Intrinsic synchronous fluctuations of the functional magnetic resonance imaging signal are indicative of the underlying ‘functional connectivity’ (FC) and serve as a technique to study dynamics of the neuronal networks of the human brain. Earlier studies have characterized the functional connectivity of a distributed network of brain regions involved in swallowing, called brain swallowing network (BSN). The potential modulatory effect of esophageal afferent signals on the BSN, however, has not been systematically studied.
Fourteen healthy volunteers underwent steady state functional magnetic resonance imaging across three conditions: (i) transnasal catheter placed in the esophagus without infusion; (ii) buffer solution infused at 1 mL/min; and (iii) acidic solution infused at 1 mL/min. Data were preprocessed according to the standard FC analysis pipeline. We determined the correlation coefficient values of pairs of brain regions involved in swallowing and calculated average group FC matrices across conditions. Effects of subliminal esophageal acidification and nasopharyngeal intubation were determined.
Subliminal esophageal acid stimulation augmented the overall FC of the right anterior insula and specifically the FC to the left inferior parietal lobule. Conscious stimulation by nasopharyngeal intubation reduced the overall FC of the right posterior insula, particularly the FC to the right prefrontal operculum.
Conclusions & Inferences
The FC of BSN is amenable to modulation by sensory input. The modulatory effect of sensory pharyngoesophageal stimulation on BSN is mainly mediated through changes in the FC of the insula. The alteration induced by subliminal visceral esophageal acid stimulation is in different insular connections compared with that of conscious somatic pharyngeal stimulation.
The functional connectivity (FC) of brain swallowing network (BSN) is amenable to modulation by peripheral sensory input. Insula plays a pivotal role in the modulation of BSN by subliminal esophageal acid stimulation.
The primary aim of the present study was to determine if the functional connectivity of the BSN is altered by subliminal esophageal acid exposure.
We used functional magnetic resonance imaging (fMRI) to study 14 healthy volunteers during esophageal infusion. Analysis of functional neuroimages (AFNI) software was used to study the spontaneous synchronousfluctuations of the fMRI signal to characterize the functional connectivity of the BSN.
Subliminal esophageal acidification augmented the functional connectivity of the right anterior insula, and conscious somatic pharyngeal stimulation diminished the functional connectivity of right posterior insula in the BSN.
Functional magnetic resonance imaging (fMRI) studies on swallowing have identified consistent positive blood-oxygenation-level-dependent (BOLD) activity in the middle cingulate and prefrontal cortex, insula/operculum, lateral pericentral, and perisylvian cortices that collectively have been identified as the BSN.[1-5] Earlier studies have shown that these swallow-associated cortical activity maps are reproducible utilizing paradigm driven event-related fMRI analysis. The intrinsic FC of the BSN has recently been characterized with use of the resting state functional MRI techniques.
Functional connectivity is based on the observation that functionally related brain regions display intrinsic synchronized fluctuations of BOLD contrast. Biswal et al. were the first to demonstrate that the BOLD signal time series from the motor cortex were correlated with all the regions within the motor system without performance of any explicit task. The FC of various brain networks has been successfully mapped by utilizing these task-independent approaches.[9, 10] Seed based FC analysis constitutes the primary strategy that has been applied to the resting state fMRI data. In this technique, FC is the correlation between the BOLD signal fluctuation of a region of interest (ROI) and all other voxels in brain, or BOLD signal correlation in between several distributed ROIs. Region of interest(s) are a-priori demarcated as seed regions (anatomically or functionally), and results are depicted as connectivity brain maps and connectivity matrices respectively. The latter approach offers the benefit of reducing the dimensionality of the data from many thousands (the number of brain voxels) by several orders of magnitude to a few brain regions that have been identified as seed regions, thereby reducing the problem of multiple inferences in the analysis. Therefore we used this approach to study brain regions associated with swallowing, which were a-prior identified during an independent fMRI study.
Recent evidence has emerged showing that even cognitive load and subliminal perceptual processing[12-14] may modulate the FC of the brain networks. Gastroesophageal reflux events are common occurrences in healthy individuals and majority of these events are subliminal and therefore not perceived. Swallowing is the main clearing mechanism of acid from the esophagus, and previous studies have shown increased swallow frequency during esophageal acid exposure. Furthermore, subliminal esophageal acid infusion has been shown to increase cerebral cortical activity and potentiate swallow related cortical BOLD activation in healthy humans. Therefore, the interaction of subliminal esophageal acid stimulation (i.e. simulation of unperceived acid reflux events) and cortical governance of swallowing has physiologic importance and may provide the basis for future studies of the pathophysiologic esophageal conditions.
While we have studied the FC of the BSN at rest and during episodic swallows, the FC of the BSN during sensory manipulation has not been investigated. This would be an important step in understanding the neurophysiological basis of swallowing in interaction with gastroesophageal reflux. The primary aim of the present study was to determine the effects of subliminal esophageal acid exposure on the FC of the BSN.
Fourteen healthy right-handed volunteers (24 ± 4 years, seven female) participated in the study. The institutional review board at the Medical College of Wisconsin approved the study. Subjects were recruited by advertisement and all volunteers gave written informed consent before the study. All volunteers were evaluated by a personal interview, detailed health questionnaire (MCW), physical examination, and a transnasal upper endoscopy. Exclusion criteria included history of any gastrointestinal disease or functional gastrointestinal disorder. The subjects were asked to fast for 6 h before the study. We placed a double lumen 2.4 mm external diameter infusion tube transnasally following topical lubricant and anesthetic application. Perfusion ports were placed 5–7 cm above the upper border of the lower esophageal sphincter. The location of the lower esophageal sphincter was identified by either transnasal upper GI endoscopy or high resolution manometry. We waited 15 min until subjects became accustomed to the transnasal catheter prior to acquiring fMRI data. All subjects tolerated the procedure well and completed the protocol.
Subjects were supine in a 3.0 T General Electric Signa LX scanner (General Electric Medical Systems, Waukesha, WI, USA) equipped with an eight-channel receiver head coil and body quadrature transmit radiofrequency coil. Cardiorespiratory monitoring was performed at a sampling rate of 40 Hz with a pulse oximeter and respiratory bellows equipment provided in the MRI system. Subjects were instructed to relax and keep their eyes closed. Functional MRI data were acquired during four distinct 540 s scans in three different conditions: (i) awake steady state with eyes closed and infusion tube in place at the beginning and end of the scanning session (tube-i and tube-ii); (ii) awake steady state with eyes closed and phosphate buffer solution with pH = 7.0–7.5 continuously infusing at 1 mL/min (buffer); and (iii) awake steady state with eyes closed and subliminal 0.1 normal hydrochloric acid solution with pH = 1.0–2.0 continuously infusing at 1 mL/min (acid). We started the buffer and acid infusions at least three minutes prior to the onset of their respective scans (during interscan resting period) to establish an esophageal equilibrium state with the infusate. To maintain consistency across study conditions we waited 3 min before imaging data acquisition in the tube i and tube ii conditions as well. The acid scan was systematically performed after the buffer infusion in all subjects due to the potential lasting effects of esophageal acid exposure. We obtained a second scan with the infusion tube in place at the end of the study to account for time effect on the functional connectivity of the BSN and to remedy the hypothetical presence of an order effect. An anatomical scan was acquired in the middle of the scanning session after the second functional run using a high-resolution spoiled gradient recalled acquisition (SPGR) technique consisting of 140 sagittal whole brain 1 mm-thick slices over a 240 mm field of view (FOV) and 256 × 224 within slice pixel resolution. The high-resolution anatomical images were used for subsequent superposition of the lower-resolution echo planar BOLD contrast image data in each subject. Echo planar images (EPI) were acquired as 34 contiguous 4-mm thick sagittal slices over the whole brain volume in an interleaved fashion without any gap or overlap. Echo planar images images were acquired with a slice-wise pixel resolution of 64 × 64 pixels over a 240 mm field of view yielding a within-slice resolution of 3.75 × 3.75 mm, captured with an echo time (TE) of 23.4 ms and a repetition time (TR) of 2000 ms.
Functional magnetic resonance imaging signal conditioning and analysis were carried out using Analysis of Functional NeuroImages (AFNI) software. Functional EPI images were reconstructed into four-dimensional time dependent datasets wherein each voxel was associated with 540 s time series of fluctuating BOLD contrast data. We conditioned the data according to a rigorous series of signal preprocessing steps based on a previously published fMRI connectivity analysis pipeline. All participants maintained a peak-to-peak global (EPI volume) head motion of <1 mm. Physiologic (cardiac and respiratory) related signal changes expressed as second order Fourier series expansion were retrospectively corrected. In the anatomical data set, the brain was extracted from surrounding tissue and used as the reference for alignment for all EPI datasets. The reference anatomical scan was spatially normalized to match the standard Talaraich-Tournoux stereotaxic template. We performed slice timing correction and modeled head motion using 12 df indexed by time (motion parameters from a general affine transformation matrix representing shift, rotational, and shear motion), which was used to interpolate the time series back to the original acquisition grid. We computed the alignment matrix between the EPI data sets and the anatomical scan. The alignment matrix was then used to register, resample (2 × 2 × 2 mm voxel size), and align all datasets to the anatomical scan grid. Voxel-wise extreme signal fluctuations of the signal (spike values) were replaced by a fitted smooth curve to the time series. Functional magnetic resonance imaging BOLD signal trend components were removed over the course of time series voxel by voxel independently using linear least squares. White matter and cerebrospinal fluid containing voxels were identified automatically based on their signal intensity and manually verified within each subject. The average time courses within the white matter and cerebrospinal fluid voxels and global noise were extracted for use as nuisance regressors. General linear modeling techniques with orthogonal least squares estimation were used to remove undesirable signal contamination correlated with motion parameters, white matter, cerebrospinal fluid and global noise temporal components as covariates of no interest. Residual time series were then spatially smoothed to full-width -half-maximum (FWHM) of 6 mm and bandpass filtered to a frequency target range of 0.015–0.1 Hz using Fourier transformation. Subsequently the first 60 s of each run was discarded (scanner equilibration period).
We adopted previously identified positive (increased BOLD signal during swallow task) and negative (decreased BOLD signal during swallow task) activated regions from an earlier study (the same research participants less than 2 weeks apart from current study) and utilized them as seed regions for functionally guided connectivity analysis of the brain swallow network. The average time series for each seed region was individually extracted (averaged time course across all voxels within a seed region) individually during each condition. We assessed functional connectivity between any two seed regions by calculating the Pearson product moment-correlation coefficient (r) between the regions' preprocessed 480 s time series. We also used the resting state data of these same subjects who participated in our previously published resting state protocol to determine the effects of constant noisome presence of the nasopharyngeal tube on FC of the BSN.
Statistical analysis was performed using customized MATLAB® software (MathWorks, Natick, MA, USA). We calculated cross-correlation coefficients (CC) between preprocessed 480 s time series (240 data points) of the paired ROIs, with values ranging from −1 to +1. Large positive values (i.e. closer to 1) are indicative of high functional connectivity, while large negative values (i.e. closer to −1) are indicative of highly anticorrelated functional connectivity. Small values (i.e. close to zero) are suggestive of low functional connectivity between corresponding pairs. A Pearson product-moment correlation coefficient (r) obtained from 240 data-points (df = 238) is considered significant if the absolute r-value is higher than 0.165 (|r| > 0.165, two-tailed p < 0.01). For each subject there were 780 [(40 × 39)/2] pairwise CC values among 40 ROIs of the deglutition connectome. These CC values were arranged into an individual connectivity matrix for each study condition and a frequency distribution of CC values was determined. The matrices of correlation coefficients, hereafter called FC matrix, were Fisher z-transformed to produce approximately normally distributed variables. These values were then averaged across subjects to generate a group FC matrix for each condition (tube i, tube ii, buffer, and acid). We plotted the CC values as a color matrix to confirm consistency of the swallow network in the presence of the nasopharyngeal tube. We performed a paired t-test between tube-i and tube-ii conditions and plotted CC of group FC matrices and ran linear regression analyses to confirm reproducibility of the data and identify potential order effect. Then we compared CC matrices of buffer and acid conditions to investigate subliminal esophageal infusate effect. Finally, the two CC matrices of tube i and ii conditions were compared with two resting state CC matrices of the same 14 subjects (resting i and ii) adopted from our previous study to evaluate the effect of nasopharyngeal intubation (tube effect). In all comparisons, false discovery rate (FDR) was used as the statistical method to correct for multiple comparisons. False discovery rate controlling procedures exert a less stringent control compared with family wise error rate procedures (such as the Bonferroni correction). This provides a useful compromise between the loss of power attributable to the Bonferroni correction and the lack of control of Type I errors associated with comparisons unadjusted for multiple comparisons. The FDR method has been widely applied in genetic and neuroimaging research, where quantitative analyses took a new dimension and the number of hypotheses tested in an experiment reached thousands. In this statistical method a new quantity called the ‘q-value’, which is the FDR analogue of the p-value has been utilized to test the null hypothesis.
After studying the individual connection dynamics of the large scale network, we defined the following FC scores for each seed region across different study conditions to further investigate the neural components as previously used by investigators for motor networks[6, 31]: (i) Average CC score of every cortical positive BOLD activated seed to remainder of cortical positive BOLD activated regions; (ii) Average CC score of every cortical negative BOLD activated seed to the remainder of cortical negative BOLD activated seeds; (iii) Average CC score of subcortical activated seeds to the remainder of subcortical regions; and (iv) Average CC score of subcortical activated seeds to the cortical BOLD positive and negative BOLD seeds. We used anova and post hoc analysis of FC scores to evaluate the effects of various study conditions.
Volunteers did not report any heartburn, nausea, regurgitation, or chest discomfort during or after esophageal acid infusion.
Group CC matrices of tube-i, tube-ii, buffer, and acid infusion states are shown in Fig. 1. The diagonal of the CC matrix from the upper left to the lower right corners represents connectivity of each seed region to itself, which (by definition) is always 1 (fully correlated). The lower left section of the CC matrix below this diagonal represents CC values and the right upper part of the matrix above the diagonal represents the standard deviation of each corresponding connection. High positive CC values (closer to +1) represent high FC, depicted as warm colors (orange – red). Negative CC values (below 0) represent anticorrelations, demonstrated as cold colors (blue – navy). Smaller CC values (close to 0) represent weak FC as shown in green color (Fig. 1). Consistent with our previous report, the CC matrices showed a consistent pattern of strong FC amongst cortical BOLD positive activated, and negative BOLD activated seeds throughout all study conditions as depicted by the upper left and lower right triangles respectively. Interaction of BOLD negative activated seeds and BOLD positive activated seeds demonstrated mainly an anticorrelation pattern (or no significant correlation), outlined by a light green-blue square in the lower left section of the CC matrices. Subcortical seeds showed persistent low level of FC connectivity across both cortically positive and negative BOLD activated seeds as depicted in the middle yellow band of the FC matrices. Variance of the data was homogeneously distributed as shown by a uniform yellowish green representation of standard deviation above the diagonal of the CC matrices (Fig. 1). CC values of the BSN were not normally distributed and were skewed to the right as shown in their frequency histograms of each condition (Fig. 2). This pattern was similar across study conditions, and hereafter CC values were z-transformed to assure normalized distribution of the data for further analyses.
In the cortical BOLD positive component of the FC matrix (top left triangle of FC matrix) consisting of 153 possible connections, 79, 89, 76, and 82 of them displayed significant FC (r > 0.165) during tube-i, tube-ii, buffer, and acid conditions respectively that is suggestive of a coherent functional network (50–58% significantly correlated and 0% anti-correlated). In the BOLD negative component of the FC matrix (bottom right triangle of all matrices) consisting of 153 possible connections, 64, 71, 66, and 59 of the connections displayed significant FC (r > 0.165) during tube-i, tube-ii, buffer, and acid conditions respectively (39–46% significantly correlated and 0% anticorrelated).
Within the interaction of BOLD positive and negative component of the FC matrix (bottom left square of all matrices), however, only one of possible 324 connections (left middle cingulate and left anterior cingulate) displayed significant correlation (r > 0.165) across all study conditions. 42, 42, 40, and 40 of the connections displayed significant anticorrelation (r < −0.165) during tube-i, tube-ii, buffer, and acid conditions respectively (0% significantly correlated and 12–13% anti-correlated). In the subcortical BOLD positive component of the FC matrix (middle band the BSN matrix), no link amongst all 150 connections displayed significant FC (0% significantly correlated or anticorrelated).
Effects of subliminal acid infusion
We compared the FC of all 780 connections of the swallow network across the acid and buffer conditions using paired t-tests. Significant difference was observed in cortical positive BOLD activated regions. Comparison of the acid and buffer infusion states showed that the FC of right anterior insula and contralateral inferior parietal lobule (Fig. 3) is significantly increased during subliminal esophageal acid stimulation (q < 0.05). We compared tube-i and tube-ii (preinfusion and postinfusion) to study the possible order effect on the observed differences between buffer and acid conditions. If the time/sequence of scan during fMRI session was an effective factor resulting in some differences, the postinfusion scan should also show similar difference compared with the preinfusion scan. This comparison did not show any significant FC difference (q > 0.05, Fig. 4A). Furthermore, as shown in Fig. 4B, the linear regression analysis of group FC matrix during tube-i and tube-ii conditions demonstrated highly reproducible values (r2 = 0.94, p < 10−5). Thus, the observed difference between the acid and buffer conditions could not have been simply related to the timing of scan or order effect. We then performed a FC score analysis of all the regions comprising the BSN. Acid infusion significantly increased the FC score of the right anterior insula region (p < 0.05), further confirming the role of right anterior insula in cortical processing of esophageal acid exposure (Fig. 5). No other cortical BOLD positive, cortical BOLD negative or subcortical region showed a significant difference across study conditions.
Effects of nasopharyngeal intubation
Significant differences were observed in cortical and subcortical positive BOLD activated regions. Intrinsic FC between the right posterior insula and the ipsilateral ventrolateral prefrontal cortex was significantly diminished with nasopharyngeal intubation (q < 0.05, Fig. 6). Amongst the subcortical regions, the FC between the right thalamus and the red nucleus was significantly diminished after nasopharyngeal intubation (q < 0.05, Fig. 6). We then performed an analysis of FC scores of all the BSN regions. This analysis showed that the FC score of the right posterior insula significantly changed during nasopharyngeal intubation (Fig. 7, p < 0.05). No other cortical BOLD positive, cortical BOLD negative, or subcortical region showed any significant difference between resting and intubated state or across infusion states.
In this study, we determined the alterations in the FC of brain regions involved in swallowing during sensory stimulation. We found that although the overall architecture of the BSN is stable, the FC of a subset of the BSN regions is modulated by pharyngoesophageal sensory input. Subliminal esophageal acid stimulation augments the FC of right anterior insula to the BOLD positive component of the BSN and specifically between the right anterior insula and the left inferior parietal lobule. The conscious somatic stimulation induced by nasopharyngeal intubation significantly reduces the FC of right posterior insula to the BOLD positive component of the BSN and particularly the FC between the right posterior insula and the right prefrontal operculum, and diminishes the FC between the thalamus and the red nucleus. The functional connectivity of the negative BOLD component, or the interaction of BOLD negative and positive components was not affected by the conscious somatic or subliminal visceral pharyngoesophageal sensory stimulation.
Determining the effects of sensory stimulation on FC of cortical regions associated with swallowing activity is important from a physiological and even pathophysiological viewpoint. Our earlier studies have documented the accentuating effect of esophageal acid stimulation on cortical bold activity associated with swallowing. Given the fact that esophageal reflux occurs in various degrees in healthy individuals and patients with gastroesophageal reflux disease, and that swallowing plays an essential mechanistic role in the clearance of the potentially hazardous chemicals from the esophagus, modulatory effects of esophageal acid stimulation has both physiologic and potentially clinical relevance. The present study further supports the finding of a previous study that subliminal esophageal acid stimulation modulates the underlying fMRI activity of cortical regions governing the swallow motor function. Furthermore, it shows that this modulation extends beyond a temporary increase in swallow related cortical activity and includes how these regions functionally connect to each other at rest without an explicit task. These findings provide a window into the depth of involvement of the cerebral cortex not only in conscious but also unperceived physiologic stimulations occurring in the upper gastrointestinal tract. In addition, they clearly establish the influence of an unperceived signal beyond a sensory network and onto a motor related network such as swallowing.
Another important finding of the present study is the observation that mere transnasal esophageal intubation of healthy volunteers affects the functional connectivity of a distinct subset of the swallow-related cortical regions. Considering the invasive nature of the transnasal placement of infusion catheter this effect is not unexpected. It is reassuring that alterations in these FC findings, at the last phase of study was not different from that of the beginning of the study alleviating our concern about the order effect as a cause of observed subliminal acid induced changes. Limitations of current technology have not allowed for systematic and simultaneous investigation of the effects of these modulations on swallowing frequency, efficacy, and biomechanical characteristics. However, it is important to remember that the acid effect compared with buffer, manifested itself above and beyond the alteration induced by the presence of the infusion tube and was observed at a stimulation level that was not perceived by subjects.
Although functional brain imaging has greatly advanced our understanding of the neural basis of swallowing and has validated the crucial role of cerebral cortex during deglutition in humans,[1, 3, 32, 33] the exact neurophysiology of cortical control of deglutition and functional role of swallow-related cortical regions remain elusive. The lateral sensorimotor cortex, both primary motor and sensory (MI and SI), plays a fundamental role not only in the execution of the semiautomatic deglutitive motor sequence[35, 36] but also monitoring of sensory signals from oropharyngeal structures.[37, 38] Earlier studies have shown that when the network was engaged in the performance of episodic swallow task the FC of bilateral sensorimotor regions was augmented signifying the crucial role of this region in execution of the swallow task. Supplementary motor area (SMA) and mid-cingulate cortex activations overlap with brain regions identified in speech production and word reading that employ similar oropharyngeal neuromuscular apparatus. These two areas are also activated in a range of other volitional motor activities and probably play a role in attention and motor planning. It is not surprising that during current study which sensory stimulation was not associated with a constant volitional response, these areas did not show any alteration in the functional connectivity.
The insula is one of the primitive regions of the brain and receives direct afferents from the thalamus. It is believed to serve as the central cortical processing region of the visceral afferents[42-44] and is organized in a hierarchical fashion.[45, 46] In addition to swallowing,[34, 37] it is involved in cognitive, socio-emotional, gustatory, and sensorimotor functions. The insula has shown changes in fMRI activity during esophageal acid exposure, and esophageal distention., Anterior insula is the proposed final site in intrainsular processing of visceral afferent signals before these signals are projected to other cortical regions such as prefrontal cortex. A recent study has revealed that right anterior insula is the most active among various insular regions associated with swallowing during gustatory or electrical sensory stimulation. Our finding of altered functional connectivity of the right anterior insula with other cortical regions involved in swallowing during esophageal acid exposure concurs with the pivotal role of insula in sensory aspects of swallowing. The inferior parietal lobule has shown activity associated with deglutition and is considered a component of the brain swallowing network.[34, 47] The lateral part of the inferior parietal lobule, parietal operculum (BA 40,43), probably represents the neuroanatomical correlate of secondary somatosensory cortex (SII) and has been viewed as a projection area of visceral stimulation. In that, the current study also shows amplified FC of right anterior insula to this region during subliminal esophageal acid infusion pointing to the putative visceral sensory role of this region.
The thalamus, putamen (dorsal striatum), and cerebellum,,,, have previously shown activation associated with volitional swallowing. Striatum is usually considered the gateway of primary motor cortex to thalamus, which in turn has projections to SMA, and may play a role in sensorimotor integration.[57, 58] Cerebellar cortex may play an important role in the adaptive coordination, and timing of the stereotypical contractions of complex oropharyngeal musculature during deglutition. Swallow related BOLD activity of the red nucleus has not been reported in previous studies, but another study focusing on the subcortical regions reported fMRI activity in adjacent substantia nigra. Red nucleus receives projections from motor cortex (BA 4, 6 and 8) and deep cerebellar nuclei through medial cerebellar systems and plays an important role in motor control. As very little is known about the function of the red nucleus in humans and no direct anatomic rubrothalamic connections are described in the literature, we hesitate to speculate regarding the importance of diminished FC of thalamus-red nucleus to statistically insignificant level (r < 0.165) during nasopharyngeal intubation. Furthermore, the pathophysiologic impact of the observed alterations at a functional level is not currently known and whether any of these functional connectivity modulations can affect behavioral changes merits further investigation with availability of newer technology and equipment.
In summary, following our previous report of modulation of FC of bilateral sensorimotor region within the swallow network during engagement in repetitive swallowing, the present study shows modulation of right posterior insular FC during conscious nasopharyngeal tube presence, and alteration of right anterior insular FC during subliminal esophageal acid exposure in healthy humans. Taken together, these findings indicate that the brain swallowing network is distinctly responsive to both perceived somatic physical and unperceived visceral chemical afferent input from the deglutitive axis.
Supported in part by NIH grant 5R01DK025731-29 and 2T32DK061923-06.
The authors have no conflict of interest to disclose.
AB had a major role in all aspects of this manuscript including study concept and design, data acquisition, analysis and interpretation of data, drafting of the manuscript, and critical revision of the manuscript for important intellectual content; RMS contributed to data analysis and interpretation; MK and BDW contributed to data analysis and interpretation of data along with statistical analysis; SJL and RS had a major role in study concept and design, drafting of the manuscript, supervision, and critical revision of the manuscript.