European Journal of Endocrinology (2007) 157 75–83 ISSN 0804-4643

CLINICAL STUDY

Insulin, C-peptide and proinsulin for the biochemical diagnosis

of hypoglycaemia related to endogenous hyperinsulinism
D Vezzosi, A Bennet, J Fauvel1 and P Caron
Department of Endocrinology, Centre Hospitalier Universitaire Rangueil, 1 Avenue Jean Poulhès, TSA 50032, 31059 Toulouse Cedex 9, France and
1
Department of Biochemistry, Centre Hospitalier Universitaire Purpan, 1 Place du Docteur Baylac, 31059 Toulouse Cedex 9, France
(Correspondence should be addressed to P Caron; Email: caron.p@chu-toulouse.fr)

Abstract
Objective: We evaluated the respective value of insulin, C-peptide and proinsulin levels in 33 patients
with endogenous hyperinsulinism and in 67 controls to determine the best parameters and thresholds
to make or to rule out the diagnosis of endogenous hyperinsulinism.
Results: When blood glucose levels were below 2.5 mmol/l, insulin was !21 pmol/l in 8–35% of the
patients and in all controls; C-peptide was O0.2 nmol/l in all insulinomas but not in the
nesidioblastosis or in the controls; proinsulin was O5 pmol/l in all patients but not in the controls.
When fasting blood glucose levels reached 2.5–3.3 mmol/l, proinsulin was !22 pmol/l in all the
controls and O22 pmol/l in 74% of the patients. Proinsulin after an overnight fast was below
22 pmol/l in all non-obese controls and above 22 pmol/l in 73% of non-obese patients.
Conclusion: Proinsulin levels above 5 pmol/l with blood glucose levels below 2.5 mmol/l during a 72 h
fast test represent the best criterion for the diagnosis of endogenous hyperinsulinism, reaching 100%
diagnostic specificity and sensitivity. Concomitant C-peptide levels above 0.2 nmol/l also make the
diagnosis of all our insulinoma patients, not the diagnosis of nesidioblastosis, while insulin levels have
much less diagnostic accuracy. Whether proinsulin levels above 22 pmol/l could also make the
diagnosis of endogenous hyperinsulinism in part of the patients at the time of fasting blood glucose
levels between 2.5 and 3.3 mmol/l or after an overnight fast in non-obese subjects needs further study.

European Journal of Endocrinology 157 75–83

Introduction 3 mIU/l (21 pmol/l) at the time of symptomatic

hypoglycaemia (4–6). Finally, the availability of proin-
The biological criteria for the diagnosis of fasting sulin assays led to the use of serum proinsulin thresholds
hypoglycaemia related to endogenous hyperinsulinism as diagnostic tools; a 5 pmol/l proinsulin cut-off level at
have been challenged within recent years. the time of hypoglycaemia below 0.45 g/l was rec-
They were previously based on the finding of ommended by Service et al. as a diagnostic criterion (1),
inappropriate serum levels of insulin and C-peptide, but others selected a higher threshold (22 pmol/l (7),
and, more recently, of proinsulin, at the time of fasting 20 pmol/l (8)), for the diagnosis of insulinomas.
hypoglycaemia. However, to date, there is no general In addition, since the proportion of proinsulin
agreement regarding the diagnostic thresholds to reach secreted by insulinoma cells is generally higher than
in order to conclude that insulin, C-peptide or proinsulin that secreted by normal b cells, high proinsulin levels
levels are inappropriate. Several years ago, ratios were suggested to be a diagnostic tool for insulinomas
calculated from insulin and blood glucose levels were whatever concomitant blood glucose levels may be.
employed. Then, a 6 mIU/l (43 pmol/l) insulin threshold Measurement of proinsulin levels could lead to the
and a 0.6 ng/ml (0.2 nmol/l) C-peptide threshold diagnosis of insulinoma for blood glucose levels ranging
concomitant with symptomatic hypoglycaemia below from 2.5 to 3.3 mmol/l during the fast test (9) or when
0.45 g/l (2.5 mmol/l) was established by Service et al. evaluated after an overnight fast, even without
(1), while others recommended cut-off values of 5 mIU/l concomitant hypoglycaemia (7).
(36 pmol/l) and 0.9 ng/ml (0.3 nmol/l) at the time of The objective of our study was to assess, in 33
hypoglycaemia below 0.5 g/l (2.8 mmol/l; (2)). With the patients with hypoglycaemia related to endogenous
insulin and C-peptide-specific assays devoid of signi- hyperinsulinism and in 67 controls, the respective
ficant cross-reaction with proinsulin which are now in values of serum levels of insulin, C-peptide and
use, a lower insulin threshold of 3 mIU/l (21 pmol/l) was proinsulin levels and, among these three parameters
recommended by Service et al. (3), but already there are now in use and traditional ratios derived from these
reports of insulinoma patients with insulin levels below parameters, to determine the best parameter and

q 2007 Society of the European Journal of Endocrinology DOI: 10.1530/EJE-07-0109

Online version via www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
76 D Vezzosi and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157

threshold level to make or to rule out the diagnosis of Assays

endogenous hyperinsulinism during the fast test. The
second objective of our study was to evaluate whether Blood glucose was measured by standard hexokinase
method.
measurement of proinsulin levels after an overnight
Insulin was measured with an immunoradiometric
fast could allow to make or to rule out diagnosis of
(IRMA) kit provided by Pasteur Diagnostics, Marnes-la-
endogenous hyperinsulinism in order to avoid the fast
Coquette (Bi-Insulin Bio-Rad kit) in 17 patients and in
test.
18 controls. There is no significant cross-reaction with
proinsulin. The inter- and intra-assay coefficients of
variation (CV) are 8.0 and 3.8% respectively, and the
smallest detectable level is 1.4 pmol/l (0.2 mIU/l; to
Subjects convert insulin values from pmol/l to mIU/l, divide by
7.175).
Patients with endogenous hyperinsulinism The normal range for fasting serum insulin levels is
The 33 patients (11 men and 22 women) aged 57G16 14.3–122 pmol/l (2–17 mIU/l).
years (meanGS.D.; 18–85 years) had a body mass index Insulin was measured with an automated immuno-
(BMI) of 25.2G5.1 kg/m2 (19–42.8 kg/m2; BMIR chemiluminometric (ICL) assay provided by Bayer
30 kg/m2 in five patients). In 32 patients, the diagnosis Diagnostics (ADVIA Centaur insulin assay) in 28 patients
of insulinoma was confirmed by histopathological and in 59 controls (12). The normal range for fasting
examination after surgery, and mean insulinoma size serum levels is 12.2–222.4 pmol/l (1.7–40 mIU/l). The
was 21G13 mm (8–60 mm). Diagnosis of nesidioblas- inter- and intra-assay CV are 5.3 and 4.7% respectively.
tosis was confirmed in one patient by histopathological The minimum detectable concentration is 3.6 pmol/l
examination after a left-sided resection of the pancreas. (0.5 mIU/l). There is no significant cross-reaction with
In this patient, the reported investigations were proinsulin.
performed when hypoglycaemic symptoms recurred Insulin was evaluated with both IRMA and ICL
one year after surgery. Liver and renal functions were assays in 12 patients and in 10 controls.
normal in all patients. C-peptide levels were measured with the automated
immunochemiluminometric method, ADVIA Centaur
(Bayer Diagnostics; (12)). The smallest detectable level
is 0.02 nmol/l (0.05 ng/ml, to convert C-peptide values
Controls from nmol/l to ng/ml, multiply by 3). The inter- and
The 67 controls (22 men and 45 women) were aged intra-assay CV are 8.3 and 3.7% respectively. The
43G16 years (17–85 years). BMI was 23.3G5.3 kg/m2 normal range for fasting C-peptide levels is 0.26–
(17.4–44.5 kg/m2; BMIR30 kg/m2 in seven controls). 0.63 nmol/l (0.78–1.89 ng/ml). There is no significant
Hypoglycaemia related to endogenous hyperinsulinism cross-reaction with proinsulin.
was excluded if, during the 72-h fast test, blood glucose In our study, the thresholds of 21 pmol/l (3 mIU/l;
did not decrease below 3.3 mmol/l or if serum b-hydro- (1)) for insulin-ICL and insulin-IRMA and 0.2 nmol/l
xybutyrate concentrations reached 2700 mmol/l with (0.6 ng/ml; (1)) for C-peptide were used for comparison.
concomitant blood glucose between 2.5 and 3.3 mmol/l Proinsulin was measured using the Human Proinsulin
(10, 11). All controls had normal liver and renal RIA kit, provided by Linco Research (St Charles, MO,
functions. USA). It yields no significant cross-reactivity with insulin
(!0.1%) or C-peptide (!0.1%). The inter- and intra-
assay CV are 7.7 and 6.9% respectively. The smallest
detectable level is 2 pmol/l. The normal levels for fasting
serum proinsulin levels are 7.9G1.5 pmol/l. Two
Methods previously recommended proinsulin threshold levels
The 72-h fast test (5 pmol/l, (1), 22 pmol/l, (7)) were used for comparison.
b-Hydroxybutyrate levels were measured by an
Patients and controls were fasted for up to 72 h under automated kinetic method in samples collected in
close medical supervision. Blood glucose, insulin, vacutainer tubes containing sodium fluoride and
C-peptide and proinsulin levels were measured after potassium oxalate (13).
an overnight fast and then every 4 h during the 72-h
fast test. Additional samples were taken when blood
glucose was !2.5 mmol/l or if clinical symptoms of Calculation of insulin/blood glucose,
hypoglycaemia occurred. The fast test was discontinued
proinsulin/insulin and proinsulin/blood
before 72 h when symptomatic hypoglycaemia with
glucose indices
blood glucose below 2.5 mmol/l was presented by the The ratio of insulin to blood glucose was calculated as
patients (1). follows: insulin (mIU/l)/blood glucose (g/l). Turner’s

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157 Diagnosis of insulinoma 77

ratio was calculated as follows: insulin (mIU/l)/(blood Table 2 shows the sensitivity and specificity observed
glucose (mg/dl)-30)!100. The ratio of proinsulin to with the various parameters studied using the cut-off
insulin was calculated as proinsulin (pmol/l)/insulin values previously established or determined with the
(mIU/l) and the ratio of proinsulin to blood glucose as ROC curves after an overnight fast. Proinsulin levels
proinsulin (pmol/l)/blood glucose (g/l). appear to be the best and most simply evaluated
The thresholds used in our study as references for parameter to distinguish patients from controls, either
comparison were 30 for insulin/blood glucose ratio and with the already established cut-off value of 22 pmol/l
50 for Turner’s ratio. or with the 20.7 pmol/l cut-off value established with
the ROC curves. The sensitivity and specificity obtained
Statistical analysis with the proinsulin/blood glucose and the proinsulin/
insulin-IRMA ratios with the cut-off values determined
Non-parametric methods such as the Mann–Whitney using the ROC curves are similar to those of proinsulin
U test were used for statistical analysis (Statview 5 levels. Insulin, C-peptide levels, Turner’s ratio or
program, SAS Institute Inc., Cary, NC, USA). The results insulin/blood glucose ratio did not improve the
were considered to be significant if P!0.05. diagnostic accuracy with the cut-off values previously
Receiver-operating characteristic (ROC) curves were reported and those established with the ROC curves
constructed using the NCSS 2000 program (Kaysville, UT, (data not shown).
USA) to examine the diagnostic test performance. Proinsulin levels were above 22 pmol/l in 24 out of
Sensitivity against 1-specificity was plotted at each 33 patients (73%), whereas they were below
threshold level and the area under the curve (AUC) was 22 pmol/l in 66 out of 67 controls (sensitivityZ
computed by the non-parametric Wilcoxon test. AUC
73%, specificityZ98%). The only control with a
represents the probability of correctly identifying controls
proinsulin level above 22 pmol/l was obese (BMIZ
and patients with endogenous hyperinsulinism. Avalue of
31 kg/m 2 ). Considering the non-obese subjects,
0.5 means that the result is no better than chance.
proinsulin levels were above 22 pmol/l in 73% of the
patients and in none of the controls. Thus, in non-
obese patients and controls (BMI!30 kg/m2), the
specificity of proinsulin after an overnight fast with a
Results threshold of 22 pmol/l reached 100% and was better
Blood glucose, insulin, C-peptide and than in the entire population, while the sensitivity
proinsulin after an overnight fast was similar (Z70%). Sensitivity and specificity in
non-obese patients and controls were not improved by
The patients with endogenous hyperinsulinism had the cut-off value of 20.7 pmol/l established with the
lower blood glucose, higher proinsulin, insulin and ROC curves (data not shown).
C-peptide levels than controls (Table 1). Out of 33, 20
(61%) patients had blood glucose levels between 2.5
and 3.3 mmol/l and 9 patients (27%) had levels below Blood glucose levels during the fast test
2.5 mmol/l, whereas all controls had levels above
The fast test lasted 8G11 h (0–48 h) in the patients
3.3 mmol/l. Since the diagnosis of endogenous hyper-
insulinism is easily made by measuring C-peptide or and 72 h in all the controls. At the end of the fast test,
proinsulin in patients with morning blood glucose levels blood glucose levels were 2.0G0.3 mmol/l (1.2–
below 2.5 mmol/l (1), we selected only those 24 out of 2.4 mmol/l) in patients, whereas they were 4.5G
33 patients with morning blood glucose levels above 0.5 mmol/l (3.5–6.0 mmol/l) in controls (P!0.0001).
2.5 mmol/l and compared their results with those of the All the patients and 4 controls reached blood glucose
67 controls in order to determine if proinsulin, insulin levels below 2.5 mmol/l. The controls who reached
or C-peptide levels after an overnight fast could make such blood glucose levels (2.3–2.4 mmol/l) were four
the diagnosis of insulinoma even when blood glucose women aged 52G23 years (32–85 years) with a BMI of
levels do not decrease below 2.5 mmol/l. 19.8G2.3 kg/m2 (17.4–22.3 kg/m2).

Table 1 Blood glucose, insulin, C-peptide and proinsulin levels in 33 patients and in 67 controls after an overnight fast. To convert insulin
values to mIU/l, divide by 7.175. To convert C-peptide values to ng/ml, multiply by 3.

Patients Controls
MeanGS.D. Range MeanGS.D. Range P

Blood glucose (mmol/l) 3.0G1.0 1.3–5.1 4.5G0.5 3.5–6.0 !0.0001

Proinsulin (pmol/l) 79.2G100.6 3.3–460 8.4G4.9 2–31 !0.0001
Insulin (pmol/l) 81.1G58.8 14.3–265.5 45.9G27.2 7.2–143.5 !0.004
C-peptide (nmol/l) 0.8G0.4 0.2–2 0.5G0.2 0.03–1 !0.0002

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
78 D Vezzosi and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157

Table 2 Sensitivity and specificity of hormonal parameters estab- levels were below 5 pmol/l in 22 controls (47%), in the
lished with previous thresholds or with receiver-operating charac- patient with nesidioblastosis and in none of the
teristic (ROC) curves and evaluated after an overnight fast in patients insulinoma patients. On the other hand, proinsulin
and controls with fasting blood glucose level O2.5 mmol/l. The
proinsulin threshold of 22 pmol/l, the insulin threshold of 21 pmol/l levels were above 22 pmol/l in 23 (74%) insulinoma
(3 mUI/l), the C-peptide threshold of 0.2 nmol/l (0.6 ng/ml), the patients and in none of the controls.
insulin/blood glucose threshold of 30, the Turner’s ratio threshold of Neither insulin nor C-peptide levels can be used to
50 were the previously established thresholds (1, 3, 7); notice that the make or to rule out the diagnosis of inappropriate
22 pmol/l proinsulin threshold is the only one that has been reported
to identify the patients with endogenous hyperinsulinism whatever insulin secretion when blood glucose is 2.5–3.3 mmol/l
the concomitant blood glucose level may be (7). The other thresholds (Tables 5 and 6). However, with the cut-off values
were established using ROC curves in 23 patients with blood glucose established by the ROC curves (31.6 pmol/l for insulin-
levels after an overnight fast above 2.5 mmol/l and in 67 controls. For IRMA and 28 pmol/l for insulin-ICL), diagnostic
calculation of the ratios, insulin was expressed in mIU/l, blood
glucose levels in g/l and proinsulin in pmol/l (see Methods).
specificity reaches 89–95%.
Insulin-IRMA or Insulin-ICL and C-peptide levels
Threshold Sensitivity Specificity were higher (P!0.0001) in patients than in controls
value (%) (%) (Table 4). However, there was an overlap between the
results observed in patients and those of the controls
Insulin-IRMA 21 pmol/l 67 19
Insulin-ICL 21 pmol/l 90 10 when blood glucose levels reached 2.5–3.3 mmol/l
C-peptide 0.2 nmol/l 100 12 during the fast test.
Proinsulin 22 pmol/l 73 98 The ratios evaluated in the present study had either
20.7 pmol/l 76 97 poor sensitivity or specificity in comparison with those
Insulin-IRMA/blood 30 100 0
glucose 11.4 29 100
of proinsulin levels, except the proinsulin/blood glucose
ratio, which gave sensitivity and specificity almost
Insulin-ICL/blood 30 100 0 similar to those of proinsulin levels (Table 3).
glucose 11 70 86
Turner’s ratio 50 29 100
(with insulin-IRMA) 20.7 64 95 Diagnostic accuracy of proinsulin, insulin and
Turner’s ratio 50 5 100
C-peptide levels and derived ratios during the
(with insulin-ICL) 21.8 59 90 72-h fast test at the time when blood glucose
levels are below 2.5 mmol/l
Proinsulin/blood 15.4 91 87
glucose Proinsulin levels were above the 5 pmol/l threshold in
Proinsulin/insulin- 3.8 83 90 all patients, below the 22 pmol/l threshold in 7 out of
IRMA
Proinsulin/insulin-ICL 2.9 56 93 33 patients (22%) and below 5 pmol/l in the four
controls who reached blood glucose levels below
2.5 mmol/l at the end of the fast (Tables 5 and 6).
Insulin-IRMA levels were below 21 pmol/l in 6 out of
the 17 (35.3%) patients and insulin-ICL levels were
Diagnostic accuracy of proinsulin, insulin and below 21 pmol/l in 3 out of the 28 (11%) patients. In
C-peptide levels and derived ratios when blood the four controls with blood glucose level below
glucose levels were 2.5–3.3 mmol/l during the 2.5 mmol/l, insulin-ICL levels were low.
72-h fast test C-peptide levels were above 0.2 nmol/l in all insuli-
noma patients. The only patient who had C-peptide
Thirty-one patients and 47 controls had blood glucose levels below 0.2 nmol/l (Z0.1 nmol/l) was the patient
levels between 2.5 and 3.3 mmol/l during the fast test. with nesidioblastosis. The four controls who reached
Such blood glucose levels were reached after 7G10 h of blood glucose levels below 2.5 mmol/l had concomitant
fasting (0–34 h) in patients versus 47G17 h (24–72 h) C-peptide levels below 0.2 nmol/l.
in controls. Regarding the traditional ratios calculated with
Table 3 shows the sensitivity and specificity of the insulin levels, their diagnostic accuracy was lower
various parameters studied using the cut-off values than that of proinsulin or C-peptide levels (data not
previously established or determined with the ROC shown). If the thresholds established with the ROC
curves. Proinsulin levels appear to be the best curves for blood glucose levels of 2.5–3.3 mmol/l were
parameter, either with the already established cut-off used for the results observed when blood glucose levels
value of 22 pmol/l or with the 24 pmol/l cut-off value were below 2.5 mmol/l, the diagnostic accuracy of these
established with the ROC curves (specificityZ100%, ratios improved but remained poor. The insulin-
sensitivityZ73–76%). IRMA/blood glucose ratio still did not reach the
Proinsulin levels were higher in the 31 patients who threshold of 11.4 in 53% of the patients (9 out of 17
presented with blood glucose levels between 2.5 and patients) and the same ratio calculated with insulin-ICL
3.3 mmol/l during the fast test than in the 47 controls did not reach the threshold of 7.3 in 11% of the patients
with similar blood glucose levels (Table 4). Proinsulin (3 out of 28 patients). Turner’s ratio calculated with

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157 Diagnosis of insulinoma 79

Table 3 Sensitivity and specificity of hormonal parameters estab- and 3.3 mmol/l were used for the values observed with
lished with previous thresholds or with receiver-operating charac- concomitant blood glucose levels below 2.5 mmol/l, the
teristic (ROC) curves and evaluated in patients and controls at the time four controls had normal results but proinsulin/insulin-
of blood glucose between 2.5 and 3.3 mmol/l. The proinsulin
thresholds of 5 and 22 pmol/l, the insulin threshold of 21 pmol/l IRMA ratio was below 7.8 in 9 out of 17 (53%) patients,
(3 mIU/l), the C-peptide threshold of 0.2 nmol/l (0.6 ng/ml), the proinsulin/insulin-ICL ratio was below 7.2 in 12 out of
insulin/blood glucose threshold of 0.3, and the Turner’s ratio threshold 28 (43%) patients and proinsulin/blood glucose ratio
of 50 were the previously established thresholds (1, 3, 7). The other was below 26.8 in 3 out of 33 (9%) patients.
thresholds were established using ROC curves in 31 patients and in 47
controls with blood glucose levels between 2.5 and 3.3 mmol/l. For
calculation of the ratios, insulin was expressed in mIU/l, blood glucose
levels in g/l and proinsulin in pmol/l (see Methods).
Diagnostic accuracy of proinsulin during the
72-h fast test at the time when blood glucose
Sensitivity Specificity levels are below 2 mmol/l
Threshold (%) (%)
Since no control subject reached blood glucose levels
Insulin-IRMA 21 pmol/l 56 50 below 2 mmol/l during the fast test, we studied the
(3 mIU/l) values found in those 21 insulinoma patients in whom
31.6 pmol/l 41 95
(4.4 mUI/l)
the fast test was prolonged until they experienced
Insulin-ICL 21 pmol/l 81 45 symptomatic hypoglycaemia below 2 mmol/l (1.8G
(3 mUI/l) 0.2 mmol/l, range 1.2–2 mmol/l). At the time of such
28 pmol/l 76 89 blood glucose levels, proinsulin levels above 5 pmol/l
(3.9 mIU/l) and C-peptide levels above 0.2 nmol/l were still the best
C-peptide 0.20 nmol/l 90 55
(0.6 ng/ml) parameters to identify an inappropriate insulin
0.23 nmol/l 90 84 secretion, while insulin levels or insulin/glucose ratio
(0.7 ng/ml) were below the diagnostic thresholds in some patients.
Proinsulin 5 pmol/l 97 42 Proinsulin levels were 95.3G89.8 pmol/l (range 7.4–
22 pmol/l 76 100
24 pmol/l 73 100
425), insulin-ICL levels were 90.3G56.8 pmol/l (range
Insulin-IRMA/blood 30 100 0 7.2–186.6), insulin-IRMA levels were 91.5G
glucose 11.4 29 100 99.8 pmol/l (range 7.2–301.4) and C-peptide levels
Insulin-ICL/blood 30 100 2
were 0.9G0.5 nmol/l (range 0.4–2.6).
glucose 7.3 71 92
Turner’s ratio 50 25 100
(with insulin-IRMA) 21 41 95
Discussion
Turner’s ratio 50 33 100
(with insulin-ICL) 16 76 92 The biological diagnosis criteria of endogenous hyper-
Proinsulin/blood 26.8 77 97 insulinism have been challenged within the last years.
glucose In the past years, ratios calculated with insulin, blood
Proinsulin/insulin- 7.8 47 95 glucose levels, then with proinsulin and insulin levels
IRMA
had been recommended (1, 3, 7). New assays leading to
sensitive and specific evaluation of insulin, C-peptide
and proinsulin are now in use and several thresholds
insulin-IRMA did not reach the threshold of 21 in 6 out were determined in order to make the diagnosis
of 16 (37.5%) patients and the same index calculated of endogenous hyperinsulinism. At the time of
with insulin-ICL remained below the threshold of 16 in hypoglycaemia below 2.5 mmol/l, threshold levels of
3 out of 28 (11%) patients. 0.2 nmol/l for C-peptide and 3 mIU/l (21 pmol/l) for
Regarding the ratios calculated from proinsulin insulin (with an insulin-specific immunochemilumino-
levels, if the thresholds established with the ROC curves metric assay) were recommended by Service (3); the
at the time when blood glucose levels were between 2.5 proinsulin threshold was found to be 5 pmol/l for such

Table 4 Insulin, C-peptide and proinsulin levels in 31 patients and in 47 controls when blood glucose level was between 2.5 and 3.3 mmol/l
during the 72-h fast.

Patients Controls
MeanGS.D. Range MeanGS.D. Range P

Proinsulin (pmol/l) 90G115 2.8–540 5.9G3.6 2–21 !0.0001

Insulin-IRMA (pmol/l) 54.5G63.8 3.2–208.1 15.8G9.3 2.1–46.6 !0.0009
Insulin-ICL (pmol/l) 63.1G58.8 7.2–344.4 15.8G10.0 0.7–64.1 !0.0001
C-peptide (nmol/l) 0.7G0.5 0.07–2 0.17G0.1 0.03–0.6 !0.0001

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
80 D Vezzosi and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157

Table 5 Insulin, C-peptide and proinsulin levels in 33 patients and several patients (11–35%) with endogenous
in 4 controls when blood glucose level was below 2.5 mmol/l. In all hyperinsulinism have insulin levels below 21 pmol/l
patients, we collected at least two samples with blood glucose,
insulin, C-peptide and proinsulin levels at the time of such blood
(4). Using two insulin-specific assays, both devoid of any
glucose levels. significant cross-reaction with intact proinsulin, 35% of
the patients with endogenous hyperinsulinism had an
Patients Controls insulin-IRMA level below 21 pmol/l and 11% of them
MeanGS.D. Range MeanGS.D. Range had an insulin-ICL level below 21 pmol/l. Thus, at the
time of symptomatic hypoglycaemia below 2.5 mmol/l,
Proinsulin 113G140 7–720 2.3G0.6 2–3.2 an insulin level below 21 pmol/l cannot rule out the
(pmol/l)
Insulin-IRMA 66G76 5–344 ND ND diagnosis of endogenous hyperinsulinism (4–6).
(pmol/l) On the other hand, a concomitant serum C-peptide
Insulin-ICL 70G47 7–186 12.5G3.6 7–14.2 level above a 0.2 nmol/l cut-off value is an excellent
(pmol/l) criterion to establish the diagnosis of insulinoma. In our
C-peptide 0.8G0.5 0.1–2.6 0.1G0.03 0.07–0.13
(nmol/l) study, serum C-peptide was above 0.2 nmol/l in all
insulinomas. The only patient with a C-peptide level
below 0.2 nmol/l, at the time of hypoglycaemia below
blood glucose levels (1), while others suggested a 2.5 mmol/l was the patient with nesidioblastosis. In this
22 pmol/l threshold (7). Others recommended different patient, such results were observed when hypoglycae-
cut-off levels (2, 7). mia reappeared 1 year after left-sided pancreatectomy.
In the present study, we confirm our previous finding The lower secretion of glucagon consecutive to the
that at the time of hypoglycaemia below 2.5 mmol/l, partial pancreatectomy might at least partly explain
Table 6 Data at discontinuation of fasts in 33 patients with endogenous hyperinsulinism. All patients were symptomatic at the time of
discontinuation of the fast test. Duration of fastZ0 indicates that the patient had spontaneous symptomatic hypoglycaemia at the beginning
of their stay in the hospital for the fast test and as of medical supervision of the fast was started; in such patients, blood samples were
collected twice within the first 20–30 min of the test and then the fast was discontinued.

Duration Blood glucose Insulin levels C-peptide levels Proinsulin levels Insulin Turner’s
Patients of fast (mmol/l) (pmol/l) (nmol/l) (pmol/l) (mIU/l/blood glucose g/l) ratio

1 10 2.09 57.4a 0.9 400 21.1 100

2 1 2.47 100.5a 1.1 60 31.1 93.3
3 4 2.31 21.5a 0.4 20 7.1 25
4 0 2.42 31.6a 0.4 14 10 31.4
5 4 1.82 208.1b 1.8 37 65.8 312.5
6 30 1.93 14.4b 0.4 24 5.7 40
7 2 1.98 93.3a 1.0 425 36.1 216.7
8 4 2.15 86.1a 1.4 35 30.8 133.3
9 1 2.37 122b 0.9 11 39.5 130.8
10 4 2.37 7.2b 0.4 34 2.3 7.7
11 4 1.27 301.4b 2.6 96 182.6 25.8
12 1 2.15 35.9a 0.5 16 12.8 55.6
13 1 1.21 150.7a 1.3 198 95.5 1300.0
14 1 1.93 14.4a 0.7 82 5.71 40
15 4 1.93 43.1a 0.6 71 17.1 120
16 1 2.15 93.3a 1.0 111 31 108.3
17 10 1.87 157.9a 1.2 13 64.7 550
18 48 2.20 7.2a 0.1 7 2.5 10
19 20 2.09 107.6a 0.7 132 39.5 187.5
20 4 2.26 93.3a 0.8 550 31.7 118.2
21 24 2.47 64.6a 0.9 240 20 60
22 24 1.98 157.9a 0.6 37 61.1 366.7
23 4 1.98 35.9a 0.3 210 15.6 250
24 24 1.92 165a 1.4 71 67.6 575.0
25 1 2.14 35.9a 0.4 59 12.8 55.6
26 4 2.42 35.9a 0.3 14 11.4 35.7
27 0 1.98 93.3a 0.7 160 36.1 216.7
28 10 1.87 57.4a 0.9 96.8 18.2 57.1
29 0 1.70 122a 1.0 73 54.9 1700
30 4 1.98 43.1a 0.6 7.4 16.7 100
31 0 2.03 57.4a 0.9 92 9.1 28.6
32 10 1.80 35.9a 0.5 82 15.2 166.7
33 8 2.42 7.2a 0.3 25 2.27 7.1
a
ICL insulin.
b
IRMA insulin.

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157 Diagnosis of insulinoma 81

that symptomatic hypoglycaemia could occur in this First, we confirm that healthy subjects can have
patient with lower insulin and C-peptide levels than blood glucose levels below 3.3 mmol/l or even
those of the insulinoma patients. In addition, it has below 2.5 mmol/l during the fast test (1, 22), the
already been observed that patients with nesidioblas- lowest level being above 2 mmol/l (2.3 mmol/l) in our
tosis can present a lower secretion of C-peptide and study.
insulin than insulinoma patients (14, 15), so that the When blood glucose levels are between 2.5 and
diagnosis of nesidioblastosis is often more difficult than 3.3 mmol/l during the fast test, insulin and C-peptide
that of insulinomas. Moreover, it has been shown that levels overlapped between controls and patients (9).
patients with nesidioblastosis and even some rare Thus, they cannot be used as diagnostic tools even
glucose-sensitive insulinoma patients can undergo a when measured with new specific assays. For such
72-h fast test without hypoglycaemia (15), such blood glucose levels, it is known that b-hydroxybutyrate
patients being detected only with an oral glucose levels represent the most accurate diagnostic criterion.
tolerance test (OGTT) or a meal test. In our study, at the time of such blood glucose levels,
In our study, proinsulin levels above 5 pmol/l during proinsulin levels above 22 pmol/l would make the
the fast test at the time of hypoglycaemia below diagnosis of endogenous hyperinsulinism in part of the
2.5 mmol/l appears to be the best criterion to make patients. Conversely, proinsulin levels below 5 pmol/l
the diagnosis of endogenous hyperinsulinism with both would rule out the diagnosis of insulinoma in our
100% sensitivity and specificity. The four controls that patients, not the diagnosis of nesidioblastosis. However,
experienced blood glucose levels below 2.5 mmol/l had no conclusion can be drawn for subjects whose
concomitant proinsulin below 5 pmol/l. Conversely, all proinsulin levels are between 5 and 22 pmol/l for
the patients with endogenous hyperinsulinism had such blood glucose levels. The use of ROC curves did not
proinsulin levels above 5 pmol/l but 18% of them had allow to establish a new proinsulin threshold improving
proinsulin below 22 pmol/l. The 5 pmol/l, not the the diagnostic accuracy of proinsulin compared with
22 pmol/l, also identifies all the insulinoma patients the thresholds previously suggested. The diagnostic
who reach blood glucose levels not reached by any of the accuracy of the ratios calculated from insulin and
controls, i.e. levels below 2 mmol/l. Therefore, a proinsulin levels is not better than that of proinsulin
itself, either with the previously cut-off values or with
proinsulin level above 5 pmol/l is an excellent criterion
the new thresholds established with the ROC curves.
to make the diagnosis of endogenous hyperinsulinism.
The fast test is a long and painful test requiring
On the contrary, the 22 pmol/l threshold would lead to
hospitalization for a few days. It would have been
several false negative diagnoses and it cannot be
interesting to find either a screening test that is easy to
recommended as a diagnostic criterion at the time of
carry-off in outpatients and could permit to select all the
hypoglycaemia below 2.5 mmol/l.
patients with endogenous hyperinsulinism, or a diag-
Traditional blood glucose/insulin ratio and Turner’s nostic test that could make the diagnosis of endogenous
ratio cannot differentiate accurately patients with hyperinsulinism without performing the fast test in
endogenous hyperinsulinism and controls, either with patients who do not present with spontaneous sympto-
the thresholds previously recommended or with the matic hypoglycaemia below 2.5 mmol/l.
new threshold values established in our study with ROC Proinsulin levels after an overnight fast could be of
curves. In the same way, the ratio of blood glucose levels interest in the screening of patients with endogenous
to proinsulin and that of proinsulin level to insulin level hyperinsulinism (7). In our study, its sensitivity and
did not result in a better diagnostic accuracy than specificity (with a threshold of 22 pmol/l) were 73 and
proinsulin levels themselves at the time of hypoglycae- 98% respectively. All controls had proinsulin levels
mia below 2.5 mmol/l. Therefore, our results confirm below 22 pmol/l except an obese subject. Therefore,
that these ratios should not be used at present. proinsulin levels above 22 pmol/l would have made the
In our study, all patients with endogenous hyper- diagnosis of endogenous hyperinsulinism with a 100%
insulinism experienced hypoglycaemia below specificity in our non-obese subjects, so that the fast
2.5 mmol/l either spontaneously or during the fast test would not have been necessary in 73% of the non-
test. However, it has been reported that some rare obese patients with endogenous hyperinsulinism. On
insulinoma patients might not reach blood glucose the other hand, based on our study, it must be pointed
levels below 2.5 mmol/l even after 72-h fast (16–21). In out that morning proinsulin levels below 22 pmol/l do
addition, the 72-h fast test is not well tolerated in all not rule out the diagnosis of endogenous hyperinsulin-
subjects. Thus, it would have been interesting to have ism. The ROC curves did not enable us to determine a
biochemical diagnostic criteria allowing the fast test to new threshold value that improved the diagnostic
be shortened or even to avoid it in patients with accuracy of proinsulin levels after an overnight fast. In
endogenous hyperinsulinism who do not present obese subjects, after an overnight fast, proinsulin levels
spontaneously with symptomatic hypoglycaemia above 22 pmol/l do not make the diagnosis of
below 2.5 mmol/l, while also ruling out the diagnosis endogenous hyperinsulinism: in 110 obese subjects
in subjects without endogenous hyperinsulinism. who did not present with glycaemic disorders (blood

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
82 D Vezzosi and others EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157

glucose levels below 6.05 mmol/l (1.10 g/l)) or suspected) is and remains the gold standard for the
hypoglycaemia related to endogenous hyperinsulin- diagnosis of endogenous hyperinsulinism.
ism), proinsulin level was 16.8G11.4 pmol/l (3.5–
78 pmol/l) and 23 of these (21%) had proinsulin levels
above 22 pmol/l after an overnight fast (unpublished
data). Other factors can contribute to insulin resistance References
and increased insulin and proinsulin levels; and such
factors might be overlooked or unknown in individual 1 Service F. Hypoglycemic disorders. New England Journal of Medicine
patients, so that one must be very cautious with the 1995 332 1144–1152.
2 Marks V & Teale JD. Investigation of hypoglycaemia. Clinical
use of proinsulin levels after an overnight fast for the Endocrinology 1996 44 133–136.
diagnosis of insulinoma patients. Based on our study, a 3 Service F. Diagnostic approach to adults with hypoglycemic
morning proinsulin level above 22 pmol/l in a non- disorders. Endocrinology and Metabolism Clinics of North America
obese, non-insulin-resistant subject should lead to high 1999 28 519–532.
suspicion of insulinoma, but whether this can be used 4 Vezzosi D, Bennet A, Fauvel J, Boulanger C, Tazi O, Louvet J-P &
Caron P. Insulin levels measured with an insulin-specific assay in
as a diagnostic criterion cannot be assessed and would patients with fasting hypoglycaemia related to endogenous
need further study in a large cohort of subjects. hyperinsulinism. European Journal of Endocrinology 2003 149
413–419.
5 Chia CW & Saudek CD. The diagnosis of fasting hypoglycemia
due to an islet-cell tumor obscured by a highly specific insulin
assay. Journal of Clinical Endocrinology and Metabolism 2003 88
Conclusion 1464–1467.
6 Chammas N, Teale J & Quin J. Insulinoma: how reliable is the
In conclusion, based on our data, among the biochemical evidence? Annals of Clinical Biochemistry 2003 40
parameters evaluated, the most reliable method to 689–693.
7 Hirshberg B, Livi A, Bartlett DL, Libutti SK, Alexander HR,
make or to rule out the diagnosis of fasting Doppman JL, Skarulis MC & Gorden P. Forty-eight-hour fast: the
hypoglycaemia related to endogenous hyperinsulinism diagnostic test for insulinoma. Journal of Clinical Endocrinology and
is to use a 5 pmol/l serum proinsulin threshold at the Metabolism 2000 85 3222–3226.
time of hypoglycaemia below 2.5 mmol/l (0.45 g/l) 8 Marks V. Recognition and differential diagnosis of spontaneous
hypoglycaemia. Clinical Endocrinology 1992 37 309–316.
during a fast test. For such blood glucose levels, all
9 Kao P, Taylor R & Service F. Proinsulin by immunochemilumino-
insulinoma patients and the nesidioblastosis patient had metric assay for the diagnosis of insulinoma. Journal of Clinical
proinsulin levels exceeding 5 pmol/l, while none of the Endocrinology and Metabolism 1994 78 1048–1051.
control subjects reached such proinsulin levels with 10 Service F & O’Brien P. Increasing serum betahydroxybutyrate
similar blood glucose levels. No ratio calculated with concentrations during the 72-hour fast: evidence against
hyperinsulinemic hypoglycemia. Journal of Clinical Endocrinology
proinsulin, insulin or blood glucose levels gave a better and Metabolism 2005 90 4555–4558.
diagnostic accuracy. C-peptide levels with a 0.2 nmol/l 11 O’Brien T, O’Brien P & Service F. Insulin surrogates in
(0.6 ng/ml) minimum threshold at the time of insulinoma. Journal of Clinical Endocrinology and Metabolism
hypoglycaemia below 2.5 mmol/l (0.45 g/l) also made 1993 77 448–451.
the correct diagnosis in all our insulinoma patients and 12 El Kenz H & Bergmann P. Evaluation of immunochemilumino-
metric assays for the measurement of insulin and C-peptide using
missed the diagnosis only in the patient with nesidio- the ADVIA Centaur. Clinical Laboratory 2004 50 171–174.
blastosis. We confirm that the diagnostic accuracy of 13 McMurray C, Blanchflower W & Rice D. Automated kinetic
insulin levels is much less than that of C-peptide and method for D-3-hydroxybutyrate in plasma or serum. Clinical
proinsulin, and that insulin levels below 3 mIU/l Chemistry 1984 30 421–425.
14 Wiesli P, Perren A, Saremaslani P, Pfammatter T, Spinas G &
(21 pmol/l) concomitant with hypoglycaemia below Schmid C. Abnormalities of proinsulin processing in functioning
2.5 mmol/l (0.45 g/l) cannot rule out the diagnosis of insulinomas: clinical implications. Clinical Endocrinology 2004 61
endogenous hyperinsulinism, when insulin is measured 424–430.
with the insulin-specific assays now in use. 15 Starke A, Saddig C, Kirch B, Tschahargane C & Goretzki P. Islet
Proinsulin levels with a cut-off value of 22 pmol/l hyperplasia in adults: challenge to preoperatively diagnose non-
insulinoma pancreatogenic hypoglycemia syndrome. World
(not insulin, C-peptide levels or ratios calculated with Journal of Surgery 2006 30 670–679.
insulin and blood glucose levels) are partly of interest for 16 Service F, Natt N, Thompson G, Grant C, van Heerden J,
the diagnosis of insulinoma when measured in patients Andrews J, Lorenz E, Terzic A & Lloyd R. Noninsulinoma
with blood glucose of 2.5–3.3 mmol/l during the fast pancreatogenous hypoglycemia: a novel syndrome of hyperinsu-
linemic hypoglycemia in adults independent of mutations in
test or after an overnight fast in non-obese subjects, but Kir6.2 and SUR1 genes. Journal of Clinical Endocrinology and
their diagnostic accuracy is much less than that found Metabolism 1999 84 1582–1589.
with the 5 pmol/l threshold when concomitant blood 17 Service F. Endocrine causes of postprandial hypoglycemia. In:
glucose levels are below 2.5 mmol/l. No basal data, Hypoglycemia, pp 45–54. Eds D Andreani, N Marks, PJ Lefebvre.
including normal proinsulin levels after an overnight New York: Raven Press, 1987
18 Thompson G, Service F, Andrews J, Lloyd R, Natt N, Van Heerden J
fast, can rule out the diagnosis of endogenous & Grant C. Noninsulinoma pancreatogenous hypoglycemia
hyperinsulinism, so the fast test (if necessary combined syndrome: an update in 10 surgically treated patients. Surgery
with an OGTT, especially when nesidioblastosis is 2000 128 937–945.

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access
EUROPEAN JOURNAL OF ENDOCRINOLOGY (2007) 157 Diagnosis of insulinoma 83

19 Rayfield E, Pulini M, Golub A, Rubenstein A & Horwitz D. Non- 22 Merimee T & Fineberg S. Homeostasis during fasting. II. Hormone
autonomous function of a pancreatic insulinoma. Journal of substrate differences between men and women. Journal of Clinical
Clinical Endocrinology and Metabolism 1976 43 1307–1311. Endocrinology and Metabolism 1973 37 698–702.
20 Jordan R & Kammer H. Case report: an insulinoma without fasting
hypoglycemia. American Journal of the Medical Sciences 1976 272
205–209.
21 Power L. A glucose-responsive insulinoma. JAMA 1969 207 Received 21 February 2007
893–896. Accepted 2 April 2007

www.eje-online.org

Downloaded from Bioscientifica.com at 12/03/2022 12:04:57PM

via free access