduction The clinical diagnosis of insulinoma is based on the presence of Whipple s triad, 1 consisting of (1) the presence of symptoms during fasting which are (2) associated with hypoglycaemia and (3) relieved by glucose. The biochemical diagnosis of insulinoma has previously relied on the demonstration of unequivocally measurable serum insulin concentrations during prevailing hypoglycaemia. The concurrent measurement of C-peptide together with the quantitation of ketones has provided an additional indicator of insulin secretion. 2 The reliability of the biochemical data is especially important because of the relative ine ectiveness of non-invasive tumour localization techniques (CT, MRI) or specialized imaging methods, and the lack of availability of expertise. 3 Here we report a case where diagnosis was challenging and examine the reliability of biochemical tests in the diagnosis of insulinoma. Case report A 53-year-old housewife presented with a prolonged clinical history of episodes of confusion, disorientation and irritability, relieved rapidly with carbohydrate-containing drinks. Her husband, who had type 2 diabetes mellitus, recorded her glucose concentration during one of these episodes as 1.7 mmol/l. In the past she had had various investigations for atypical epilepsy, including a normal CT scan of the brain. She also su ered from migraine, which responded to simple analgesia. Clinical examination was entirely normal. Although she had gained one stone in weight over the past year her body mass index was 23.4. An initial supervised fast for 24 h failed to induce hypoglycaemia and she remained asymptomatic during her hospital stay and was discharged home. Her symptoms recurred following discharge and on repeat formal fasting her glucose dropped to1.9 mmol/ L with simultaneous insulin concentrations of 26 pmol/ L and C-peptide of 134 pmol/ L. A raised insulin-like growth factor 1 (IGF-1) concentration of 40.5 nmol/ L (age-related reference range: 7.5-30.0), an adequate 9 a.m. cortisol of 502 nmol/ L and a growth hormone concentration of 8 miu/l did not suggest pituitary insu ciency. The b-hydroxybutyrate (BHB) concentration was 372 mmol/l. On four subsequent fasts over a period of about 2 months, similar results for glucose, insulin, C-peptide and BHB were obtained, with the exception of IGF-1 which was normal on one occasion but otherwise near the upper limit of the range.whilst of interest, the IGF-1 data are not of signi cance in isolation and, together with the growth hormone result, were not 2003 The Association of Clinical Biochemists 689
690 Chammas et al. thought typical of an acromegalic state. During the hypoglycaemic episodes increases in blood pressure were noted. A 24-h urine vanillylmandelic acid excretion was normal, thereby excluding phaeochromocytoma. A urine drug screen for opiates, benzodiazepines, cannabis, cocaine, amphetamine and sulphonylurea was negative. During her hospital stay she was commenced on diazoxide and there was no signi cant hypoglycaemic attacks subsequently, despite troublesome peripheral oedema. Autoimmune hypoglycaemia was not considered to be a possible option since in these cases severe hypoglycaemia occurs only post-prandially. 4 In such cases, total serum insulin concentrations are high with disproportionately low C-peptide. The general consensus at this stage therefore suggested an insulinoma but a computed tomography (CT) scan and an ultrasound scan of abdomen and pelvis failed to identify any de nite lesion. Magnetic resonance imaging (MRI) of the abdomen and pituitary gland was also normal, and an octreotide scan was not indicative of insulinoma. She was therefore referred to a specialist centre. Endoscopic ultrasound scanning revealed a pancreatic lesion which was thought to be consistent with an insulinoma. She underwent pancreatoduodenectomy and cholecystectomy and a tumour, measuring 0.25 cm61.0 cm, was identi ed in the mid-pancreas. The tumour was palpable only after the head of the pancreas was mobilized. The anatomical position of the insulinoma directly below the division of the gastroduodenal artery made it di cult to enucleate and the decision was made to excise the tumour as this was felt to be safer than enucleation. Histological examination of the tumour (see Fig. 1) suggested that it may be malignant. Immunostaining showed positive staining for chromogranin, P- glycoprotein 9.5, neurone-speci c enolase and synaptophysin, all of which are neuroendocrine tumour markers found in insulinomas. The patient made a satisfactory initial post-operative recovery and on follow-up appointment she reported steady progress with no further episodes of hypo- or hyperglycaemia. Methods Case investigation Serum insulin was measured by a commercial enzyme-linked immunosorbent assay (ELISA) system (Mercodia Isoinsulin, available from Diagenics Ltd, Milton Keynes, UK). This method was selected for its reactivity with proinsulin (60% relative to insulin), since elevated serum proinsulin concentrations are often seen in insulinomas. The method showed a Figure 1. Clusters of neoplastic islet cells with eosinophilic cytoplasm and mildly pleomorphic enlarged enlarged hypochromatic nuclei with areas of brosis visible in between the nests and clusters of cells. Immunostaining showed strong positive staining for chromogranin and moderate staining for P- glycoprotein 9.5, synaptophysin and neurone-speci c enolase, all of which highlight neurosecretory granules in neuroendocrine tumours such as insulinomas. between-assay imprecision of 13.3% at 33 pmol/ L, 8.2% at 61pmol/L and 8.2% at 120 pmol/l. A precision- dose pro le based on these data produced an assay sensitivity of 4 pmol/l and an assay detection limit of 10 pmol/ L (2.56sensitivity). C-peptide was measured by a commercial ELISA (Mercodia) which exhibited negligible reactivity with insulin and proinsulin. The between-assay imprecision was 10.3% at 300 pmol/ L, 6.2% at 520 pmol/l and 4.9% at 735 pmol/l, with a sensitivity of 40 pmol/ L and an assigned detection limit of 100 pmol/l. Serum b-hydroxybutyrate (BHB) was measured by an enzymatic spectrophotometric method. 5 This assay had a between-assay imprecision of 13.9% at 720 mmol/l and 8.2% at 4270 mmol/l. Retrospective data Data from 45 cases of insulinoma (histologically con rmed) referred over a 16-year period were analysed. None of the patients had renal impairment which caused an increase in C-peptide. In the majority of these cases insulin and C-peptide were measured by in-house radioimmunoassays (RIAs). The assay characteristics were similar to the ELISA methods described above with the exception that the detection limit of the insulin RIA was 25 pmol/ L. Glucose concentration was measured on all serum/ plasma samples to con rm hypoglycaemia. Results The biochemical results obtained in our patient are shown in Table1. Data from the rst sample in our case
Biochemical diagnosis of insulinoma 691 Table 1. Summary of biochemical results during the course of diagnosis Date Glucose Insulin C-peptide BHB IGF-1 (nmol/l) 19/7/99 1.9 26 134 372 40.5 13/8/99 1.8 25 225 268 19.5 3/9/99 2.3 16 180 187 31.6 7/9/99 2.2 32 260 240 27.9 9/9/99 2.2 37 207 277 38.9 Age-related reference range for insulin-like growth factor 1 (IGF-1): 7.5-30.0 nmol/l. BHB, b-hydroxybutyrate. (number 3) and from 45 other cases are shown in Table 2 in increasing order of serum insulin. Insulin and C-peptide were measured by RIA methods in 37 cases and by ELISA methods in nine. The median age was 48 years (range 22-84) and 59% were female, similar to the data from the Mayo clinic. 6 In the rst eight cases, serum insulin concentrations were in the zone of diagnostic uncertainty (up to 50 pmol/ L). This concentration was selected when the less sensitive RIA method was in use and was retained when the ELISA system came into operation. For C-peptide, a concentration up to 300 pmol/l was considered of borderline signi cance. In all cases, BHB concentrations were less than 600 mmol/ L during the hypoglycaemia. Discussion Most insulinomas are small, 90% being less than 2 cm in diameter and 50% less than 1.3 cm; they are therefore di cult to detect. 7 A wide range of imaging techniques, including ultrasound, 8 CT 8 and MRI, 9 has been used to localiz e insulinoma pre-operatively, but so far no single localization modality has been able to provide a diagnosis with a high level of speci city and were unsuccessful in the present case. Localization by highly selective visceral angiography, 8 transhepatic portal vein sampling or venous sampling after stimulation 10 has achieved high rates of success in experienced centres. Somatostatin receptor scintigraphy is of limited value because of the absence, in a substantial proportion of patients, of receptors which bind octreotide. 8 Intra-operative ultrasonography is advocated as the most e ective method for localizing an insulinoma at surgery. 11 More recently, endoscopic ultrasound has been reported to be extremely e ective in localiz ing small pancreatic endocrine tumours, with detection rates of 80-85% and a speci city of 95%, despite the fact that most insulinomas are found in the body or tail of the pancreas. 11 This technique proved successful in locating the insulinoma in the present case. The most characteristic feature of insulinomas is fasting hypoglycaemia, together with inappropriate serum insulin and C-peptide concentrations. 12,13 The de nition of hypoglycaemia is generally accepted as a blood glucose concentration below 2.2 mmol/ L in patients up to 60 years of age and 2.5 mmol/ L in older patients, although values below 3.0 mmol/ L warrant further investigation. The recommended procedure for safe and thorough evaluation of hypoglycaemia is an inpatient fast, of up to 72 h if necessary, and on more than one occasion, although most patients, particularly those with a longer symptomatic history, will become hypoglycaemic within 24 h. The hypoglycaemia should be established by laboratory measurement and reliance should not be placed solely on glucose meter measurements. Neither should hypoglycaemia be diagnosed solely on the appearance of symptoms that are relieved by food ingestion. Serum insulin concentrations during a fast must be interpreted in the light of simultaneous glucose values and are of no value if the latter are normal or elevated. Some laboratories use fasting reference ranges for insulin and C-peptide and insulin:glucose ratios for the diagnosis of insulinoma. It is evident from the insulin data in Table 2 that some of the serum insulin concentrations could be dismissed as `normal on the basis of such guidelines. The biochemical diagnosis of endogenous hyperinsulinaemia appears increasingly to be fraught with uncertainties associated with analytical and physiological variation, particularly in the measurement of insulin, serum levels of which are susceptible to rapid uctuation. The short half-life of insulin (4-6 min) is caused both by the rapidly pulsatile nature of its secretion (by normal pancreas and tumour) and its rst-pass clearance through the liver. The wide uctuations in peripheral blood concentrations of insulin often encountered in cases of insulinoma can confuse a de nitive diagnosis when based on a single insulin measurement. The simultaneous measurement of C-peptide has proved to be an indispensable adjunct to insulin measurement in cases of accidental or deliberate injection of exogenous insulin. 2 It can also make an important contribution in the diagnosis of endogenous hyperinsulinaemia because of its relatively longer half-life (30-40 min).
692 Chammas et al. Table 2. Data from 46 cases of histologically con rmed insulinoma investigated using insulin and C-peptide measurement by radioimmunoassay or *enzyme-linked immunosorbent assay Case no. Age (years) Sex Glucose Insulin C-peptide BHB 1* 28 M 2.4 15 788 75 2 43 M 0.9 25 413 176 3* 53 F 1.9 26 134 372 4 60 F 2.3 28 474 402 5 44 F 1.9 30 158 64 6 45 M 1.7 31 255 90 7 74 F 1.5 38 522 125 8* 78 M 2.5 45 252 107 9 32 M 2.2 52 713 30 10 60 F 2.4 55 413 123 11 68 M 2.0 58 682 110 12 46 F 1.9 58 802 372 13 47 F 1.4 62 678 104 14 48 F 1.2 70 553 520 15 47 F 1.6 82 943 336 16 27 F 2.0 85 660 59 17 82 F 1.7 98 896 280 18* 65 F 1.6 100 1083 20 19 22 F 2.3 107 821 520 20* 57 F 2.2 116 402 58 21 75 F 1.7 120 884 520 22 38 F 0.9 123 868 580 23 45 M 2.0 130 1256 99 24 68 F 1.9 155 641 190 25 40 F 2.1 156 415 158 26 52 F 1.6 170 590 520 27* 67 F 1.0 187 806 520 28 42 M 2.1 218 652 65 29 39 M 2.2 225 1426 50 30 53 M 1.4 264 1587 122 31 25 F 1.9 301 1426 20 32 27 M 1.7 320 1052 313 33 63 F 2.3 344 1749 520 34 42 M 2.2 378 1077 637 35 56 M 1.9 400 1957 520 36 58 F 1.3 426 1060 271 37* 55 F 2.7 430 1618 102 38 50 M 1.6 443 1999 520 39 79 M 1.7 455 807 520 40 49 M 2.7 546 1023 520 41* 42 F 2.5 614 1818 520 42 68 M 2.4 666 1315 347 43 56 M 1.5 831 3200 80 44* 51 F 1.8 907 2381 56 45 40 M 2.5 944 1588 520 46 84 M 0.9 2054 4595 58 BHB, b-hydroxybutyrate. Wide targets for diagnostic reliability have been set for both insulin (up to 50 pmol/l) and C-peptide (up to 300 pmol/ L). Modern automated methods involve necessarily short incubation times and as a consequence are susceptible to errors in the diagnostically critical range of low concentrations. The data in Table 2 were obtained using RIA and ELISA methods operating at their limits of sensitivity. It is possible that some cases of insulinoma with serum concentrations below these limits may have been missed. The application of ketone (BHB) measurement to the diagnosis of hypoglycaemia has proved a useful screening test in the past. 5 Presence of ketosis during fasting indicates that hyperinsulinaemia is unlikely and further investigations are unnecessary. In cases with borderline insulin and C-peptide results BHB
Biochemical diagnosis of insulinoma 693 measurement provided further biochemical evidence. In all insulinoma cases, serum BHB concentrations were either not elevated or, when borderline, were associated with unequivocally inappropriate insulin and C-peptide concentrations (see cases 22 and 34 in Table 2). Proinsulin measurement can be useful but only where the immunoreactive insulin value (insulin+ 60% of the proinsulin) is su ciently high. Insulin concentrations such as those encountered in the present case would not provide reliable proinsulin data. Interestingly in three cases of proinsulinoma, unequivocally inappropriate concentrations of immunoreactive insulin (using cross-reacting methods) and proinsulin (using speci c methods) were accompanied by signi cant concentrations of C-peptide (measured by speci c methods) (unpublished observations). The biochemical diagnosis of insulinoma in the present case was therefore based on consistently inappropriate serum insulin and C-peptide with low BHB concentrations. Importantly, the possible presence of sulphonylurea, which can produce a biochemical pro le similar to that in insulinoma cases, 14 was excluded. Serum IGF-1 concentrations close to the upper limit of the relevant reference range can also provide additional diagnostic support, as shown previously in many, but not all, cases of insulinoma. 15 The application of all these procedures to the present case resulted in the correct diagnosis. This report highlights the degree of diagnostic di culty associated with the biochemical diagnosis of insulinoma. We conclude that careful optimization of immunoassays for insulin and C-peptide is an absolute prerequisite for this application and that measurement of ketones (BHB) is a valuable adjunct to the diagnostic process. Acknowledgement We thank Professor GWH Stamp from Imperial College School of Medicine, London, for supplying the histopathology slide. References 1 Whipple AO. The surgical therapy of hyperinsulinism. J Int Chir 1938; 3: 237-76 2 Marks V, Teale JD. Investigation of hypoglycaemia. Clin Endocrinol 1996; 44: 133-6 3 Grant CS. Surgical aspects of hyperinsulinaemic hypoglycaemia. Endocrinol Metab Clin N Am 1999; 28: 533-54 4 Redmon JB, Nuttall FQ. Autoimmune hypoglycaemia. Endocrinol Metab Clin N Am 1999; 28: 603-18 5 Teale JD, Pearse AG, Marks V. The measurement of insulin, C- peptide and b-hydroxybutyrate in the differential diagnosis of spontaneous hypoglycaemia. In: Andreani D, Marks V, Lefebvre PJ, eds. Hypoglycemia. New York: Raven Press, 1987: 281-2 6 Service FJ. Classi cation of hypoglycemic disorders. Endocrinol Metab Clin N Am 1999; 28: 501-17 7 Service FJ, Dale AJ, Elveback LR, Jiang NS. Insulinoma: clinical and diagnostic features of 60 consecutive cases. Mayo Clinic Proc 1976; 51: 417-29 8 Hammond PJ, Jackson JA, Bloom SR. Localisation of pancreatic endocrine tumours. Clin Endocrinol 1994; 40: 3-14 9 Liessi G, Pasquali C, D Andrea AA, Scandellari C, Pedrazzoli S. MRI in insulinomas: preliminary ndings. Eur J Radiol 1992; 14: 46-51 10 Doppman JL, Miller DL, Chang R, Shawker TH, Gordon P, Norton JA. Insulinomas: localisation with selective intraarterial injection of calcium. Radiology 1991; 178: 237-41 11 Dolan JP, Norton JA. Occult insulinoma. Br J Surg 2000; 7: 385-7 12 Marks V. Insulinoma. In: Grossman A, ed. Clinical Endocrinology. Oxford: Blackwell Science, 1998: 531-9 13 Marks V, Rose FC. Hypoglycaemia, 2nd edn. Oxford: Blackwell Science, 1980 14 Teale JD, Starkey BJ, Hampton SM, Robinson J, Fisher RA, Levine DF. The prevalence of factitious hypoglycaemia due to sulphonylurea abuse in the UK: a preliminary report. Pract Diab 1989; 6: 177-8 15 Labib M, Teale D, Marks V. Insulin-like growth factor I in patients with hypoglycaemia. Ann Clin Biochem 1990; 27: 107-9 Accepted for publication 9 May 2003