Diabetes mellitus affects 387 million people around the world; in Brazil, 9 million people above 18 years of age have this disease, and 10% of them have type 1 diabetes mellitus. Autonomic diabetic neuropathy is one of the most common complications of type 1 diabetes and generally impairs the cardiovascular system, resulting in autonomic neuropathy,Reference Iser, Stopa and Chueiri 1 significantly influencing the mortality rate of diabetic patients.Reference Traon, Fontaine, Tap, Guidolin, Senard and Hanaire 2 – Reference Rolim, Sá, Chacra and Dib 4
In this context, we highlight the importance of studies that assess the dynamics of the autonomic nervous system, providing new elements on how neuropathies are developed and how they can change the sympathetic and parasympathetic systems.Reference Pop-Busui, Evans and Gerstein 3
A method that is widely used to assess cardiac autonomic modulation is the analysis of heart rate variability.Reference Villegas, Espinosa, Moreno, Echeverry and Rodrigues 5 Heart rate variability describes the oscillations of the intervals between consecutive heart beats (R–R intervals)Reference Vanderlei, Silva, Pastre, Azevedo and Godoy 6 – Reference Vanderlei, Rossi and Souza 8 and it provides information on the diagnosis and prognosis of several diseases.Reference Ferreira, Souza, Bernardo, Vitor, Valenti and Vanderlei 9 Analysis of heart rate variability can be performed by linear methods, which include the indices in the time and frequency domains, and by nonlinear methods based on chaos theory.Reference Vanderlei, Pastre, Hoshi, Carvalho and Godoy 7
Studies using heart rate variability that investigated type 1 diabetes mellitus by linear methods showed that this population may exhibit reduced modulation of the autonomic nervous system on both systemsReference Traon, Fontaine, Tap, Guidolin, Senard and Hanaire 2 , Reference Chen, Lee, Chiu and Jeng 10 , Reference Seyd, Joseph and Jacob 11 or only one system.Reference Lucini, Zuccotti and Malacarne 12 , Reference Javorka, Trunkvalterova, Tonhajzerova, Javorkova, Javorka and Baumert 13 Studies using nonlinear methods to investigate type 1 diabetes mellitus are scarce and have provided inconclusive results, indicating loss of complexity of the autonomic nervous system only in some indices.Reference Javorka, Trunkvalterova, Tonhajzerova, Javorkova, Javorka and Baumert 13 – Reference Vitor, Souza and Lorenconi 16
It is known that the nonlinear behaviour is prevalent in the human systems, and the study of heart rate variability by nonlinear methods has gained increasing interest, as the recurrence plot.Reference Javorka, Trunkvalterova, Tonhajzerova, Lazarova, Javorkova and Javorka 17
The recurrence plot is a graphical representation of recurrence in a dynamic system and it evaluates the complexity of a system.Reference Marwan and Kurths 18 , Reference Ferreira, Messias, Vanderlei and Pastre 19 The plot allows a qualitative analysis through vertical, horizontal, and diagonal variables and it also allows a quantitative analysis by means of its indices,Reference Marwan and Kurths 18 thus enabling a more sensitive analysis of cardiac autonomic regulation.
In this sense, studying the nonlinear behaviour, which is more representative of the physiological behaviour of individuals, can lead to a better understanding of the influence of the autonomic nervous system on type 1 diabetes mellitus subjects and assist in the assessment and risk stratification of these patients. Therefore, better and specific treatment plans can be designed depending on autonomic nervous system action.
Thus, we aimed to evaluate the cardiac autonomic modulation of type 1 diabetes mellitus subjects at rest through qualitative and quantitative analyses of the recurrence plot.
Materials and methods
Study population
We analysed data from 43 young adult volunteers of both sexes who were divided into two groups: control and type 1 diabetes mellitus groups. We chose type I diabetes patients for this study because of the lower prevalence of co-morbidities associated with type 1 diabetes mellitus compared with type 2 diabetes mellitus, because co-morbidities of type 2 diabetes mellitus could influence autonomic nervous system modulation.
The type 1 diabetes mellitus group was composed of 20 volunteers diagnosed with type 1 diabetes mellitus, the time of diseases was 9.8±4.97 years, and the control group consisted of 23 healthy volunteers. Table 1 contains the characterisation of the groups studied.
Mean values, followed by their respective standard deviations of physical and clinical characteristics of the studied volunteer groups.
BMI=body mass index; bpm=beats per minute; DBP=diastolic blood pressure; DM1=diabetes mellitus type 1; HR=heart rate; kg=kilograms; m=metres; min=minute; mmHg=millimetres of mercury; PA=physical activity; SBP=systolic blood pressure; WHR=waist/hip circumference
Mean±standard deviation [minimum–maximum]
The inclusion criteria for selection were as follows: individuals who did not use drugs that influence cardiac autonomic regulation – such as β-blockers and adrenergic agonists – those who did not have cardio-respiratory diseases, and those who were non-smokers and did not consume alcohol. We excluded from the study volunteers who did not comply with the recommendations for performing the experimental procedure and those who had a series with more than 95% sinus rhythm.
All the study procedures were approved by the Ethics Committee in Research of our Institution (47/2011), and we followed the rules established by the 466/2012 Resolution of The National Health Council. The volunteers were fully informed about the procedures and objectives of this study, and all of them signed a consent form.
Procedures
Data were collected in our sound-proof laboratory. The temperature (23.87±2.54°C) and humidity (54.22±8.51%) were recorded by a digital thermo-hygrometer (Incoterm, Rio Grande do Sul, Brazil).
Initially, volunteers answered a questionnaire to identify and obtain information on age, signs and symptoms of type 1 diabetes mellitus, drug use, presence of associated diseases, evaluation of physical activity time performed in a week, measured through the IPAQ,Reference Hallal, Gomez and Parra 20 and the age of disease onset among the volunteers. We performed physical assessments – anthropometric measurement, heart rate and blood pressure at rest – and subsequently cardiac autonomic regulation was evaluated. All the procedures were conducted by a group of trained researchers.
Physical assessment
We measured the weight and height of the participants to calculate body mass index, and we measured waist circumference and hip circumference to obtain the waist/hip ratio.
Body mass was measured using a digital scale (Welmy R/I 200, Brazil) and height was measured using a stadiometer (Sanny, São Paulo, Brazil). Using the weight and height scores, body mass index was calculated using the formula weight (kg)/height (metres) squared. 21
To obtain the waist/hip ratio, we used a measuring tape (Sanny) for the measurement of waist circumference – which was obtained by measuring the smallest circumference between the lower costal margin and the anterior superior iliac crest – and hip circumference was obtained by measuring the diameter at the trochanters. To obtain the waist/hip ratio, the value of waist circumference was divided by the value of hip circumference.Reference Ferreira, Valente, Gonçalves-Silva and Sichieri 22
Blood pressure was measured indirectly, using the stethoscope (Littman, Saint Paul, Minnesota, United States of America) and aneroid sphygmomanometer (WelchAllyn – Tycos, New York, New York, United States of America), on the left arm of the volunteer following the criteria established by the VI Brazilian Guidelines on Hypertension blood. 23
Heart rate was measured by palpation of the brachial artery for one minute.
Autonomic assessment
Volunteers were instructed to avoid consuming alcohol, caffeine and/or autonomic nervous system stimulants such as coffee, tea, and chocolateReference Vanderlei, Rossi and Souza 8 for 24 hours before evaluation. Data were collected between 1 p.m. and 6 p.m. All the procedures necessary for data collection were explained to the individuals, and the subjects were instructed to remain at rest and not to talk during the data collection process.
After all the procedures of data collection were explained to the volunteers, a strap that captures the heart electrical activity was placed on the sternum and the Polar S810i monitor was placed on the wrist (Polar Electro, Kempele, Finland).Reference Vanderlei, Silva, Pastre, Azevedo and Godoy 6 , Reference Gamelin, Berthoin and Bosquet 24 The participants were then placed supine on a stretcher where they remained at rest for 30 minutes with spontaneous breathing.
The R–R intervals recorded by the portable validated heart rate monitor Polar S810iReference Rolim, Sá, Chacra and Dib 4 , Reference Villegas, Espinosa, Moreno, Echeverry and Rodrigues 5 – with a sampling rate of 1000 Hz – were uploaded to the Polar Precision Performance programme (v. 3.0; Polar Electro). The software enabled the visualisation of heart rate and the extraction of a cardiac period (R–R interval) file in downloadable “.txt” format. Following digital filtering complemented with manual filtering for the elimination of premature ectopic beats and artefacts,Reference Godoy, Takakura and Correa 25 1000 R–R intervals were used for the data analysis. Only series with more than 95% sinus rhythm was included in the study.
For calculation of the indices, we used heart rate variability Analysis software (Kubios HRV v.1.1 for Windows; Biomedical Signal Analysis Group, Department of Applied Physics, University of Kuopio, Kuopio, Finland), whereas the nonlinear indices derived from the recurrence plot were obtained by Visual Recurrence Analysis – VRA (Visual Recurrence Analysis Software, by Eugene Kononov, United States of America) software.
Analysis of heart rate variability indices
All the following analyses of heart rate variability were conducted by a single researcher, and therefore all of quantitative and qualitative indices.
Linear methods
To analyse heart rate variability in the frequency domain, the low frequency (LF=0.04–0.15 Hz, sympathetic component) and high frequency (HF=0.15–0.40 Hz, parasympathetic component) spectral components were used in absolute (ms2) and normalised units (nu), which represent a value relative to each spectral component in relation to the total power minus the very low frequency components, and the ratio between these components (LF/HF). The spectral analysis was calculated using the Fast Fourier Transform algorithm.Reference Vanderlei, Pastre, Hoshi, Carvalho and Godoy 7
The analysis in the time domain was performed by means of SDNN (standard deviation of normal-to-normal R–R intervals; overall variability of heart rate) and RMSSD (root-mean square of differences between adjacent normal R–R intervals in a time interval, parasympathetic modulation).Reference Vanderlei, Pastre, Hoshi, Carvalho and Godoy 7
Nonlinear methods
The nonlinear indices evaluated were derived from the recurrence plot. The following indices were obtained: recurrence rate (REC), determinism (DET), and Shannon entropy (ES), as well as the qualitative analysis of the recurrence plot.
The recurrence plot is the visualisation of a square matrix in which the elements of the matrix correspond to the moments in which a state of a dynamic system repeats – columns and rows correspond to a particular couple of times. The recurrence plot reveals the moments when the trajectory in the phase space of the dynamical system returns approximately to the same area in the phase space,Reference Souza 26 and it may be quantitatively and qualitatively analysed.
Qualitative analysis was carried out by visualisation of the vertical, horizontal, and diagonal lines of the recurrence plot. The diagonal lines reflect the repeating sequence of states in the system, and the horizontal and vertical lines result from a persistent state during a time interval.Reference Souza 26 In healthy individuals, the recurrence plot has a diagonal square and is less apparent, indicating higher heart rate variability, whereas in subjects with cardiac abnormalities the recurrence plot presents more apparent squares, indicating the inherent frequency and low heart rate variability related to an autonomic nervous system recurrence. Furthermore, if the recurrence plot is randomness it will show no structure and an uniform distribution, and there will be no pattern identifiable in the plot. Thus this plot is related to a autonomic nervous system with randomness trend.Reference Ferreira, Messias, Vanderlei and Pastre 19
Quantitative analysis of the recurrence plot enables the extraction of three indices: REC, DET, and ES. The REC is the ratio of all recurrent states (return points) for all possible states, and therefore the probability of recurrence of a particular state; it takes into account the ratio of ones and zeros in the matrix of the recurrence plot, measuring their return points. The DET is the number of points present in the recurrent formation of the diagonal lines over all the points of relapse and it is related to the degree of predictability of the system.Reference Ferreira, Messias, Vanderlei and Pastre 19
The ES is a measure that describes the irregularity, complexity, or uncertainty of an experimental time series, representing the energy expended to produce work.Reference Ferreira, Messias, Vanderlei and Pastre 19 Mathematically, the calculation of ES signal is negative – the higher the entropy value the more information and greater adaptability to the environment.Reference Selig, Tonolli, Silva and Godoy 27 Thus, if ES is large, then the distribution is flat – all the random patterns are distributed and the series takes as much information as possible. On the other hand, if small, there is likely a subset of patterns.Reference Kunz, Souza, Takahashi, Catai and Silva 28
Mathematics series
Mathematical and linear random series were createdReference Baptista 29 in order to compare the patterns of behaviour of these series with those obtained by the series of the volunteers in this study. The series obtained were also analysed using the visual recurrence analysis software to obtain the indices derived from the recurrence plot.
The random series were constructed in Excel using the formula Random ()×100, random values between 0 and 100 were obtained after excluding the decimal places. The linear series was obtained through a series constructed with prime numbers between 2 and 3800.Reference Baptista 29
Statistical analysis
Standard statistical methods were used for the calculation of means, standard deviations, maximum and minimum values, and confidence intervals of 95%. Normal Gaussian distribution of the data was verified by the Shapiro–Wilk goodness-of-fit test (z value >1.0). For parametric distribution, we applied the Student t-test, and for non-parametric distribution we applied the Mann–Whitney test. Differences were considered significant when the probability of a Type I error was lower than 5% (p<0.05). The calculation of the study power with the number of subjects analysed and significance level of 5% (two-tailed test) confirmed a power higher than 80% to detect differences between the variables.
Results
Table 1 shows the characteristics of both the groups. There were no significant differences between the groups.
All the volunteers from the type 1 diabetes mellitus group were insulin dependent. In addition, three subjects (15%) used medication to control blood pressure, three (15%) for thyroid disorders, two (10%) for peripheral neuropathy symptoms, and two (10%) used contraceptives. Regarding co-morbidities, the control group included two volunteers classified as obesity grade 1, and one subject in the type 1 diabetes mellitus group was classified as obesity grade 3. 21
Table 2 presents the values of the DET, REC, and ES indices derived from recurrence plot quantitative analysis. Smaller values of these indices were observed for the type 1 diabetes mellitus group compared with the control group (p<0.05).
Mean values, followed by their respective standard deviations of DET, REC, and ES indices derived by RP for the studied groups.
DET=determinism; DM1=diabetes mellitus type 1; ES=Shannon entropy; REC=recurrence rate; RP=recurrence plot
Mean±standard deviation [95% CI]
Regarding the qualitative analysis of recurrence plot, representative samples of the groups are presented in Figure 1. The selected recurrence plot of the individuals presented the DET and REC values closer to the average for their respective groups.
( a ) Recurrence plot of young healthy controls (REC: 26.35 and DET: 97.64), ( b ) recurrence plot of young diabetic individuals (REC: 17.69 and DET: 96.14). It is noticed that figure A has more squares and lines along the whole image, compared with figure B, and this pattern is present in all the images of the control group; however, no image pattern is present in the type 1 diabetes mellitus group, in which the number of lines and squares formed are different between the diabetes subjects.
The recurrence plot of the healthy subjects presented a pattern with more square and vertical and horizontal lines apparent in comparison with the diabetic recurrence plot, which showed the least square lines. Furthermore, the intra-group analysis showed that the control group recurrence plot patterns were between individuals, which did not occur in the diabetic subjects’ recurrence plot, which showed greater variation.
Table 3 shows the results of the mathematical models used to compare behaviour of the time random and linear series with the average values of the groups studied. It is worth noting that the average value of the type 1 diabetes mellitus group is tending towards randomness when compared with the average value of the control group.
Values of REC, DET, and ES of the studied groups and the mathematical models of time series: random, diabetic group, control group, and linear.
DET=determinism; ES=Shannon entropy; REC=recurrence rate
Table 4 displays values of the heart rate variability indices obtained in the time and frequency domains. We observed significantly increased values of LF (ms2), HF (ms2), RMSSD, and SDNN in the control group.
Mean values, followed by the respective standard deviations of HRV indices obtained in the time and frequency domains of the studied groups.
DM1=diabetes mellitus type 1; HF=high frequency; HRV=heart rate variability; LF=low frequency; LF/HF=low and high frequency ratio; RMSSD=root mean square of the sum of squares of differences between iRR adjacent; SDNN=standard deviation of all normal R–R intervals recorded in a time interval
Mean±standard deviation [95% CI]
Discussion
This study was carried out to evaluate nonlinear dynamics of heart rate in young type 1 diabetic subjects through analysis of recurrence plots. Our results indicate that young diabetic subjects demonstrate reduced DET, REC, and ES indices compared with healthy young control subjects, suggesting that this group has a tendency to random cardiac autonomic regulation behaviour. In other words, they present impaired cardiac autonomic regulation.Reference Peng, Havlin, Hausdorff, Mietus, Stanley and Goldberger 30 Recurrence plot qualitative analysis showed that diabetic patients had a pattern with less squares and apparent vertical and horizontal lines, as well as greater figures of variation between individuals compared with the recurrence plot of controls.
These findings were accompanied by reduced sympathetic and parasympathetic regulation of the heart in the type 1 diabetes mellitus group, observed by heart rate variability linear indices analysis.
Using the REC, DET, and ES values obtained from random linear mathematical series to compare the values found in all groups, we observed that the type 1 diabetes mellitus group had values close to random series, suggesting that type 1 diabetes mellitus subjects have a standard tendency towards randomness, which may explain the lower values of the DET and REC indices reported in our study.
The presence of randomness produces unpredictable cardiac autonomic modulation behaviour,Reference Higgins 31 and this random behaviour can influence, directly or indirectly, the occurrence of complications of type 1 diabetes mellitus.Reference Hotta, Otsuka and Murakami 32
Unlike the results presented in our investigation, a previous study evaluated heart rate variability in healthy young subjects and type 1 diabetes mellitus patients. The authors described higher values of DET for individuals with type 1 diabetes mellitus, indicating a reduced complexity in the control of heart rate in the group with diabetic patients;Reference Javorka, Trunkvalterova, Tonhajzerova, Lazarova, Javorkova and Javorka 17 however, the methodology used in the mentioned study included restriction of ANS stimulants for 12 hours before evaluation, the heart rate variability analysis was performed in the morning, and the authors did not exclude smokers. These criteria are different from that of our study, which may have influenced the results and contributed to the differences in results.
Regarding ES, we observed higher values for healthy subjects compared with diabetic individuals. The ES reduction is related to sympathetic–vagal imbalance, which can predict depressed global modulation of heart rate as well as pathological conditions such as malignant cardiac arrhythmias, heart attack, and even sudden death,Reference Kunz, Souza, Takahashi, Catai and Silva 28 indicating that patients with type 1 diabetes are more prone to such conditions.
For the recurrence plot qualitative analysis of the type 1 diabetes mellitus group, we reported a pattern graph with lower recurrence, which was against the lowest values for DET and REC indices observed in the qualitative analysis of this group. Volunteers in this group also showed a pattern of more diverse recurrence plot figures, indicating autonomic regulation of the heart with greater variation.
The index that evaluates global modulation of heart rate, the SDNN index,Reference Vanderlei, Pastre, Hoshi, Carvalho and Godoy 7 was reduced in the type 1 diabetes mellitus group, suggesting that these individuals have reduced overall regulation of the heart, which is an indicator of abnormal function and insufficient adaptation of the cardiac autonomic regulation.Reference Vanderlei, Pastre, Hoshi, Carvalho and Godoy 7 , Reference Mogensen, Jensen and Kober 33 This finding may indicate a high predictor of mortality, possibly due to heart attack, atrium fibrillation, other cardiac arrhythmias, congestive heart failure, and ischaemic heart disease.Reference Rolim, Sá, Chacra and Dib 4 , Reference Villegas, Espinosa, Moreno, Echeverry and Rodrigues 5 , Reference Mogensen, Jensen and Kober 33
In agreement with this result, a study with 17 type 1 diabetes mellitus patients, aged between 12 and 30 years, reported decreased SDNN index in these individuals.Reference Javorka, Trunkvalterova, Tonhajzerova, Javorkova, Javorka and Baumert 13 In contrast, Trunkvalterova et alReference Trunkvalterova, Javorka and Tonhajzerova 15 analysed the SDNN index in 14 type 1 diabetes mellitus patients (mean age 22.3±1.2 years old) and found no significant differences between the control and the type 1 diabetes mellitus group. The authors explain that the lack of differences may be related to the small sample size used.
Our findings also indicate that the indices that determine the parasympathetic (RMSSD and HF ms2) and the sympathetic (LF ms2) modulation of the heart were lower in the type 1 diabetes mellitus group compared with the control group. This condition can produce responses in subjects’ physiology that are not efficient to induce alarm reactions and considered as an increased risk of cardiovascular diseases and significantly increased mortality rate.Reference Xhyheri, Manfrini, Mazzolini, Pizzi and Bugiardini 34 , Reference Rodrigues, Ehrlich, Hunter, Kinney, Rewers and Snell-Bergeon 35 Similar results for the LF and HF indices in children with type 1 diabetes melllitus aged between 8 and 12 years were also reported in the literature.Reference Chen, Lee, Chiu and Jeng 10
Our study presents some points that are noteworthy. The different time of diagnosis among the type 1 diabetes mellitus subjects and the presence of obesity in both groups are limitations that should be mentioned. Regarding obesity, although the type 1 diabetes mellitus group presented two volunteers classified as obesity grade 1 and 3, previous research pointed out that this situation does not affect the autonomic nervous system function.Reference Maser and Lenhard 36
In this context, it has been observed that the presence of type 1 diabetes mellitus causes major changes in the cardiac autonomic modulation, and these changes are already present in the young population, which has no other typical complications triggered by the type 1 diabetes mellitus. Change in the heart rate variability dynamics is an important factor to be prevented and treated, as its occurrence is associated with a worse prognosis and quality of life,Reference Rolim, Sá, Chacra and Dib 4 and it may influence the occurrence of various diseases such as severe arrhythmias, dysfunction of the vascular atherosclerosis, stroke, and even sudden death.Reference Soydan, Bretzel, Fischer, Wagenlehner, Pilatz and Linn 37 , Reference Matei, Popescu, Ignat and Matei 38
Thus, it is understood that early identification of cardiac autonomic changes is important, as the treatment strategies can be prepared in a more direct manner, and therefore prevent the onset or prevent complications that this disease induces, creating a better prognosis for type 1 diabetes mellitus.
Acknowledgements
The authors are grateful to CNPq’s (National Council of Scientific and Technological Development) financial support for this study (Protocol no. 477442/2012-9).
Financial Support
We received financial support from CNPq (National Council of Scientific and Technological Development).
Conflicts of Interest
The authors declare that there are no conflicts of interest.
Ethical Standards
The authors declare that all study procedures were approved by the Ethics Committee in Research of our Institution (Proc: 47/2011) and followed the rules established by the 466/2012 Resolution of The National Health Council.
