Identification and genotoxicity evaluation of two carbamate impurities in rasagiline

Abstract

During the synthesis of a second-generation monoamine oxidase-B inhibitor rasagiline, two unknown impurities (impurity A and impurity B) were detected and isolated by preparative liquid chromatography. Based on mass spectroscopy and NMR, these two impurities were characterized as the by-products with a propargyl carbamate structure, whose generation was related to the carbon dioxide in the alkaline reaction solution. Because the carbamate structure has been highlighted as a class of potentially genotoxic impurities (GTIs), the genotoxicity of these two impurities was evaluated by the malformation test and the comet assay using zebrafish embryos. The results showed that the genotoxicity of impurity B was significant higher than that of rasagiline and other impurities. Thus, a HPLC-MS method was developed and validated for the determination of impurity B in rasagiline. The established method showed a good specificity, linearity, precision and accuracy. The detection limit of this method was 2.0 ppm with 0.1 mg ml⁻¹ rasagiline mesylate.

---

1. Introduction

Rasagiline, (R)-N-(prop-2-ynyl)-2,3-dihydro-1H-inden-1-amine, is a second-generation selective and irreversible monoamine oxidase-B (MAO-B) inhibitor.¹ It is able to ameliorate motor symptoms and prevent motor complications in Parkinson's disease (PD).² Unlike the first-generation MAO-B inhibitor, rasagiline was not metabolized to L-methamphetamine derivatives in vivo, making rasagiline more effective and well tolerated.^3,4 It gained UK- and EU-marketing authorization in 2005 and US FDA approval in 2006,⁵ and gradually became a valuable therapeutic option for use in all stages of Parkinson's disease. Separation and identification of impurities in drugs is an important issue for drug development. A literature survey reveals that two HPLC methods with different chiral stationary phases were developed to separate the rasagiline from its analogues and its (S)-enantiomer, respectively.^6,7 There are also several reports on the quantitative analysis of rasagiline mesylate in microspheres⁸ and separation of process related impurities and degradation products of rasagiline mesylate by HPLC.^9,10 During the synthesis of rasagiline mesylate, we detected and identified two new carbamate impurities by HPLC-MS and NMR. Concern about the presence of these two impurities was extremely crucial for the drug's safety, because the carbamate structure has been highlighted as a class of structural alerts or potentially genotoxic impurities (PGIs).¹¹

Genotoxicity means damage to the DNA and accumulation of DNA damage in specific genes can lead to cancer, a process that develops in the course of several years. Therefore, genotoxic impurities are garnering increased attention from regulatory agencies and the pharmaceutical industry in recent years.^12,13 The control of these impurities is more strict than that of the common impurities and their permitted limits in the long-term treatment drugs are usually in part per million (ppm) levels.¹⁴ Commonly used analytical methods for general impurity profile, i.e., HPLC-UV and GC-FID, are not suitable for determination of such impurities.¹⁵ The higher cost methods, such as MS based methods, are often needed for trace analysis of genotoxic impurities.^14–16 Based on the bacterial reverse mutation assay (Ames test), some structurally alerting functional groups have been known to be involved in reactions with DNA.¹¹ However, in vitro genotoxicity tests have a high rate of false positive results with rodent carcinogenicity.¹⁷

Zebrafish embryos have been proposed as an in vitro animal model which could bridge the gap between simple assays based on cell or tissue culture, and biological validation in whole animals such as rodents during the drug discovery.¹⁸ In recent years, the zebrafish embryos method has been proven to be a practicable and suitable avenue for detecting DNA damage induced by various genotoxicants.^19,20 The present study, therefore, attempts to differentiate the genotoxic impurities from the common impurities present in rasagiline mesylate using zebrafish embryos. Then a HPLC-MS method was developed and validated for the genotoxic impurity control. Our investigations will help the safety of this active pharmaceutical ingredients during its long-term clinical treatment.

2. Experimental

2.1 Chemicals and reagents

Rasagiline mesylate, the N-propargyl derivative, and (R)-1-aminoindan were synthesised in Changzhou No. 4 Pharmaceutical Co., Ltd. The purity of these compounds (>97.0%) were determined by respective HPLC methods. All reagents and solvents used in the synthesis were obtained from Nanjing Chemical Reagent Co., Ltd (Nanjing, China). Purified water was obtained by a Milli-Q system from Millipore (Bedford, MA, USA). Triethylamine, acetic acid, formic acid, HPLC grade acetonitrile and methanol were purchased from Merck (Darmstadt, Germany), which were used for the preparation of mobile phases.

2.2 Synthesis of rasagiline crude product

The synthesis of rasagiline crude product followed the procedures reported by the ref. 21 and the synthesis scheme was shown in Fig. 1. Briefly, (R)-1-aminoindan (1 mol) was dissolved by DMF into a round bottom flask with potassium carbonate (1 mol), followed by dropwise addition of propargyl methanesulfonate ester (1 mol) in DMF. The reaction mixture was stirred at 25–35 °C for more than 20 h. After completion of the reaction, required amount of water was added. The aqueous layer was extracted twice with dichloromethane. The combined dichloromethane layer was washed with acid water and the aqueous layer was separated. Sodium hydroxide solution was added to the aqueous layer, then it was extracted twice with ethyl acetate. The combined ethyl acetate layer was dried and distilled off under vacuum to get the title compound.

Fig. 1 The synthesis scheme for rasagiline, impurity A and impurity B.

2.3 Conditions for the detection of impurities

The rasagiline crude product was dissolved in acetonitrile–water (1 : 1, v/v) to form 1 mg ml⁻¹ solution. An Agilent 6240 LC-TOF mass spectrometer (Agilent Technologies, Santa Clara, USA) with an electrospray ionization (ESI) interface was used for the detection of impurities. Chromatographic separations were performed on an Agilent Poroshell 120 EC-C₁₈ (50 × 4.6 mm, 2.7 μm) column. A mixture of acetonitrile–water with 0.1% formic acid (1 : 1, v/v) was used as the mobile phase. The flow rate was 0.3 ml min⁻¹ and the column temperature was kept at 30 °C. The injection volume was 3 μl and the UV spectra were recorded at 215 nm. The MS operating conditions were as follows: positive ionization mode, scan range from m/z 100 to 500, drying gas (N₂) flow rate of 10 l min⁻¹, drying gas temperature of 350 °C, nebulizer pressure of 40 psi, capillary voltage of 4000 V.

2.4 Preparative HPLC conditions for the isolation of impurities

These two impurities were isolated using an Agilent 1200 series preparative HPLC system which was equipped with an automated fraction collector, a photodiode array detector and a Chemstation software. An Hypersil Prep. C₁₈ (100 × 50 mm, 10 μm) column was applied. Solvent A was water with 0.1% formic acid and solvent B was acetonitrile. The gradient program [time (min)/% solvent B] was set as 0/20, 30/50 and 60/50. The flow rate was set at 40 ml min⁻¹ and the injection volume was 500 μl. The impurities containing fractions were collected separately and were concentrated to powder using Rotavapor under high vacuum.

2.5 Nuclear magnetic resonance (NMR) spectrometer

NMR experiments were performed at ambient temperature (300 K) on a BRUKER AV-500 spectrometer, and were operated at 500 MHz. CDCl₃ was used as the solvent and tetramethylsilane (TMS) was used as the internal standard. The ¹H and ¹³C chemical shift values were reported on the δ scale in ppm, relative to TMS (δ = 0.00 ppm). For exchangeable proton experiments, 10% D₂O was added to provide a lock signal. Two-dimensional spectra (HMBC, HMQC and COSY) were recorded using standard impulse sequences provided by BRUKER.

2.6 Conditions for the determination of impurity B

Determination of impurity B was performed using Agilent 1200 SL series HPLC system interfaced with a quadrupole mass spectrometer equipped with an electrospray source (Agilent Technologies, CA, USA). The chromatographic condition was almost the same as mentioned in Section 2.3 except for using methanol–water with 0.1% acetic acid (80 : 20, v/v) as the mobile phase. Selective ion monitoring (SIM) mode was applied and the impurity B was monitored with its adduct ion [M + Na]⁺ m/z 276.1. The other MS operating conditions were the same as mentioned in Section 2.3.

2.7 Zebrafish embryos assays

Adult AB line zebrafish (Danio rerio) were obtained from a local aquarium and maintained by standard culture techniques. Fish were kept at temperature 28 ± 0.5 °C, pH 7.0 ± 0.5, a constant light cycle (14 h light/10 h dark) and fed a tropical fish commercial flake food twice a day. Embryos were obtained from healthy adult fish and cultured using a standard procedure.¹⁹

For the malformation test, each tested compound at different concentrations was incubated with 50% epiboly embryos for 48 h using water as the solvent. Standard culture solution without any toxicant was used as a control. The development status of zebrafish embryos were observed with an inverted microscope (IMT 2, Olympus, Tokyo, Japan) and documented photographically at the specified time.

For the comet assay, 10 zebrafish embryos after 48 h incubation with 2 μg ml⁻¹ impurity A or impurity B, were homogenized in 1 ml ice-cold potassium phosphate buffer, pH 7.4. The homogenate was centrifuged for 10 min at 4 °C and 200 × g. The pellet was resuspended in 1 ml phosphate buffer and again centrifuged 10 min at 4 °C and 180 × g. The pellet was resuspended in 100 μl phosphate buffer and mixed with 90 μl low melting agarose. The comet assay followed the procedures reported by the ref. 19. Comets were analysed by means of image analysis software Comet Assay IV, version 4.3 (Perceptive Instruments Ltd).

3. Results and discussion

3.1 Detection and structural elucidation of impurity A and impurity B

The rasagiline crude product was analyzed by HPLC-UV for the detection of possible impurities. As shown in Fig. 2, there were four main impurity peaks in the chromatogram except the rasagiline peak. Table 1 shows the accurate molecular weight and formulas of rasagiline and each impurity in positive ionization mode. By comparison of the mass spectra and retention time with those of authentic standard, the impurity peak at 2.5 min was identified as the 1-aminoindan and the impurity peak at 23.7 min was identified as the N-propargyl impurity. However, the mass spectral data of impurity A (9.0 min) and impurity B (16.7 min) did not match with the formula of any rasagiline related compound. Therefore, these two unknown impurities were isolated by preparative liquid chromatography for the further structural elucidation.

Fig. 2 Chromatogram of rasagiline and its impurities.

Table 1 Accurate molecular weight and formulas of rasagiline and each impurity

Peak	[M + H]^+ (m/z)	Error^a (ppm)	Formula
a Difference between the measured and calculated mass.
2.5 min	134.0968	−3.47	C9H11N
5.6 min	172.1127	−3.17	C12H13N
9.0 min	216.1026	−3.37	C13H13NO2
16.7 min	254.1180	−1.76	C16H15NO2
23.7 min	210.1282	−2.02	C15H15N

---

The molecular mass of impurity A was 44 units (CO₂) more than that of rasagiline. Compared with rasagiline, impurity A showed an additional ¹³C-NMR signal at 155.2 ppm corresponded to a carbonyl carbon, suggesting that the additive structure may be a carboxyl group or an ester group. However, no exchangeable proton was observed in the D₂O exchange experiment, indicating the absence of a carboxyl proton or other reactive protons in the structure of impurity A. Interestingly, the ¹H–¹H COSY spectra exhibited clear correlations between 5.05 ppm and 5.22 ppm (H₁), which indicated the presence of amide proton at 5.05 ppm in the structure of impurity A. A reasonable expectation of this non-exchangeable amide proton was that the formation of carbamate ester resulted in the low reaction activity of amide proton. The other evidences for the existence of a carbamate structure were that the ¹H and ¹³C chemical shift values of position 11, 12 and 13 of impurity A were higher than that of rasagiline, and no correlated proton of the amide proton was observed in the ¹H–¹H COSY spectra except the H₁. Two-dimensional spectra HMBC, HMQC and COSY, were used for the assignments in the Fig. 3. Based on the above data, the molecule formula of impurity A was confirmed as C₁₃H₁₃NO₂ (Fig. 2).

Fig. 3 ¹H and ¹³C-NMR of rasagiline impurity A.

The molecular mass of impurity B was 38 units (C₃H₂) higher than that of impurity A. Therefore, the structure of impurity B might contain another propargyl group when compared with the structure of impurity A. In the ¹H-NMR spectra of impurity B (Fig. 4), alkyne proton signals (2.15 ppm) and methylene proton (3.51 ppm, 4.01 ppm) were observed, while the amide proton was absent. This indicated that another propargyl group was connected with the N atom. Based on the above spectral data, impurity B was confirmed as C₁₆H₁₅NO₂ (Fig. 2).

Fig. 4 ¹H-NMR of rasagiline impurity B.

3.2 The formation reason of impurity A and impurity B

The formation reason of the carbamate structure in impurity A and impurity B was analyzed next. Previous reports have indicated that carbon dioxide readily reacts with nucleophiles.^22,23 For example, urea is industrially produced by a reaction with ammonia and secondary or primary amines react with CO₂ to give carbamic acids (Scheme 1).

Scheme 1

Rasagiline is currently available by multiple synthetic schemes and most of its synthetic routes encountered the same intermediate (R)-1-aminoindan.^24–26 Then (R)-1-aminoindan was used to react with different propargyl ester under various alkaline conditions, which provided a suitable environment for the existence of an amount of CO₂. Therefore, the side reaction resulting in the formation of impurity A and impurity B (Fig. 1) might occur at these conditions.

3.3 Genotoxicity evaluation of impurity A and impurity B

At present it is extremely difficult to experimentally prove the existence of a threshold for the genotoxicity of a given impurity.²⁷ In this study, the malformation rates of impurity A and impurity B with other impurities present in rasagiline were compared to differentiate the genotoxic impurities from the common impurities. Then the genotoxicity of impurity A and impurity B was tested using the comet assay.

In the malformation test, the digital images of embryonic malformations were used for the morphological assessment. All impurity-exposed embryos exhibited several morphological defects in a concentration-dependent manner and the ED₅₀ values (Fig. 5), determined using the probit method with the R software,²⁸ were used to represent the concentration of each compound that produces 50% malformation rate. As shown in Table 2, the ED₅₀ value of impurity B was lowest among the tested compound. The ED₅₀ value of impurity A was higher than those of rasagiline and (R)-1-aminoindan, but equal with that of N-propargyl impurity. In the comet assay, the percentage of DNA in the tail (% tail DNA) was used to represent the extent of DNA strand breaks.¹⁹ As shown in Fig. 6, the % tail DNA was significant higher in the group that was incubated with 2 μg ml⁻¹ impurity B than in the control group. But there was no significant difference in the % tail DNA between the control and the group that was incubated with 2 μg ml⁻¹ impurity A.

Fig. 5 The embryonic malformation rate after 48 h incubation with rasagiline or its impurities at different concentration (n = 3).

Table 2 The 50% of embryonic malformation concentration (ED₅₀) of rasagiline and its impurities

Compound	Rasagiline	(R)-1-Aminoindan	N-Propargyl impurity	Impurity A	Impurity B
EC50 (μg ml^−1)	280	167	31	33	2

---

Fig. 6 The percentage of DNA in the tail (% tail DNA) after 48 h incubation with 0.5% water (control), 2 μg ml⁻¹ impurity A or 2 μg ml⁻¹ impurity B (n = 3).

The malformation test and the comet assay indicated the highest DNA damage ability of impurity B among these impurities. Therefore, a MS based method was needed to differentiate impurity B from the other impurities present in rasagiline.

3.4 Development and validation of HPLC-MS method for the determination of impurity B

Different ions were assessed for quantitation, including molecular ion [M + H]⁺ and adduct ions [M + Na]⁺, [M + K]⁺. In addition, different mobile phase such as acetonitrile and methanol have been tested. Quantitation of impurity B was finally achieved by monitoring with adduct ion [M + Na]⁺ using methanol–water with 0.1% formic acid (80 : 20, v/v) as the mobile phase. Then the analytical performance of the proposed method was investigated with respect to specificity, linearity, precision, accuracy, limits of detection (LODs) and limits of quantitation (LOQs).

As shown in Fig. 7, the retention time of impurity B was 3.23 min and blank solvent (mobile phase) had no interference with it. Different standard concentrations of impurity B at 0.20–20 ng ml⁻¹ were prepared in mobile phase to evaluate the method linearity. Standard curve was constructed by plotting the peak area of impurity B (Y) against the impurity B concentration (X). The linearity correlation equation obtained was Y = 247.8 × X − 25.9, r = 0.998.

Fig. 7 Chromatograms of blank solvent (a) and impurity B (b).

The limit of detection (LOD) was obtained based on a signal-to-noise ratio (S/N) of 3, and the limit of quantification (LOQ) was obtained based on a S/N of 10. The determined LOD value was 0.07 ng ml⁻¹ and the LOQ value was 0.20 ng ml⁻¹, respectively.

The 0.20 ng ml⁻¹ and 20 ng ml⁻¹ impurity B solutions were separately prepared and injected for six times to evaluate the injection precision and % RSD values for peak responses of impurity B were separately calculated and it was found to be well within acceptance criteria of not more than 2.15%.

Solution stability was studied by injecting the 0.20 ng ml⁻¹ and 20 ng ml⁻¹ solutions at T = 0 h, 30 min, 1 h, 2 h, 4 h and 8 h and no significant difference in the area of impurity B solutions were observed within the 8 h (less than 2.03%).

Accuracy of the method was demonstrated though spiked recovery experiments. Authentic impurity B was spiked into 0.1 mg ml⁻¹ rasagiline mesylate in triplicates at three levels (2 ppm, 20 ppm and 200 ppm). The average recovery was 88.74% with RSD of 5.68% (n = 9).

4. Conclusion

Two carbamate impurities were detected and isolated during the synthesis of rasagiline, and their identity was established by the mass spectra and NMR. These two impurities may come from the intermediate (R)-1-aminoindan reacting with the carbon dioxide in the alkaline reaction solution. Because the carbamate structure has been highlighted as a class of potentially genotoxic impurities (GTIs), the genotoxicity of these two impurities was evaluated by the malformation test and the comet assay using zebrafish embryos. The results showed that the genotoxicity of impurity B was significant higher than those of rasagiline and other impurities. Thus, a HPLC-MS method was developed and validated to determine impurity B in rasagiline mesylate. This study will help the safety of this active pharmaceutical ingredients during its long-term clinical treatment.

Acknowledgements

This is a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions and the Open Project Program of MOE Laboratory of Drug Quality Control and Pharmacovigilance (No. DQCP2015MS04).

References

1. J. J. Chen and A. V. Ly, Rasagiline: a second-generation monoamine oxidase type-B inhibitor for the treatment of Parkinson's disease, Am. J. Health-Syst. Pharm., 2006, 63, 915–928.
2. S. Minguez-Minguez, J. Solis-Garcia del Pozo and J. Jordan, Rasagiline in Parkinson's disease: a review based on meta-analysis of clinical data, Pharmacol. Res., 2013, 74, 78–86.
3. A. H. Schapira, Monoamine oxidase B inhibitors for the treatment of Parkinson's disease: a review of symptomatic and potential disease: a review of symptomatic and potential disease-modifying effects, CNS Drugs, 2011, 12, 1067–1071.
4. Z. A. Abassi, O. Binah and M. B. Youdim, Cardiovascular activity of rasagiline, a selective and potent inhibitor of mitochondrial monoamine oxidase B: comparison with selegiline, Br. J. Pharmacol., 2004, 143, 371–378.
5. S. E. Lakhan, From a Parkinson's disease expert: rasagiline and the future of therapy, Mol. Neurodegener., 2007, 2, 13–15.
6. K. M. Tak, E. J. Park and M. H. Hyun, Liquid chromatographic resolution of racemic rasagiline and its analogues on a chiral stationary phase based on (+)-(18-crown-6)-2,3,11,12-tetracarboxylic acid, J. Sep. Sci., 2013, 36, 3682–3687.
7. P. S. Reddy, K. S. Babu and N. Kumar, A validated normal phase LC method for enantiomeric separation of rasagiline mesylate and its (S)-enantiomer on cellulose derivative-based chiral stationary phase, Chirality, 2013, 25, 324–327.
8. M. Fernández, E. Barcia and S. Negro, Development and validation of a reverse phase liquid chromatography method for the quantification of rasagiline mesylate in biodegradable PLGA microspheres, J. Pharm. Biomed. Anal., 2009, 49, 1185–1191.
9. P. G. Shelke and A. V. Chandewar, Study of stress degradation behavior of rasagiline mesylate under hydrolytic conditions by high performance liquid chromatography, Drug Res., 2014, 64, 182–185.
10. P. S. Reddy, K. Sudhakar Babu and N. Kumar, Development and validation of a stability-indicating RP-HPLC method for the simultaneous estimation of process related impurities and degradation products of rasagiline mesylate in pharmaceutical formulation, J. Chromatogr. Sci., 2013, 51, 242–249.
11. L. Müller, R. J. Mauthe and C. M. Riley, et al., A rationale for determining, testing, and controlling specific impurities in pharmaceuticals that possess potential for genotoxicity, Regul. Toxicol. Pharmacol., 2006, 44, 198–211.
12. US FDA, Guidance for Industry: Genotoxic and Carcinogenic Impurities in Drug Substances and Products, Recommended Approaches, 2008.
13. ICH, M7: Assessment and Control of DNA Reactive (mutagenic) Impurities in Pharmaceuticals to Limit Potential Carcinogenic Risk, 2013.
14. D. Q. Liu, M. Sun and A. S. Kord, Recent advances in trace analysis of pharmaceutical genotoxic impurities, J. Pharm. Biomed. Anal., 2010, 51, 999–1014.
15. N. V. V. S. S. Raman, A. V. S. S. Prasad and K. R. Reddy, Strategies for the identification, control and determination of genotoxic impurities in drug substances: a pharmaceutical industry perspective, J. Pharm. Biomed. Anal., 2011, 55, 662–667.
16. T. Guo, Y. Shi, L. Zheng, F. Feng, F. Zheng and W. Y. Liu, Rapid and simultaneous determination of sulfonate ester genotoxic impurities in drug substance by liquid chromatography coupled to tandem mass spectrometry: comparison of different ionization modes, J. Chromatogr. A, 2014, 1355, 73–79.
17. D. Kirkland, M. Aardema, L. Henderson and L. Müller, Evaluation of the ability of three in vitro genotoxicity tests to discriminate rodent carcinogens and non-carcinogens. I. Sensitivity, specificity and relative predictivity, Mutat. Res., 2005, 584, 1–257.
18. S. Ali, D. L. Champagne, H. P. Spaink and M. K. Richardson, Zebrafish embryos and larvae: a new generation of disease models and drug screens, Birth Defects Res., Part C, 2011, 93, 115–133.
19. T. O. Sogbanmu, E. Nagy, D. H. Phillips, V. M. Arlt, A. A. Otitoloju and N. R. Bury, Lagos lagoon sediment organic extracts and polycyclic aromatic hydrocarbons induce embryotoxic, teratogenic and genotoxic effects in Danio rerio (zebrafish) embryos, Environ. Sci. Pollut. Res. Int., 2016, 23, 14489–14501.
20. J. A. Adeyemi, A. da Cunha Martins-Junior and F. J. Barbosa, Teratogenicity, genotoxicity and oxidative stress in zebrafish embryos (Danio rerio) co-exposed to arsenic and atrazine, Comp. Biochem. Physiol., Part C: Toxicol. Pharmacol., 2015, 172–173, 7–12.
21. Y. Q. Sun, Y. M. Yan, X. H. Feng, W. Jiang, M. Y. Qian and Y. R. Tu, Process for preparation of rasagiline and its analogues, CN103804200A, 2014.
22. T. Sakakura, J. C. Choi and H. Yasuda, Transformation of carbon dioxide, Chem. Rev., 2007, 107, 2365–2387.
23. D. B. Dell'Amico, F. Calderazzo, L. Labella, F. Marchetti and G. Pampaloni, Converting carbon dioxide into carbamato derivatives, Chem. Rev., 2003, 103, 3857–3898.
24. M. B. H. Youdim, J. P. M. Finberg, R. Levy, J. Sterling, D. Lerner, T. Berger-Paskin and H. Yellin, R-Enantiomer of N-propargyl-1-aminoindan, its preparation and pharmaceutical compositions containing it, EP 0436492, 1994.
25. D. G. Sathe, S. V. Damle, K. D. Sawant, P. S. Gatne and T. A. Naik, Process for preparation of rasagiline and salts thereof, US 8741962, 2014.
26. J. Sterling, D. Lerner, H. Rosen, L. Bronov, D. Medini-Green, B. Iosefzon, T. Berger-Peskin, R. Lidor-Hadas and E. Bahar, Preparation of rasagiline mesylate and chloro derivative impurities, US 7572834, 2009.
27. H. Lee, Pharmaceutical Industry Practices on Genotoxic Impurities, CRC Press, 2014, pp. 18–23.
28. D. J. Finney, Probit analysis, Cambridge, University Press, 1952, p. 318.

---

Footnote

† These two authors contributed equally to this article.

---