Abstract
Objectives
Injectable tetracosactide hexa-acetate, ACTH 1-24 (Synacthen), is not marketed in many countries including India, whereas Injectable long acting porcine sequence, ACTH 1-39 (Acton Prolongatum®) is easily available and much cheaper. This study aimed to find the diagnostic accuracy of ACTH stimulation test using i.m. Acton Prolongatum® (acton prolongatum stimulation test, APST) in comparison with Synacthen (short synacthen test, SST) for the diagnosis of glucocorticoid insufficiency.
Methods
Subjects with a suspicion of adrenal insufficiency based on clinical features underwent a SST with 250 µg Synacthen followed by APST using 30 units of Acton Prolongatum®. Serum cortisol levels were measured at 60 and 120 min following injection of Acton Prolongatum®. Stimulated peak cortisol of less than 18 µg/dL on SST was considered as adrenal insufficiency.
Results
Forty seven patients with mean age of 36.7 ± 14.4 years were enrolled for the study. Based on SST, twenty (n = 20) persons were classified as having adrenal insufficiency, whereas twenty-seven (n = 27) were found to be normal. Area under the curve of APST (at 120 min) was 0.986 when compared to SST, thus proving its high accuracy. A serum cortisol cut off value of 19.5 µg/dL at 120-min following stimulation with Acton Prolongatum® showed a sensitivity of 100% and specificity of 88%.
Conclusion
ACTH stimulation test using Acton Prolongatum® is an economical and accurate alternative to the short Synacthen test.
Introduction
Early diagnosis of adrenal insufficiency (AI) can be lifesaving (1). Morning cortisol level is useful for the detection of adrenal insufficiency in cases where it is markedly low or clearly in the upper reference range (2, 3). Still, in a majority of patients, stimulation tests are required for the definitive diagnosis of AI. Although the reference test for the diagnosis of AI is measurement of cortisol response to insulin-induced hypoglycemia, it is contraindicated in patients with ischemic heart disease and epilepsy and may be risky in patients with primary AI. Current guidelines suggest the use of standard dose (250 μg) Short Synacthen Test (SST) for the diagnosis of both primary (4) and secondary AI (5). Synacthen or Cosyntropin (tetracosactide hexaacetate) is a synthetic adrenocorticotropic hormone (ACTH) composed of the first twenty-four amino acids of the ACTH 1-39 peptide. Synacthen is not marketed in India; hence, the cost is prohibitively high. Constraints of procurement have led to avoidance of ACTH stimulation test even when indicated, with undue reliance on morning cortisol values. Synthetic porcine sequence corticotrophin in carboxymethylcellulose is freely available as Acton Prolongatum® in a 5 mL multidose vial manufactured by Ferring’s Pharma. When used at a dose of 30 IU per test it costs 14 times less than 250 µg of Synacthen. Reversible binding of ACTH to carboxymethyl cellulose protects the peptide against enzymatic breakdown thereby prolonging its action. Acton Prolongatum is preserved at 2–10°C and administered intramuscularly (6).
Although Acton Prolongatum stimulation test (APST) is used in many centers in India for diagnosing AI, there is paucity of literature to strongly support this practice. This study was done to evaluate the performance of APST and validate it against the standard SST.
Subjects and methods
The study was conducted in a prospective observational design, in the Department of Endocrinology, Government Medical College Hospital, Thiruvananthapuram, India. The study was approved by the Human Ethics Committee of Govt. Medical College Thiruvananthapuram and funded by the State Board of Medical Research, Kerala. All patients with clinical suspicion of AI (anorexia, nausea, weight loss, unexplained hyponatremia, progressive hyperpigmentation, history of pituitary macroadenoma or surgery) who attended the outpatient clinic during the study period were included. Written, informed consent was obtained from all the subjects after explaining the procedure. Demographic details and symptomatology of each individual were entered into a pre-designed proforma.
Baseline serum cortisol was measured at between 8:00 h and 9:00 h in all subjects. Short Synacthen Test (SST) was carried out by administering 250 µg of Synacthen intramuscularly to deltoid followed by venous blood sampling for serum cortisol at 30 (SST 30) and 60 (SST 60) min. Three days later APST was performed by injecting 0.5 mL (30 units) of Acton Prolongatum® intramuscularly to deltoid followed by venous blood sampling for serum cortisol at 60 (APST 60) and 120 (APST 120) min. A dose of 30 units was chosen to ensure ease of dispensing (0.5 mL) from a 300 unit in 5 mL formulation. Stimulated peak cortisol cut off of less than 18 µg/dL on SST was considered as AI based on guideline recommendations (4). Patients were closely monitored for any discomfort or adverse events. Delta cortisol was calculated as the maximum increment in serum cortisol following the administration of Acton Prolongatum®. Serum cortisol estimation was done by electrochemiluminescence method using COBAS e 411 analyzer with Cortisol II kits, which has an intra- and inter-assay coefficient of variation of 1.6–2.4% and measuring range of 0.054–63 μg/dL.
Data was tabulated using Microsoft excel and analyzed using SPSS, version 25. Continuous variables were described using mean and s.d. and categorical variables as frequency and percentages. Receiver-operating characteristic (ROC) curve was plotted using a uniform threshold for any selected variables according to 95% specificity. Area under the curve (AUC) was used to express the overall diagnostic accuracy of the index criterion. P value was considered to be significant at 5% level of significance for all comparisons. Youden index was used to find out the cut off levels at which the best possible sensitivity and specificity was obtained.
Results
During the study period, 54 patients with clinically suspected AI were encountered in the outpatient clinic, of whom four were excluded due to liver disease. Thus, 50 persons were included to undergo the SST and APST. The SST was done in 50 patients, of whom three patients did not turn up for the second test and were considered as having withdrawn consent. The remaining 47 patients were included in the final analysis (Fig. 1). The mean age of the study population was 36.7 (±14.46) years. There were 26 (55.3%) female patients and 21 (44.6%) male patients. The mean BMI of the study group was 25.3 ± 5.2 kg/m2. None of the patients experienced any immediate adverse events like nausea, vomiting, rash or anaphylaxis with Synacthen or Acton Prolongatum®.
Of the 47 patients with suspected AI, 20 (42.5%) were confirmed to have AI based on the SST using a stimulated peak cortisol cut off of less than 18 µg/dL. Of these, four patients had primary AI and the remaining had secondary AI based on clinical features and ACTH levels.
If 14.5 μg/dL is taken as the cut off for diagnosis of adrenal insufficiency on a SST, as suggested by recent studies which had used the same equipment and kits (Roche COBAS e 411 and Cortisol II), only 16 patients would be classified as having the disease (7, 8, 9).
Performance of APST in subjects with normal adrenal reserve
Among the patients without AI based on SST, the mean basal serum cortisol levels were 7.2 (±1.8) µg/dL and the values at 30 and 60 min were 20.8 (±3.2) µg/dL and 20.3 (±4.0) µg/dL respectively after Synacthen administration. The mean serum cortisol values with APST were 22.6 (±5.1) and 26.1 (±5.5) µg/dL at 60 and 120 min, respectively. The mean serum cortisol values at 1 h after stimulation with Acton Prolongatum® were numerically higher by approximately 2.3 µg/dL when compared to that after Synacthen, but the difference was not statistically significant (Table 1) (P = 0.072; 95% CI, −4.65 to 0.21).
Comparison of cortisol values at different time points during APST and SST in patients with and without AI.
| Patients without AI | Patients with AI | |||
|---|---|---|---|---|
| Time (min) | SST cortisol (μg/dL) | APST cortisol (μg/dL) | SST cortisol (μg/dL) | APST cortisol (μg/dL) |
| 0 | 7.3 ± 1.9 | 7.3 ± 1.9 | 4.6 ± 3.5 | 4.6 ± 3.5 |
| 30 | 20.9 ± 3.2 | NA | 7.8 ± 5.0 | NA |
| 60 | 20.3 ± 4.1 | 22.6 ± 5.2 | 9.3 ± 6.2 | 9.8 ± 6.1 |
| 120 | NA | 26.1 ± 5.6 | NA | 10.6 ± 6.37 |
Performance of APST in subjects with AI
Among the patients with AI, the mean basal serum cortisol levels were 4.5 ± 3.5 µg/dL. During the SST the cortisol levels had a rise to a mean of 7.9 (±5.1) µg/dL at 30 min and 9.3 (±6.2) µg/dL at 60 min. The mean serum cortisol values with APST were 9.8 (±6.0) µg/dL at 60 min and 10.6 (±6.4) µg/dL at 120 min respectively. S. cortisol values at 1 h after stimulation with Synacthen and Acton Prolongatum® were similar (P = 0.796; 95% CI −4.44 to 3.42) (Table 1).
Receiver-operating characteristic
Receiver-operating characteristic (ROC) analysis was performed for the basal as well as post APST serum cortisol levels (at 60 and 120 min) and area under the curve (AUC) was found out for basal cortisol, APST 60 min, APST 120 min, peak APST cortisol value, mean of post APST cortisol values and Delta cortisol (difference between basal and peak cortisol). Acton Prolongatum® stimulated serum cortisol at 120 min (APST 120) showed maximum area under the curve among the analyzed parameters (Fig. 2 and Table 2) A diagnostic cut off of 19.5 µg/dL provided the highest sensitivity and specificity both for APST 120 (100 and 88% respectively) and peak cortisol after APST (95 and 89%). For the proposed lower (assay specific) SST cut off of 14.5 µg/dL the corresponding cut-off for APST was found to be 18.6 µg/dL both for APST 120 (sensitivity 100% and specificity 84%) and peak cortisol after APST (sensitivity 94% and specificity 84%) based on ROC analysis.
Receiver-operator characteristic (ROC) curve plot analysis for morning basal cortisol, APST 60, APST 120 and delta cortisol. The maximum area under ROC was for APST 120 (0.986).
Citation: European Journal of Endocrinology 181, 6; 10.1530/EJE-19-0558
Receiver-operator characteristic (ROC) curve plot analysis for morning basal cortisol, APST 60, APST 120 and delta cortisol. The maximum area under ROC was for APST 120 (0.986).
Citation: European Journal of Endocrinology 181, 6; 10.1530/EJE-19-0558
Receiver-operator characteristic (ROC) curve plot analysis for morning basal cortisol, APST 60, APST 120 and delta cortisol. The maximum area under ROC was for APST 120 (0.986).
Citation: European Journal of Endocrinology 181, 6; 10.1530/EJE-19-0558
Comparison of performance among various parameters obtained from APST.
| Test | AUC | s.e. | Best cutoff (sensitivity %, specificity %) (μg/dL) | Cutoff for 100% sensitivity (μg/dL) | Cutoff for 100% specificity (μg/dL) |
|---|---|---|---|---|---|
| Basal cortisol | 0.775 | 0.080 | 5.9 (73, 80) | NA | 3.6 |
| APST 60 | 0.956 | 0.027 | 17.0 (95, 88) | 20.3 | 14.9 |
| APST 120 | 0.986 | 0.012 | 19.5 (100, 88) | 19.5 | 17 |
| Peak cortisol after APST | 0.978 | 0.017 | 19.5 (95, 89) | 20.5 | 16.7 |
| Mean of APST cortisol | 0.977 | 0.017 | 18.8 (100, 85) | 18.8 | 16.5 |
| Delta cortisol with APST | 0.966 | 0.024 | 10.8 (84, 100) | 15 | 10.8 |
Performance of morning cortisol
Morning cortisol values in patients with AI and those without AI were 7.3 ± 1.9 µg/dL and 4.6 ± 3.5 µg/dL respectively. Although there was a considerable overlap between these values, a morning cortisol level below 3.6 µg/dL provided 100% specificity for diagnosing AI (Table 2). If this value of morning cortisol is taken as a cut off for diagnosis of AI, ACTH stimulation test could have been avoided in 9 (19.1%) out of the 47 patients.
Primary versus secondary AI
Patients with AI were classified as having primary or secondary AI on the basis of clinical features and plasma ACTH levels. Of the 20 patients with AI, 16 had secondary AI and 4 had primary AI. Mean age of patients with primary and secondary AI was 44.5 (19–64) years and 38 (17–80) years and mean BMI was 24.0 kg/m2 vs 26.2 kg/m2 respectively. Autoimmune adrenalitis and tuberculosis were the probable cause in patients with primary AI. Pituitary surgery, pituitary macroadenoma, stalk interruption syndrome and prior long-term steroid exposure were the etiologies in secondary AI. The mean basal cortisol values were 2.2 (±2.9) and 4.8 (±2.9) in primary and secondary AI respectively (P = 0.135). Patients with primary AI achieved a much lower mean peak cortisol level 2.1 (±2.4) μg/dL after APST compared to those with secondary AI who had a mean peak cortisol level of 12.7 (±5.1) μg/dL.
Discussion
SST is the current standard of care for the diagnosis of AI. Unavailablility of Synacthen in large parts of the world makes the diagnosis of AI fraught with challenges of under diagnosis as well as over diagnosis. Synthetic porcine sequence corticotrophin (Acton Prolongatum®) is readily available, stable and cheaper than Synacthen. Present study was done to assess the performance of APST against SST.
Animal studies comparing porcine ACTH in gelatin (1–39 amino acid sequence) and synthetic ACTH (1–24 amino acid sequence) in female dogs revealed no significant difference and concluded that either method caused maximal secretion of the adrenocortical reserve when administered in equimolar amounts (10). Friedman et al. studied 11-hydroxycorticosteroid levels following long acting corticotrophin carboxymethyl cellulose and found peak cortisol levels around 1 to 3 h (25.7 µg/dL at 1 h and 37.4 µg/dL at 3 h) (11). Hence, in the present study, serum cortisol values up to 2 h following administration of Acton Prolongatum® were measured. All patients underwent SST followed by APST, where SST was used to classify the patients to be glucocorticoid sufficient or insufficient, and results of APST were compared. It was demonstrated that APST can be safely and reliably utilized for the diagnosis of glucocorticoid insufficiency with a high sensitivity (100%) and specificity (88%). The best cut off for stimulated cortisol after Acton Prolongatum® administration was determined to be 19.5 μg/dL (corresponding to an SST cut off of 18 μg/dL). Thirty unit dose of Acton Prolongatum® was found to be as potent as Synacthen 250 µg in stimulating cortisol secretion in normal as well as glucocorticoid-insufficient patients.
The values of ACTH stimulated cortisol are assay dependent. The proposed newer cut offs for SST while using assays with better sensitivity and analyte specificity (as in the present study) range from 12.7 to 14.5 μg/dL which would be more relevant in patients with secondary AI (7, 8, 9). If a cut off value of 14.5 μg/dL on SST was used for the diagnosis of AI in this study, the corresponding value for APST would be 18.6 μg/dL based on ROC analysis.
In practice, clinical features and plasma ACTH are used to differentiate between primary and secondary AI. In the present study it was noted that patients with primary AI achieve a much lower mean peak cortisol level than those with secondary AI. Further studies may be required to validate APST for the differentiation of primary and secondary adrenal insufficiency as this may be helpful in a setting where plasma ACTH testing is not feasible.
Gundguruthy et al. studied the efficacy of porcine sequence corticotropin in healthy volunteers and those with AI albeit without a documented comparison to a gold standard test (ITT or SST) (6). The average cortisol level achieved in patients with normal adrenal reserve in our study on APST was 26.1 ± 5.6 μg/dL, whereas that in the control group of the aforementioned study was 29.6 ± 5.9 μg/dL. Another study by Wagmode et al. documented the response of both tests but, the study included healthy volunteers only (12). Sharma et al. have compared the long acting ACTH (Acton Prolongatum®) test against ITT for the diagnosis of secondary AI in children in which the peak values of cortisol obtained on the ACTH test were much higher than those on the ITT (13). This effect was also observed with the SST in several previous studies and is probably related to the supra-physiological plasma levels of ACTH attained with 250 μg of Synacthen which is same as in the case of using Acton-Prolongatum® (14). However, the present study, as well as that by Wagmode et al. have noted insignificant differences in the stimulated cortisol values with both Synacthen and Acton Prolongatum®. The Porcine sequence corticotrophin has also been used to identify AI in critically ill patients and in patients with HIV infection from two different centers in India (15, 16). The findings of previous studies related to Acton Prolongatum® are summarized in Table 3.
Summary of previous studies on porcine ACTH.
| Reference | Study subjects | Comparative standard used | Dose of porcine ACTH | Findings |
|---|---|---|---|---|
| (10) | 10 healthy volunteers | nil | 40IU | 11-hydroxyl corticosteroid peak between 1 to 3 h |
| (11) | 15 female dogs | Synacthen 250 µg | 2.2 IU/kg | Similar maximal secretion from adrenocortical reserve by both agents |
| (6) | 24 healthy controls; 24 adrenal insufficient | nil | 25 IU | Statistically significant difference in Acton Prolongatum stimulated cortisol and delta cortisol in normal and adrenal insufficient groups |
| (12) | 15 healthy volunteers | Synacthen 250 µg | 25 IU | Comparable peak serum cortisol in tetracosactide and porcine sequence corticotropin group (P = 0.686) |
| (13) | 48 suspected patients; 28 sufficient, 20 insufficient children | Insulin Tolerance Test | 25 IU | Peak cortisol value <18 μg/dL on the IM-ACTH suggestive of AI; Serum cortisol value >22 μg/dL ruled out secondary AI in children (5–14 years) |
Using Acton Prolongatum® for diagnosis of adrenal insufficiency brings down the cost of investigation considerably. Acton Prolongatum® when used at a dose of 30 units per test costs 14 times less than the 250 µg of Synacthen and using APST in lieu of SST can bring down the total cost of diagnosing glucocorticoid deficiency from USD 44 to 11 (Indian Rupees 3100 to 800).
The strength of the present study is the comparison of APST with an existing standard of care (SST), in individuals with both glucocorticoid sufficiency as well as deficiency, being the first one to do so. Limitations of our study include absence of dose titration of porcine ACTH and lack of an assay specific cut-off for Synacthen-stimulated serum cortisol in our laboratory. As opposed to previous studies using 25 units of Acton Prolongatum® for the APST, this study used 30 units considering the ease of dispensing from a 5 mL vial containing 300 units (6, 12, 13). No adverse events were observed with Synacthen or Acton Prolongatum in the present study. Anaphylactic reactions to either can occur, although rarely in the present era due to the improved methods of purification (17).
Conclusion
Acton Prolongatum® stimulation test (APST) can accurately diagnose glucocorticoid insufficiency. Use of this long-acting porcine sequence ACTH in conjunction with morning cortisol levels can bring down the cost of diagnosing glucocorticoid insufficiency considerably and make the test accessible and affordable to large parts of the world.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of this study.
Funding
This study was funded by the provincial government (State Board of Medical Research, Kerala) vide order no. A2 (SBMR.2017-2018)/15182/2017/MCT dated 10.7.2017 granted to Dr C Jayakumari.
Acknowledgements
The authors thank Dr Mathew John, Consultant Endocrinologist at Providence Endocrine Centre, Thiruvananthapuram for the logistic help provided for this study. Dr Nandini Prasad, Dr S Soumya and Dr T Ajeesh (Senior Residents), Dr T P Seena (Physiological Assistant), S Sajeev, Arya Suresh, Arya Sugesh and Sudi Sisupalan (Research Assistants) in Department of Endocrinology, Govt. Medical College, Thiruvananthapuram, are acknowledged for their role in data collection and tabulation. Dr A Arun Kumar, Assistant Professor, Depatment of Anaesthesiology, Government Medical College, Thiruvananthapuram is acknowledged for proof reading the manuscript. The authors are thankful to the State Board of Medical Research, Kerala for funding the study.
References
- 1↑
Stewart PM, Newell-Price JDC. The adrenal cortex. In Williams Textbook of Endocrinology, 13th ed., pp 490–555. Eds Melmed S, Polonsky KS, Larsen PR & Kronenberg HM. Philadelphia, PA: Elsevier, 2016.
- 2↑
Kazlauskaite R, Evans AT, Villabona CV, Abdu TA, Ambrosi B, Atkinson AB, Choi CH, Clayton RN, Courtney CH, Gonc EN et al. Corticotropin tests for hypothalamic-pituitary-adrenal insufficiency: a metaanalysis. Journal of Clinical Endocrinology and Metabolism 2008 4245–4253. (https://doi.org/10.1210/jc.2008-0710)
- 3↑
Yo WS, Toh LM, Brown SJ, Howe WD, Henley DE, Lim EM. How good is a morning cortisol in predicting an adequate response to intramuscular synacthen stimulation? Clinical Endocrinology 2014 19–24. (https://doi.org/10.1111/cen.12373)
- 4↑
Bornstein SR, Allolio B, Arlt W, Barthel A, Don-Wauchope A, Hammer GD, Husebye ES, Merke DP, Murad MH, Stratakis CA et al. Diagnosis and treatment of primary adrenal insufficiency: an Endocrine Society clinical practice guideline. Journal of Clinical Endocrinology and Metabolism 2016 364–389. (https://doi.org/10.1210/jc.2015-1710)
- 5↑
Ospina NS, Al Nofal A, Bancos I, Javed A, Benkhadra K, Kapoor E, Lteif AN, Natt N, Murad MH. ACTH stimulation tests for the diagnosis of adrenal insufficiency: systematic review and meta-analysis. Journal of Clinical Endocrinology and Metabolism 2016 427–434. (https://doi.org/10.1210/jc.2015-1700)
- 6↑
Gundgurthi A, Garg MK, Dutta MK, Pakhetra R. Intramuscular ACTH stimulation test for assessment of adrenal function. Journal of the Association of Physicians of India 2013 320–324.
- 7↑
Raverot V, Richet C, Morel Y, Raverot G, Borson-Chazot A. Establishment of revised diagnostic cut-offs for adrenal laboratory investigation using the new Roche Diagnostics Elecsys® cortisol II assay (Letter to the Editor). Annales d’Endocrinologie 2016 620–622. (https://doi.org/10.1016/j.ando.2016.05.002)
- 8↑
Kline GA, Buse J, Krause RD. Clinical implications for biochemical diagnostic thresholds of adrenal sufficiency using a highly specific cortisol immunoassay. Clinical Biochemistry 2017 475–480. (https://doi.org/10.1016/j.clinbiochem.2017.02.008)
- 9↑
El-Farhan N, Pickett A, Ducroq D, Bailey C, Mitchem K, Morgan N, Armston A, Jones L, Evans C, Rees DA. Method-specific serum cortisol responses to the adrenocorticotrophin test: comparison of gas chromatography-mass spectrometry and five automated immunoassays. Clinical Endocrinology 2013 673–680. (https://doi.org/10.1111/cen.12039)
- 10↑
Feldman EC, Stabenfeldt GH, Farver TB, Addiego LA. Comparison of aqueous porcine ACTH with synthetic ACTH in adrenal stimulation tests of the female dog. American Journal of Veterinary Research 1982 522–524.
- 11↑
Friedman M. Comparison of duration of activity of corticotrophin-gelatin and corticotrophin-carboxymethyl cellulose. BMJ 1967 409–410. (https://doi.org/10.1136/bmj.3.5562.409)
- 12↑
Wagmode N, Mathew V, Bantwal G, Ayyar V, George B, Patanjali CP, Michael J. It is time to carpe diem with porcine ACTH – a comparison of porcine sequence corticotropin to tetracosactide hexaacetate in testing the hypothalamic pituitary adrenal axis in healthy individuals. Indian Journal of Endocrinology and Metabolism 2018 636–640. (https://doi.org/10.4103/ijem.IJEM_114_18)
- 13↑
Sharma R, Madathil S, Maheshwari V, Roy K, Kumar B, Jain V. Long-acting intramuscular ACTH stimulation test for the diagnosis of secondary adrenal insufficiency in children. Journal of Pediatric Endocrinology and Metabolism 2018 57–63. (https://doi.org/10.1515/jpem-2018-0330)
- 14↑
Reimondo G, Bovio S, Allasino B, Terzolo M, Angeli A. Secondary hypoadrenalism. Pituitary 2008 147–154. (https://doi.org/10.1007/s11102-008-0108-4)
- 15↑
Das SN, Sahu NC, Darjee DR, Mishra SK, Swaroop S, Suna SP, Salim SKM & Thatoi PK. Glucocorticoid response to intramuscular ACTH stimulation in critically ill patients. Journal of Evolution of Medical and Dental Sciences 2017 13–32.
- 16↑
Sharma N, Sharma LK, Anand A, Gadpayle AK, Gaurav K, Mukherjee S, Kulshreshtha B, Dutta D. Presence, patterns and predictors of hypocortisolism in patients with HIV infection in India. Indian Journal of Medical Research 2018 142–150. (https://doi.org/10.4103/ijmr.IJMR_43_16)
- 17↑
Rosenblum AH, Rosenblum P. Anaphylactic reactions to adrenocorticotrophic hormone in children. Journal of Pediatrics 1964 387–395. (https://doi.org/10.1016/s0022-3476(64)80191-8)
