Abstract
Objective
The aim of this study was to investigate the preliminary safety and efficacy of three doses of budiodarone in patients with paroxysmal atrial fibrillation.
Background
Budiodarone is a chemical analogue of amiodarone and shares its mixed ion channel electrophysiological properties. It has a shorter half-life than amiodarone.
Methods
Patients with paroxysmal atrial fibrillation and a previously implanted dual-chamber pacemaker capable of storing electrograms for at least 4 weeks were enrolled. Pacemaker memories were used to quantify atrial tachycardia/atrial fibrillation burden (AT/AFB). All antiarrhythmic drugs were stopped for greater than five half-lives and amiodarone greater than 3 months prior to enrollment. Following a 4-week baseline period to assess AT/AFB off antiarrhythmic drugs, patients with AT/AFB between 3% and 70% were blindly randomized to placebo, 200, 400, or 600 mg BID of budiodarone for 12 weeks followed by a 4-week washout period. Pacemakers were interrogated and safety assessed every 4 weeks. Pacemaker-derived electrograms were adjudicated blinded to treatment assignment. The primary study endpoint was percent change from baseline AT/AFB over 12 weeks of treatment compared to placebo.
Results
Of 72 randomized patients, 61 completed the study. The median reduction of AT/AFB for the 400 and 600 mg BID groups vs. placebo was 54% and 74% (p = 0.01 and 0.001), respectively. The budiodarone dose–response was statistically significant (p < 0.001). Number and duration of AT/AF episodes were reduced.
Conclusions
In this preliminary study, budiodarone at both higher doses significantly reduced AT/AFB. The study is novel because dual-chamber pacemakers, previously placed for standard clinical indications, were successfully used to monitor AT/AFB.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
Clinical trials comparing rhythm and rate-control strategies in patients with atrial fibrillation have not found significant differences in morbidity and mortality [1–6]. However, for patients who are symptomatic, rhythm control continues to be preferred [7]. Currently, amiodarone is the most effective antiarrhythmic drug for maintenance of sinus rhythm [8, 9]. It is widely used in patients with atrial fibrillation. Unfortunately, amiodarone has significant toxicities that hamper its use. Discontinuation rates due to unwanted effects are as high as 18% within the first year of use [10]. Drug toxicity and discontinuation rates may have masked benefits of the sinus rhythm strategy [11]. Thus, there is an urgent need for a new effective and safe antiarrhythmic drug for the treatment of patients with atrial fibrillation.
Budiodarone was developed to capitalize on the proven efficacy of amiodarone and to avoid its side effects. It has a short plasma half-life (7 h) and a lower volume of distribution (13 L/kg). It is cleared from the body in 48 h. Like amiodarone, budiodarone has balanced, multiple cardiac ion channel inhibiting activity, giving it properties of all four Vaughan Williams antiarrhythmic drug classes [12]. Budiodarone, unlike amiodarone, undergoes rapid metabolism by plasma and tissue esterases [13]. It is thus expected to be less susceptible to drug–drug interactions with drugs that inhibit CYP450 mediated metabolism. After oral administration, it has a rapid onset of action and steady state is reached within 2 to 3 days.
The present study was designed to provide a preliminary indication of the efficacy and safety of budiodarone at three doses in patients with symptomatic paroxysmal atrial fibrillation. The study design is unique. This is the first controlled trial to use implanted permanent pacemakers capable of continuous storage of rhythm and rate for assessment of an antiarrhythmic medication. The diagnostic capabilities of these pacemakers allow precise evaluation of atrial tachycardia/atrial fibrillation (AT/AF) burden [14–17], permitting the collection of large quantities of data from individual patients not possible using non-continuous monitoring.
1 Methods
PASCAL was a randomized, double-blind, placebo-controlled, parallel-group study conducted at 44 centers in the USA, Canada, Poland, and Germany during November 2006 to May 2008. The study was conducted in accordance with the Declaration of Helsinki. Institutional Review Boards or Research Ethics Boards approved the study at each center. Written informed consent was obtained from each patient. The trial was sponsored by ARYx Therapeutics. It was designed with the assistance and advice of a steering committee. A Pacemaker Committee provided advice on the programming of dual-chamber pacemakers to optimize the detection and logging of AT/AF episodes. The data was analyzed by an independent statistician.
1.1 Study population
Patients with paroxysmal atrial fibrillation documented by electrocardiogram, Holter monitor, or pacemaker electrogram and who had a previously implanted dual-chamber pacemaker for standard clinical indications that was capable of detecting and logging AT/AF episodes and storing atrial electrograms were eligible if they also met the following criteria: an AT/AF burden between 3% and 70% during the baseline period, no evidence of persistent atrial fibrillation, and able to have pacemaker antitachyarrhythmia algorithms turned off for the duration of the trial. Exclusion criteria included: amiodarone treatment within 3 months of screening or a history of amiodarone toxicity, catheter ablation within 3 months or cardioversion within 1 month of screening, a left ventricular ejection fraction <25% or New York Heart Association class III or IV heart failure, unstable angina, myocardial infarction, percutaneous transluminal coronary angioplasty or coronary artery bypass surgery within 3 months of screening, or any history of sustained ventricular tachycardia. Concomitant treatment with other antiarrhythmic drugs was prohibited, but use of rate-control therapy was permitted.
1.2 Study procedures
After informed consent, pacemaker programming parameters, related to the detection and logging of AT/AF episodes were optimized. The pacemaker core laboratory, which was blinded to treatment assignment, provided technical training to all participating centers to standardize the programming of the pacemakers. The lower rate of sensing/tracking was set at 70 beats per minute and the upper rate at 80% of the maximum age-predicted heart rate. Investigators were permitted to set rates that were optimal for patient care. Atrial sensitivity was set to minimize under- and over-sensing of true atrial events, with optimization of the signal-to-noise ratio. Once set, atrial sensitivity was not changed for the duration of the study period (baseline, treatment, and washout phases). Inappropriate sensing or inadvertent modifications of these settings were communicated by the pacemaker core lab to the participating centers, where the devices were reprogrammed. Atrial rates above 200 beats per minute were defined as episodes of AT/AF. The AT/AF burden was calculated as the percentage time in AT/AF during the total observation period.
AT/AF burden was recorded during a 4-week baseline period. Eligible patients were then centrally randomized to placebo, 200, 400, or 600 mg BID on a 1:1:1:1 basis. The total treatment duration was 3 months, followed by a 4-week washout period. Pacemakers were interrogated at monthly visits. Reports of AT/AF episodes were downloaded and adjudicated at the pacemaker core laboratory blinded to treatment assignment. Safety evaluations at the monthly visits included adverse events, vital signs, a standard 12-lead electrocardiogram, and clinical laboratory tests (chemistry, hematology, urinalysis, and thyroid function tests). Male patients were also tested monthly for inhibin B, follicle-stimulating hormone, luteinizing hormone, and testosterone levels as earlier dog toxicology studies showed reversible inhibition of spermatogenesis. Pulmonary function tests, including carbon monoxide diffusion capacity, were performed at screening and repeated if symptoms suggestive of pulmonary toxicity developed during treatment. Chest X-ray examinations were performed at screening and at the end of treatment. Patients were contacted by telephone midway between each clinic visit to inquire about adverse events. A slit lamp examination was performed at screening for patients who had been on amiodarone within 2 years of study enrollment and at the end of therapy for all patients to assess for corneal microcrystalline deposits. In anticoagulated patients, the international normalized ratio was monitored at weekly intervals for the first 4 weeks of budiodarone therapy and then at least monthly thereafter. Digoxin plasma concentrations were measured monthly if the patient was on digoxin.
1.3 Endpoints
The primary endpoint was the percent change in AT/AF burden over the 12-week treatment period compared to the baseline AT/AF burden. Secondary endpoints were mean duration of AT/AF episodes, mean number of AT/AF episodes, and mean duration of normal sinus rhythm. Part C of the University of Toronto AFSS was assessed at each visit [18]. The AFSS includes seven common symptoms of atrial fibrillation rated on a scale of 0 to 5 where 0 was “no symptoms in the past 4 weeks” and 5 was “the patient has been bothered a great deal by symptoms.” The questionnaire was administered at screening, baseline and the end of the baseline period, monthly during treatment, and at the end of the washout period. The pretreatment score is the mean of the screening and baseline scores. The on-treatment score is the mean of the three monthly on-treatment scores. The washout score is the score at 1 month after discontinuing treatment. A patient global clinical impression questionnaire, similar to the patient global assessment used as a primary endpoint in the Ferinject assessment in patients with iron deficiency and chronic heart failure trial, was given at trial termination [19]. At the end of each follow-up visit (every 4 weeks), patients were asked two questions: “How well did the test medication control your atrial fibrillation symptoms?” and “How satisfied are you with the test medication?” A Likert scale was used to record the patients’ responses.
1.4 Statistical analysis
The statistical analysis was based on the intent-to-treat population, which included all randomized patients who had at least one assessment of AT/AF burden during the treatment period. For analysis of the primary endpoint, the Wilcoxon rank-sum test was used for pairwise comparison between each budiodarone treatment group and the placebo group. Dose response was assessed with the Jonckheere–Terpstra test. The secondary endpoints were analyzed using the same statistical tests as for the primary endpoint. Median values are presented to minimize the effect of extreme values in the pacemaker data. The patient global clinical impression questionnaire was analyzed using the Cochran–Mantel–Haenszel mean score test comparing each budiodarone dose group to placebo. A two-sided p value < 0.05 was considered statistically significant.
1.5 Role of the funding source
The sponsor (ARYx Therapeutics) participated in the study design together with the academic authors. The sponsor collected the trial data, which were statistically analyzed by a contract research organization (Synteract). The data was interpreted by the steering committee.
2 Results
2.1 Patient population
Of 167 consented patients, 57 were screen failures mostly due to insufficient AT/AF burden or because of withdrawal of consent (Fig. 1). Of the 110 patients enrolled, 38 were excluded after the baseline period because of insufficient AT/AF burden or persistent atrial fibrillation. After randomization, 11 patients discontinued prematurely and 61 completed the study. Of the 11 patients discontinuing early, 3 were for adverse events, 3 were by patient request, 3 were discontinued by the investigator or sponsor, and 2 were discontinued because they did not meet enrollment criteria. The baseline characteristics are described in Table 1. The majority of patients had failed strategies to maintain sinus rhythm with either cardioversion, prior ablation, or attempts at rhythm control with other antiarrhythmic drugs.
2.2 Effects of study drug
There was a dose-dependent decrease in AT/AF burden with increasing dosages of budiodarone (p < 0.001) (Fig. 2(a) and Table 2). Administration of 400 and 600 mg BID budiodarone resulted in a decrease in AT/AF burden of 54% (p = 0.01 vs. placebo) and 74% (p = 0.001 vs. placebo), respectively. This decrease was evident within the first month of treatment for both groups and persisted for the 3-month treatment phase (Fig. 2(b)). AT/AF burden is a summation of the total number of episodes and their duration. Pacemaker interrogation allows for the independent assessment of each of these components. Analysis of the median percent changes from baseline showed that reductions in duration of AT/AF episodes (Fig. 3(c)) and in the number of episodes (Fig. 3(a)) both contributed to the net reduction in AT/AF burden. These reductions in duration and number of AT/AF episodes led to a reciprocal increase from baseline in the duration of periods in sinus rhythm (Fig. 3(b)). The sample size was too small to perform a gender-based subgroup analysis. However, no consistent gender-based differences in the efficacy parameters were noted (Table 2).
2.3 Safety evaluations
There were no respiratory symptoms suggestive of pulmonary toxicity. There were no deaths on treatment. Serum creatinine showed a reversible and non-progressive elevation of 12% to 21% compared to baseline. Serum-free T3 concentrations decreased between 27% and 35% compared to baseline. Thyroid-stimulating hormone concentrations increased by 65% to 109%, and the only change in free T4 values was an increase of 11.8% in the 600 mg BID group. There was no evidence of clinically overt hyperthyroidism. In all male patients, there were no significant changes in testosterone, luteinizing hormone, follicle-stimulating hormone, or inhibin B hormone levels. No corneal microdeposits in budiodarone-treated patients were detected by slit lamp examination at the end of treatment. There were no end-of-study chest X-ray findings suggestive of drug-induced pneumonitis. No prolongation of the QTc interval was seen in budiodarone-treated patients during periods of sinus rhythm.
2.4 Quality of life assessment
The on-treatment individual AFSS scores were generally low, reflecting low symptom burden (Table 3). There was a positive effect on the overall AFSS score.
3 Discussion
Modern pacemakers are capable of accurately recording and storing cardiac rate and rhythm information over long periods of time, allowing for a more objective assessment of arrhythmia burden. Only patients with dual-chamber pacemakers were included in this study, as an atrial lead is necessary to detect atrial activity.
The first major conclusion of this study is that it is possible to use pacemakers to assess the effects of an antiarrhythmic compound in patients with paroxysmal atrial fibrillation. This methodology is better than less continuous monitoring [20].
The second finding was the demonstration of a significant, dose-dependent decrease in atrial fibrillation burden in patients assigned to budiodarone, with the maximum effect in the 400 and 600 mg BID dose groups, respectively. All doses evaluated were well-tolerated in this short-term study.
In the management of atrial fibrillation, drugs simultaneously targeting multiple electrophysiological mechanisms hold promise [21]. Multimechanistic drugs such as amiodarone [8, 9] and dronedarone [22] are effective in patients with atrial fibrillation. However, amiodarone is not FDA-approved for the management of atrial fibrillation in the USA and its use is hampered by its extracardiac toxicity and complex pharmacokinetic/dynamic profile. Dronedarone, while well-tolerated except in patients with advanced heart failure [23], lacks a dose–response relationship and is generally thought not to be as effective as amiodarone.
Budiodarone targets multiple ion channels [12]. It is eliminated from the body within 48 h following oral administration. In this short-term study, there was no evidence of the toxic effects attributable to amiodarone. Unlike amiodarone [24], slit lamp examinations revealed no evidence of corneal microdeposits after 3 months of therapy with budiodarone. Budiodarone does affect thyroid function with a biochemical profile of hypothyroidism. Hyperthyroidism may be possible. Close monitoring of thyroid function is needed in a phase III trial.
Physiologic pacing (DDDR) has been associated with a reduced risk of atrial fibrillation [25–27]. Since only patients with dual-chamber pacemakers were included in this study, it is important to recognize that the baseline atrial fibrillation burden and symptoms might be lower in these patients than in those without pacemakers, increasing the validity of the outcome. The small sample size and short follow-up duration are limitations. However, the comprehensive nature of the endpoint allowed a statistically significant result with fewer patients.
This study and the recently published, six-patient study of budiodarone [28] are the only studies using continuous monitoring to evaluate an antiarrhythmic drug. Past trials have relied on time to first atrial fibrillation recurrence after cardioversion [1, 2, 8, 9], providing a less comprehensive evaluation of drug efficacy.
Abbreviations
- AT/AFB:
-
Atrial tachycardia/atrial fibrillation burden
- AFSS:
-
Atrial Fibrillation Severity Scale
References
Opolski, G., Torbicki, A., Kosior, D. A., Szulc, M., Wozakowska-Capton, B., Kolodziej, P., et al. (2004). Rate control vs rhythm control in patients with nonvalvular persistent atrial fibrillation: the results of the Polish How to Treat Chronic Atrial Fibrillation (HOT CAFE) Study. Chest, 126(2), 476–486.
Roy, D., Talajic, M., Nattel, S., Wyse, D. G., Dorian, P., Lee, K. L., et al. (2008). Rhythm control versus rate control for atrial fibrillation and heart failure. The New England Journal of Medicine, 358(25), 2667–2677.
The Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Investigators. (2002). A comparison of rate control and rhythm control in patients with atrial fibrillation. The New England Journal of Medicine, 347, 1825–1833.
Van Gelder, I. C., Hagens, V. E., Bosker, H. A., Kingma, J. H., Kamp, O., Kingma, T., et al. (2002). A comparison of rate control and rhythm control in patients with recurrent persistent atrial fibrillation. The New England Journal of Medicine, 347(23), 1834–1840.
Hohnloser, S. H., Kuck, K. H., & Lilienthal, J. (2000). Rhythm or rate control in atrial fibrillation: pharmacological intervention in atrial fibrillation (PIAF): a randomized trial. The Lancet, 356, 1789–1794.
Carlsson, J., Miketic, S., Windeler, J., Cuneo, A., Haun, S., Micus, S., et al. (2003). Randomized trial of rate-control versus rhythm-control in persistent atrial fibrillation: the Strategies of Treatment of Atrial Fibrillation (STAF) study. Journal of the American College of Cardiology, 41(10), 1690–1696.
Nieuwlaat, R., Capucci, A., Camm, J., Olsson, B., Andresen, D., Davies, D. W., et al. (2005). Atrial fibrillation management: a prospective survey in ESC member countries: the Euro Heart Survey on Atrial Fibrillation. European Heart Journal, 26, 2422–2434.
Roy, D., Talajic, M., Thibault, B., Connolly, S., Eisenberg, M. J., Green, M., et al. (1997). Pilot study and protocol of the Canadian Trial of Atrial Fibrillation (CTAF). The American Journal of Cardiology, 80, 464–468.
Singh, B. N., Singh, S. N., Reda, D. J., Tang, C., Lopez, B., Harris, C. L., et al. (2005). Amiodarone versus sotalol for atrial fibrillation. The New England Journal of Medicine, 352, 1861–1872.
Zimetbaum, P. (2007). Amiodarone for atrial fibrillation. The New England Journal of Medicine, 356, 935–941.
Steinberg, J. S., Sadaniantz, A., Kron, J., Denny, D. M., Daubert, J., Campbell, W. B., et al. (2004). Analysis of cause-specific mortality in the Atrial Fibrillation Follow-up Investigation of Rhythm Management (AFFIRM) Study. Circulation, 109(16), 1973–1980.
Raatikainen, M. J., Morey, T. E., Druzgala, P., Milner, P., Gonzalez, M. D., & Dennis, D. M. (2000). Potent and reversible effects of ATI-2001 on atrial and atrioventricular nodal electrophysiological properties in guinea pig isolated perfused heart. The Journal of Pharmacology and Experimental Therapeutics, 295, 779–785.
Juhasz, A., & Bodor, N. (2000). Cardiovascular studies on different classes of soft drugs. Pharmazie, 55(3), 228–238.
Israel, C. W., Groenefeld, G., Ehrlich, J. R., Li, Y. G., & Hohnloser, S. H. (2004). Long-term risk of recurrent atrial fibrillation as documented by an implantable monitoring device. Journal of the American College of Cardiology, 43, 47–52.
Capucci, A., Santini, M., Padeletti, L., Gulizia, M., Botto, G., Ricci, R., et al. (2005). Monitored atrial fibrillation duration predicts arterial embolic events in patients suffering from bradycardia and atrial fibrillation implanted with antitachycardia pacemakers. Journal of the American College of Cardiology, 46(10), 1913–1920.
Gillis, A. M., & Rose, M. S. (2000). Temporal patterns of paroxysmal atrial fibrillation following DDDR pacemaker implantation. The American Journal of Cardiology, 85, 1445–1450.
Glotzer, T. V., Hellkamp, A. S., Zimmerman, J., MOST Investigators, et al. (2003). Atrial high rate episodes detected by pacemaker diagnostics predict death and stroke: report of the Atrial Diagnostics Ancillary Study of the Mode Selection Trial (MOST). Circulation, 107, 1614–1619.
Dorian, P., Jung, W., Newman, D., Paquette, M., Wood, K., Ayers, G. M., et al. (2000). The impairment of health-related quality of life in patients with intermittent atrial fibrillation: implications for the assessment of investigational therapy. Journal of the American College of Cardiology, 36, 1303–1309.
Anker, S. D., Comin Colet, J., Filippatos, G., Willenheimer, R., Dickstein, K., Drexler, H., et al. (2009). Ferric carboxymaltose in patients with heart failure and iron deficiency. The New England Journal of Medicine, 361, 2436–2448.
Warman, E. N., Grammatico, A., & Padeletti, L. (2004). Sample size estimates for atrial fibrillation endpoints. Heart Rhythm, 1, B58–B63.
Ezekowitz, M. D. (2007). Maintaining sinus rhythm: making treatment better than disease. The New England Journal of Medicine, 357(10), 1039–1041.
Hohnloser, S. H., Crijns, H. J., Van Eickels, M., Gaudin, C., Page, R. L., Torp-Pedersen, C., et al. (2009). Effect of dronedarone on cardiovascular events in atrial fibrillation. The New England Journal of Medicine, 360, 668–678.
Køber, L., Torp-Pedersen, C., McMurray, J. J., Gotzsche, O., Levy, S., Crijns, H., et al. (2008). Increased mortality after dronedarone therapy for severe heart failure. The New England Journal of Medicine, 358(25), 2678–2687.
Greene, H. L., Graham, E. L., Werner, J. A., Sears, G. K., Gross, B. W., Gorham, J. P., et al. (1983). Toxic and therapeutic effects of amiodarone in the treatment of cardiac arrhythmias. Journal of the American College of Cardiology, 2, 4–28.
Andersen, H. R., Thuesen, L., Baggar, J. P., Vesterlund, T., & Thomsen, P. E. (1994). Prospective randomised trial of atrial versus ventricular pacing in sick-sinus syndrome. The Lancet, 344(8936), 1523–1528.
Lamas, G. A., Lee, K. L., Sweeney, M. O., Silverman, R., Leon, A., Yee, R., et al. (2002). Ventricular pacing or dual-chamber pacing for sinus-node dysfunction. The New England Journal of Medicine, 346(24), 1854–1862.
Stambler, B. S., Ellenbogen, K. A., Orav, E. J., Sgarbossa, E. B., Estes, N. A., Rizo-Patron, C., et al. (2003). Predictors and clinical impact of atrial fibrillation after pacemaker implantation in elderly patients treated with dual chamber versus ventricular pacing. Pacing and Clinical Electrophysiology, 26(10), 2000–2007.
Arya, A., Silberbauer, J., Teichman, S. L., Milner, P., Sulke, N., & Camm, A. J. (2009). A preliminary assessment of the effects of ATI-2042 in subjects with paroxysmal atrial fibrillation using implanted pacemaker methodology. Europace, 11(4), 402–404.
Acknowledgments
Members of the Steering Committee are P. Dorian, M. Ezekowitz, A. Gillis, B. Thibault. Members of the Pacemaker Committee are A. Gillis, R. Yee, S. Saksena. M. Knowles is the independent statistician. Investigators who enrolled patients were: B. Coutu, Y.K. Chan, P. Costi, E. Crystal, P. Dorian, L. Sterns, A. Thang, D. Birnie, B. Thibault, M. Sturmer, S. Lepage, F. Ayala Paredes, J. Heath in Canada; A. Bauer, M. Borggrefe, S. Hohnloser, L. Lickfett, R. Ventura, J. Brachmann, A. Schuchert in Germany; Z. Kalarus, W. Kargul, K. Kawecka-Jaszcz, J. Kubica, A. Lubinski, P. Mitkowski, G. Opolski, G. Raczak, Z. Kornacewicz-Jach in Poland; M. Ezekowitz, T. Guarnieri, J. Ip, B. Strunk, A. Tran, A. Murphy, P. Underwood, S. Goel, S. Krueger, A. Cohen, D. Lustgarten, I. Khan, B. Pert, S. Mazer, P. Coffeen in the USA.
Funding sources
The study was funded by a grant from ARYx Therapeutics.
Disclosures
M. Ezekowitz received consulting fees and grant support from Boehringer Ingelheim, ARYx Therapeutics and Portola and consulting fees from Wyeth, Sanofi-Aventis, Bristol-Myers Squibb, Astra Zeneca, Daiichi Sankyo, and Medtronic. S. Hohnloser received consulting and lecture fees from Sanofi-Aventis, St. Jude Medical, Boehringer Ingelheim, and Procter & Gamble; consulting fees from Cardiome, ARYx Therapeutics and Bristol-Myers Squibb; research grants from Sanofi-Aventis and St. Jude Medical; and lecture fees from Boehringer Ingelheim. B. Thibault received consulting and lecture fees from St. Jude Medical, Sorin and Medtronic; research grants from St. Jude Medical and Sorin; and honoraria from ARYx Therapeutics. R. Nagarakanti and A. Lubinski have no disclosures. All other authors are employees of ARYx Therapeutics.
Author information
Authors and Affiliations
Consortia
Corresponding author
Rights and permissions
About this article
Cite this article
Ezekowitz, M.D., Nagarakanti, R., Lubinski, A. et al. A randomized trial of budiodarone in paroxysmal atrial fibrillation. J Interv Card Electrophysiol 34, 1–9 (2012). https://doi.org/10.1007/s10840-011-9636-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10840-011-9636-3