Young type 1 diabetes subjects sway more than healthy persons when somatosensory system is challenged in static standing postural stability tests

In type 1 diabetes, it is important to prevent diabetes‐related complications and postural instability may be one clinically observable manifestation early on. This study was set to investigate differences between type 1 diabetics and healthy controls in variables of instrumented posturography assessment to inform about the potential of the assessment in early detection of diabetes‐related complications. Eighteen type 1 diabetics with no apparent complications (HbA1c = 58 ± 9 mmol/L, diabetes duration = 15 ± 7 years) and 35 healthy controls underwent six 1‐min two feet standing postural stability tests on a force plate. Study groups were comparable in age and anthropometric and performed the test with eyes open, eyes closed (EC), and EC head up with and without unstable padding. Type 1 diabetics exhibited greater sway (path length, p = 0.044 and standard deviation of velocity, p = 0.039) during the EC test with the unstable pad. Also, power spectral density indicated greater relative power (p = 0.043) in the high‐frequency band in the test with EC head up on the unstable pad and somatosensory activity increased more (p = 0.038) when the unstable pad was added to the EC test. Type 1 diabetes may induce subtle changes in postural control requiring more active balancing when stability is challenged. Postural assessment using a portable easy‐to‐use force plate shows promise in detecting a diabetes‐related decline in postural control that may be used as a sensitive biomarker of early‐phase diabetes‐related complications.


| INTRODUCTION
Type 1 diabetes is an autoimmune disease which leads to a chronic condition where the body is not able to produce insulin, and without insulin secretion, people with diabetes are exposed to elevated blood glucose concentration and potentially diabetes-related complications over time (Melendez-Ramirez et al., 2010;Van Belle et al., 2011).
Microvascular complications, like neuropathy and retinopathy, often develop first and may already reduce the quality of life and increase the health care burden (Girach et al., 2006).Thus, any improvements in clinical health care directed to diabetes and diabetes-related complications should be considered.
Retinopathy and neuropathy are known to affect gait instability and postural balance (Bonnet et al., 2009;D'Silva et al., 2016).It has been shown that neuropathic diabetics were more swaying during the postural measurements when compared to non-neuropathic diabetic and normal healthy subjects (Giacomini et al., 1996;Oppenheim et al., 1999;Uccioli et al., 1995).Furthermore, diabetics with peripheral neuropathy had increased postural instability, larger sway motion and sway speed when compared to healthy controls (Boucher et al., 1995).In their review, Sempere-Bigorra et al. (2021) disclosed that only type 1 diabetics, not type 2 diabetics, showed reduced nerve excitability.In addition, they assumed that altered axonal function predates the onset of clinical neuropathy in type 1 diabetes.The altered nerve excitability may affect postural balance.
Human postural balance has been defined as maintaining, achieving, or restoring postural balance (Low et al., 2017).Postural control is a summary of multiple sensorimotor systems which include visual, vestibular, somatosensory, and higher-level motor systems in the brain (Mancini & Horak, 2010).Deficits in any postural balance system will cause an increase in postural instability.People with known postural balance deficits have the greatest decrease in postural stability when their vision is disturbed.This highlighted the reliance on vision to compensate for other sensory deficits (D 'Silva et al., 2016).In addition, body sway has increased particularly when patients stand with their eyes closed (EC) (Nardone & Schieppati, 2004).However, distractions of postural balance are commonly found in patients with polyneuropathy due to poor proprioception and motor function (Di Nardo et al., 1999).One of the reasons for this postural instability has been thought to be the loss or damage of spindle afferent fibres of large diameter (group 1a) (Nardone & Schieppati, 2004).Some authors have pointed out that the prevalence of diabetic peripheral neuropathy with type 1 diabetics is 6% at the onset of the disease, and it will increase up to 30% after 13-14 years of progression (Hicks & Selvin, 2019;Sempere-Bigorra et al., 2021).
Previous studies have identified clear differences in postural balance between patients with long-term diabetic neuropathy and healthy controls (Bonnet et al., 2009;Boucher et al., 1995); however, literature regarding early detection of diabetes related complications using postural balance assessment is mixed.Bonnet et al. (2009) did not find differences between diabetics without any complications and healthy controls during quiet standing.However, Almurdhi et al. (2017) found that patients with type 2 diabetes have alterations in balance and gait compared to healthy controls.In addition, Oppenheim et al. (1999) also proposed that there are some behavioural differences under the challenging and disturbed postural balance situations and suggested that the postural control and postural balance of diabetics may differ from healthy people.
We aimed to investigate potential differences in centre of pressure (COP) movement between type 1 diabetics with no apparent complications and healthy controls in static standing balance tests.We used a force plate to measure COP path in traditional two feet standing tests (feet side by side) with eyes open (EO), EC with neutral head posture and head up with EC with or without a soft foam pad under the feet.Obtained knowledge could influence the current clinical diagnostics and produce some new practices to type 1 diabetes care, especially for preventing the potential risk factors for falls.

| METHODS
This study is part of the DIAMES-Diabetes Mellitus Exercise and Stress study, which was approved by the Ethical Committee of Northern Savo District in June 2021.The study was conducted between September 2021 and November 2022 in Kuopio, Finland.In total 53 subjects were enroled on the study including 18 subjects with type 1 diabetes and 35 healthy controls, and all subjects signed written informed consent before participation in the study.
The inclusion criteria for the type 1 diabetes group (DM) were age 18-50 years, body mass index (BMI) 18-35 kg/m 2 , type 1 diabetes duration 3-25 years and absence of diabetes-related major complications.Three subjects from the DM group were under mild hypertension and cholesterol medication.The inclusion criteria for the control group (CON) were age 18-50 years, BMI 18-35 kg/m 2 and overall healthy without any chronic diseases.Subjects were nonsmokers, and they avoided any strenuous exercise for 48 h and refrained from alcohol, nicotine, and caffeine products for 12 h before the study visit.The participant had to be healthy 2 weeks before the visit date and no severe hypoglycaemia (<2.9 mmol/L) occurred during the last 24 h before the visit.People with type 1 diabetes were instructed to maintain blood glucose levels between 5 and 13.9 mmol/L throughout the study visit and this was frequently controlled from capillary blood obtained by fingerstick.The DM group had blood glucose concentration of 9.0 ± 3.4 mmol/L before starting the tests.Venous blood samples were collected into EDTA tubes for hemoglobin A1c (HbA1c).The concentration of HbA1c was determined according to the manufacturer's instructions (cobas 8000 (c 702)-analyzer; Hitachi High Technology Co).

| Postural stability tests
COP trajectory during the balance tests was recorded using a force plate (balance platform BT4; HUR Ltd) with a sampling rate of 50 Hz.
The platform was calibrated before starting the measurements of each participant.The postural stability tests included six 1-min postural standing tests, with a short resting period between each test.Three of the tests were carried out without and three with 3 cm thick foam padding placed on top of the platform (unstable padding).Within both groups, half of the subjects performed the tests first on the stable platform while the other half performed the tests first with the addition of unstable padding.This ensured that the effects of muscle fatigue and habituation on between-condition differences could be ruled out from group mean data.In all tests, subjects were instructed to stand on two feet with feet at a 30-degree angle, heels separated by 2 cm and arms relaxed on both sides of the body.In this position, the swaying is considered uniform in both antero-posterior and medio-lateral directions according to the force plate manufacturer.
In the first test, the subjects were instructed to maintain postural balance for 1 min with EO while looking at a black circle slightly below eye level at 4 m distance (EO).In the next test, subjects were instructed to close their eyes and maintain balance for 1 min (EC).Finally, subjects were requested to close their eyes and rotate their head upwards and maintain the balance for 1 min (EC head up, ECH).After having completed the tests on the stable platform or with the addition of unstable padding, subjects repeated the tests on the other condition.

| Centre of pressure analysis
All the parameters were derived from preprocessed continuous COP data.First, the data was zero-centred by subtracting mean values from antero-posterior and medio-lateral directions.After that, the data was filtered with a lowpass Butterworth filter with a cut-off frequency of 5 Hz and order of 20.
After preprocessing the COP signal, sway path length (later sway path), standard deviation of sway velocity (VSD), and area of postural stability sway comprising 90% of the COP data (C90) were calculated.
Sway path describes the length of the path COP travels during the test and is calculated as the sum of distances between each discrete COP point.VSD is calculated as the standard deviation of the COP signal's velocity and describes the overall variation in COP velocity during the test.Finally, C90 describes the ellipse area which encloses approximately 90% of the COP data.The ellipse was determined using principal component decomposition, in which we extracted the two most significant eigenvectors in terms of explained variance of the covariance matrix.These two eigenvectors are orthogonal and correspond to the major and minor axis of the ellipse and the corresponding eigenvalues describe the variances in the directions of the axes.By assuming that the ellipse equation follows a χ 2 distribution with two degrees of freedom, we can estimate the ellipse area that comprises 90% of the COP.Romberg's constant was also determined from EO and EC tests without the unstable pad.An example of sway path, VSD and C90 are shown in Figure 1a.
The frequency content of the data is also interesting as frequency bands have been connected to different balance compensatory mechanisms.The power in the very low-frequency band comprising of 0-0.10 Hz (or sometimes even 0-0.30Hz) has been linked to balance adjustments based on visual control, power in the frequencies between 0.1 and 0.50 Hz has been connected to controls based on vestibular stress, mid to high frequencies between 0.50 and 1.0 Hz to reflect lower extremity balance compensatory activities based on somatosensory information, and frequencies over 1.00 Hz have been connected to tremor and unregulated movements due to abnormal central nervous system function (Lyytinen et al., 2010;Nagy et al., 2004;Oppenheim et al., 1999).In this study, we have separated the frequency bands into 0-0.1 Hz, 0.1-0.5, 0.5-1.0, and The power spectral density (PSD) was calculated using Fourier transformation.Zero padding was used over the COP data to have 4096 points in the Fourier spectrum, corresponding to a frequency grid interval of 0.012 Hz.A Hanning window was used to force the data to be periodic, and thereby, avoid spectral leakage.The Fourier spectrum was calculated for both directions separately and then averaged over each frequency point.By doing so, spectral components in both directions were accounted for.The spectral power for the bands 0-0.1 Hz, 0.1-0.5 Hz, 0.5-1 Hz, and >1 Hz were calculated and divided by total power to better compare subjective activity in each of the frequency bands.Finally, the median frequency of the Fourier spectrum was calculated to provide a robust estimate for the centre frequency of the PSD.

| Statistics
All the results are presented as mean and standard deviation (SD), estimate and 95% confidence interval (CI) or as the median and interquartile range (IQR) when normality could not be met with the Shapiro-Wilk test.The descriptive parameters were compared between the groups with Welch's two-sample t test.In case abnormality, the groups were compared with Wilcoxon ranksum test.
Group differences in postural stability parameters were estimated using a mixed-effects model with a random intercept for subject to account for repeated measurements study design.The

| RESULTS
Type 1 complications free diabetics, and healthy controls were analyzed for postural balance.The groups were matched by age, anthropometry and sex and did not differ statistically (Table 1).The DM group had greater HbA1c compared to the CON group as expected (DM, 58 ± 9 mmol/mol vs. CON, 32 ± 2 mmol/mol, p < 0.001).
In group pairwise comparisons, the groups did not possess statistically different postural stability patterns on the stable platform regardless of having EO, EC or head up with EC.However, with the unstable pad in the EC test, the DM group exhibited greater sway in terms of VSD (ln-transformed VSD = 0.20, 95% CI [0.01, 0.39], p = 0.039) and sway path (ln-transformed sway path = 0.189, 95% CI [0.005, 0.374], p = 0.044) when compared with the CON group (Figure 2).As for spectral power bands and median frequency, Swaying was not different in the groups in all stable platform tests.
But, in the unstable EC head up test, the DM group had more power in the high-frequency band (ln-transformed 1-5 Hz band = 0.425 95% CI [0.014, 0.835], p = 0.043) when compared with the CON group (Figure 3).
Both study groups adjusted similarly to the addition of unstable padding during EO and head up with EC tests for all variables (Figures 2 and 3).Whereas, for EC tests, the addition of unstable padding induced a greater shift to the frequency band 0.5-1 Hz (ln-transformed 0.5-1 Hz band = 0.408, 95% CI [0.022, 0.795], p = 0.038) for the DM group when compared to the effect of the CON group (Figure 3).

| DISCUSSION
In this study, we examined centre of pressure sway in participants with type 1 diabetes and healthy control subjects during steady state standing balance tests.We hypothesized that increasing difficulty in the postural balance tests would accentuate the potential deficits in the ability to maintain postural control in the subject with diabetes.
Our findings suggest, that while subjects with type 1 diabetes have similar postural stability compared to healthy controls on a stable platform, the addition of unstable padding induced a greater increase in swaying and required more active balance control in the DM group.
The results of our study indicate that it is possible to assess modest alterations in postural control with force plate measurements at an early stage of diabetes when diabetes-related complications have not clinically manifested.Early phase investigation of differences in postural balance behaviour and tools for the examination should be emphasized as recommended by the American Diabetes Association (2017, https:// diabetes.org).Furthermore, it has been noted that the clinical symptoms of neuropathy may manifest within 5 years after type 1 diabetes onset (Shafi & Latief, 2017).
Peripheral neuropathy has been shown to be one cause of postural disturbances.There is evidence that diabetics with peripheral neuropathy demonstrate weakened postural control in quiet standing tasks compared to healthy controls (Bonnet et al., 2009;Boucher et al., 1995).In addition, various authors have pointed out the impact of diabetes per se on postural control.Centomo et al. (2007) pointed out a significant difference in the dynamic reaching test between middle-aged diabetics without peripheral neuropathy and healthy controls.Oppenheim et al. (1999) indicated that diabetics without peripheral neuropathy had weakened postural control compared to healthy controls.Moreover, compared to normal healthy subjects the diabetics oscillate significantly more (Boucher et al., 1995;Uccioli et al., 1995).On the other hand, Uccioli et al. (1997) andDi Nardo et al. (1999) reported no differences in postural control measurements between diabetics without peripheral neuropathy and healthy controls.In the current study, significant differences between diabetics and healthy controls were found only in tests where unstable foam padding was used on top of the force platform.
The finding indicates that similar observations to those presented above regarding patients with diabetes-related complications can be observed in the patient population without complications when postural control is challenged more.The results further suggest that diabetes itself, without identified complications, can cause postural control disturbances.In clinical manners, the early phase health examination is important because the disease may proceed imperceptibly, and postural balance impairments may emerge before any identified complications can be seen.There is evidence showing that postural balance control can be negatively impacted even without clinically observed neuropathic complications when type 1 patients were studied (Oppenheim et al., 1999).Thus, further focus on this early screening of diabetic patients is crucial and should be done yearly (Fulk et al., 2010).
In the present study, the frequency bands of the COP sway were also examined.Previously, Oppenheim et al. (1999) have shown that there is larger spectral power in several frequency bands for diabetics with neuropathy compared to healthy controls.In the current study, we did not observe statistically significant differences between diabetics and healthy controls during quiet standing with EO.We observed that the spectral power during EO tests was mainly in the visual band (i.e., 0-0.1 Hz), which supports that visual stimulus contributes greatly to postural stability when eyes are open.As assumed, spectral power shifted towards higher frequencies in both groups when eyes were closed.In the high frequency band (0.5-1.0 Hz), the DM group exhibited statistically greater increase when compared to the CON group.This suggests that while the somatosensory activity was similar in both EC tests, the DM group had a greater increase in the power of this band due to the added postural challenge.And, when the maintenance of postural balance may be diminished due to somatosensory information deficits, the compensatory sensory activity may have increased indicating more effort to maintain stability.This is also supported by the fact that the DM group had more power in the high-frequency band with more challenging postural balance test.While movements with frequencies higher than 1 Hz have been related to tremor or uncontrolled movement, it is likely unrelated to these causes since the differences between the groups in this frequency band were detected only when somatosensory system was challenged using foam padding.As the balance is difficult to maintain when the head points up and the eyes are closed, the interpretation of having a more difficulties maintaining a well-balanced position for the diabetes subjects seems a more plausible explanation.
Fall and fear of falls are tremendous problems for people with postural balance and gait problems (Rubenstein and Josephson 2002).
Clinically accurate and useful methods for identifying fall risk in specific populations have been questioned (Jernigan et al., 2012;Muir et al., 2008).Currently commonly used postural balance methods include Berg Balance Scale, The Tinetti Performance Oriented Mobility Assessment, and Dynamic Gait Index to name a few.These balance assessment methods are adequate for their purpose, but they generally target different patients with balance disorders and are not necessarily accurate or sensitive (Chen et al., 2021).Force plate instead is a widely used method to assess COP movement in both healthy and people with diseases like diabetes (DeBerardinis et al., 2020;Piirtola & Era, 2006).Harro et al. (2016) also concluded that force plate measurements are reliable and valid tests of balance impairment in people with Parkinson's disease.
Thus, postural balance measurements using a force plate are considered justifiable for the assessment of postural stability in health and disease.
In this study, we chose to go with relatively easy two feet standing tests to ensure that we are capturing the sway motion and not to highlight subjective skills, which may have more impact in more difficult tests for example, one-foot standing tests.This also reduces the chances of falling and variability in the data thus enabling more robust estimates.However, it is also possible that more difficult tests with simple endpoints such as the duration of maintaining stability could provide more meaningful clinical interpretation and would also allow wide clinical implementation since they do not require special equipment.While difficult tests may provide a greater challenge for stability, it is still unknown whether these would provide sufficient information to assess clinical complications related to type 1 diabetes.
In a conclusion, our study indicates that type 1 diabetes, even without clinically diagnosed complications, may impair postural control.The impairments were detected when the visual feedback was blocked, and the somatosensory system was challenged with an unstable foam padding indicating that the deficits are likely in the function of the somatosensory system.We hypothesize that instrumented postural balance assessment can provide a sensitive method to detect patients with early diabetes-related complications affecting postural balance which potentially increases their risk of falls.Early detection of the risk individual with suitable interventions could help to reduce fall incidence.
Representation of the centre of pressure (COP) sway and distribution of power in the frequency domain.(a) Blue dots represent the path COP has travelled during the test and the lines represent the fitted ellipse and its major and minor axes.(b) Corresponding power spectral density derived from the COP data.Black vertical lines represent the boundaries of the frequency bands.over 1.0 Hz as shown in Figure 1b.In addition, the median frequency from the distribution was assessed.
implemented model was, in terms of mixed model notation, 'Response ~Body mass + Stability * Test * T1D + (1| Subject)'.All parameters were log-transformed, excluding the very low-frequency band PSD, to ascertain normally distributed residuals.Restricted maximum likelihood estimation was used to fit mixed models.Estimated marginal means with Kenward-Roger degree of freedom approximation were used for comparisons between groups and postural balance tests conditions.The significance level was set to 0.05 across the study and p values were not adjusted due to exploratory nature of the study.All statistical analyses were performed with R 4.0.2(R Core Team 2020).
Results of the study groups during eyes open, eyes closed (EC) and EC head up tests: (a) Sway path; (b) standard deviation of the velocity; (c) area of the ellipse (C90); and (d) Median frequency.Blue circles correspond to the control group and red triangles correspond to the diabetes group.Results are presented as mean ± SD and the asterisk * indicates a significant difference between the groups.F I G U R E 3 Results of the study groups during eyes open, eyes closed (EC) and EC head up for power spectral densities: (a) relative power in the frequency band (0-0.1 Hz); (b) relative power in the frequency band (0.1-0.5 Hz); (c) relative power in the frequency band (0.5-1 Hz); and (d) relative power in the frequency band (1-5 Hz).Blue circles correspond to the control group and red triangles correspond to the diabetes group.Results are presented as mean ± SD and the asterisk * indicates a significant difference between the groups.
T A B L E 1 Descriptive statistics.