The association of snoring, growth, and metabolic risk factors at the age of two years

Abstract

Background and aims

This observational study examined the association of snoring and growth during early childhood and the cardiovascular and metabolic risk factors based on blood samples at the age of two years.

Methods and results

The sample comprised 78 children from the CHILD-SLEEP birth cohort with full-night polysomnography (PSG) and a questionnaire consisting of parts concerning the child's sleep and environmental factors at 24 months. The growth charts were collected from well-baby clinics. Metabolic blood samples were drawn from 31 children.

There were no statistically significant differences in the growth parameters of snoring children compared to controls during the first two years of life. However, in linear regression models, snoring time in PSG significantly predicted lower levels of HDL (β = -0.484, p = 0.007) and ApoA1 (β = -0.451, p = 0.049) and higher levels of hs-CRP (β = 0.410, p = 0.019).

Conclusion

In conclusion, in Finnish children the levels of HDL and ApoA1 were inversely related to the snoring time in PSG. In addition, the snoring time in PSG significantly predicted higher levels of hs-CRP. These results suggest that snoring in early childhood could negatively alter the serum metabolic profile, adding to the risk of cardiovascular diseases in adulthood.

Introduction

Most children snore occasionally, whereas habitual snoring is less common, affecting approximately 10% of preschool-aged children (Bonuck et al. 2011). Differentiated from snoring, obstructive sleep apnea syndrome (OSAS) is characterized by upper airway obstruction that disrupts normal ventilation and sleep patterns (Standards 1996). Approximately 1–4% of children are estimated to suffer from OSAS (Lumeng and Chervi 2008). Sleep-disordered breathing (SDB) refers to all breathing difficulties during sleep, and it is associated among other aspects with behavioral and neurocognitive consequences, such as internalizing problems and poor school performance (Bourke et al. 2011; Brockmann et al. 2012).

The influence of snoring and sleep disorders on childhood growth is two-sided. On the one hand, growth failure was reported in the 1980s among children with OSAS (Guilleminault et al. 1981). Abnormal nocturnal secretion of growth hormone is a possible explanation for growth restriction among children with sleep apnea (Chennaoui et al. 2020). Consistently, the concentrations of insulin-like growth factor 1 (IGF-1) and insulin-like growth factor binding protein 3 (IGFBP-3), which are more stable in the serum than growth hormone levels and therefore more suitable as screening methods (Chinoy and Murra 2016), have been reported to be lower in children with OSA (Nieminen et al. 2002).

On the other hand, SDB is associated with obesity in childhood (Anuntaseree et al. 2014; Bonuck et al. 2015). The links between sleep-disordered breathing and obesity are multifactorial. Fat deposition in the soft tissue diminishes the diameter of the pharyngeal lumen and increases upper airway collapse (Gulotta et al. 2019). Furthermore, OSAS induces systemic inflammation, which can lead to the same kind of morbidity as caused by obesity (Bhattacharjee et al. 2011). Children suffering from obstructive sleep apnea even with normal weight appear to have higher levels of C-reactive protein (CRP) and interleukin-6 when compared to healthy children (Tauman et al. 2004; Gozal et al. 2008). Reciprocally, the levels of anti-inflammatory interleukin-10 are reported to be lower among children with diagnosed obstructive sleep apnea (Gozal et al. 2008). It is shown that the levels of CRP and interleukin-6 return to normal when OSAS is treated by adenotonsillectomy (Kheirandish-Gozal et al. 2006).

In addition to SDB, short sleep duration is consistently found to associate with growth outcomes (Ogilvie and Pate 2017). Zhou et al. found that a short sleep duration associated with higher body weight, larger increase in weight, and shorter body length during the first two years of life (Zhou et al. 2015). This implies that the relationship between sleep disturbances and growth outcomes can be established already in early childhood.

Elevated blood total cholesterol and low density-lipoprotein (LDL) levels add to the risk for cardiovascular diseases. Apolipoprotein B (ApoB) -containing lipoproteins are important agents in atherogenic lipoprotein particles, i.e., LDL (Ference et al. 2018). Apolipoprotein A-I (ApoA1) is the major apolipoprotein in high density-lipoprotein (HDL), which has a protective effect against coronary disease. In children, only a few studies have evaluated the association between sleep quality and the risk of cardiovascular diseases. Limited evidence suggests that higher sleep disturbance scores and an irregular sleep pattern are associated with an unfavorable blood lipid profile (Quist et al. 2016). Koren et al. reported that children with OSAS had had an accompanying significant increase in their HDL levels after adenoidectomy (Koren et al. 2016). Furthermore, non-obese children with OSAS and children with adenoid or tonsillar hypertrophy are reported to have lower HDL levels (Alexopoulos et al. 2011; Zong et al. 2013).

Children with OSAS and obesity have higher levels of triglycerides and Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) than children with obesity only (Flint et al. 2007). However, insulin resistance seems to be determined mainly by the degree of obesity rather than by the severity of OSAS (Tauman et al. 2005).

The purpose of this study was to add to the knowledge of the links between snoring, growth, and metabolism during early childhood. The main objective was to examine the association of snoring and growth and to study if the unfavorable effect of SDB on children’s metabolic profile based on blood samples could be seen already in early childhood.

Methods

Participants

This study is a part of the CHILD-SLEEP STUDY, and the study protocol was approved by the Ethics Committee of Pirkanmaa Hospital District on March 9, 2011 (number R11032). A birth cohort of 1679 families was recruited in Pirkanmaa Hospital District, Finland. The infants were born from April 2011 to February 2013. The details of the recruitment procedure have been reported previously (Juulia Paavonen et al. 2017). All the families of the CHILD-SLEEP cohort received a questionnaire prenatally and follow-up questionnaires at the ages of three, eight, 18, and 24 months postnatally. In addition to the questionnaires, the CHILD-SLEEP STUDY comprised several subsamples.

The CS study had two separate subsamples undergo PSG at 24 months: an observational case control study of snoring children recruited at 8 months and their controls (Markkanen et al. 2021), and a follow up study of healthy children recruited at birth (Satomaa et al. 2016). All children with PSG recoding at 24 months (N = 117) were eligible for the present study. The questionnaires were used to estimate the background factors, and 39 children in the PSG subsamples did not have a 24-month questionnaire available leaving 78 children for further analyses. To assess the effects of sleep-disordered breathing on metabolic parameters, these children were then divided into PSG snorers (n = 19) and PSG non-snorers (n= 59) based on their snoring time in PSG recordings. This approach was selected because of the fairly low level of obstructive apnea–hypopnea index (OAHI) (median 0.00, range 0.00–6.30) in this study population. The amount of snoring in PSG was used as a substitute for OAHI, because in our previous study with a mean age of 27 months (ranging from 23 to 34 months), we found that the snoring time percentage of TST (63.6%) in PSG was significantly higher in children with OSAS than in children who snored but did not have OSAS (6.9%) (Markkanen et al. 2021), indicating that snoring time percentage associates with the likelihood of OSAS. In this study cohort (N = 78), the majority of the children had mild disease, with an OAHI of less than 2 events per hour (N = 75). Specifically, one child had an OAHI of 2.8 events per hour, another had an OAHI of 3.4 events per hour, and one had an OAHI of 6.3 events per hour. All of these children were included in the analysis based on the duration of snoring observed in the PSG recording. We analyzed the distribution and the quartiles of snoring time in percentages of total sleep time (TST) and found that the 75th percentile cut point was 12.10% of TST. The children snoring at least 12.10% of TST were denoted as PSG snorers (N = 19), whereas the children snoring less than 12.10% of TST were denoted PSG non-snorers (N = 59).

The demographic characteristics of the children were delineated by employing the parental reports included in the 3, 8, and 24-month questionnaires.

Accuracy of reported snoring

To evaluate the accuracy of parents’ reports concerning children’s snoring, we used questions of snoring frequency from the Sleep Disturbance Scale for Children. The answer options were always (daily), often (3–5 times per week), sometimes (once or twice per week), occasionally (once or twice per month or less), and never. In the analyses, we combined the frequency of snoring into a dichotomy as follows: snoring at least once a week (N = 21) versus less (N = 57).

Growth

The growth charts were collected from well-baby clinics. There was some variation in the number of growth measurement points between children. To evaluate the association of snoring and growth, we chose the measurement points at three, eight, 18, and 24 months corresponding to the follow-up questionnaire points. Growth values for these specific follow-up points were obtained by interpolating, or if not available, by extrapolating, the anthropometric measurements for the four fixed time points using the two nearest available measurements and assuming linear growth based on these values. These interpolated or extrapolated values were then used in the analyses (Tuohino et al. 2019). The study population´s growth was compared to that of all children in the CS cohort whose 24-month questionnaire was available (N = 947).

Body mass index (BMI) was categorized based on age and sex according to the Finnish growth references as ‘under-weight’ (with standard deviation score [SDS] equivalent <1.88 for boys and <1.79 for girls), ‘normal’ (SDS equivalent 1.88 to 1.31 for boys and 1.79 to 1.24 for girls), and ‘overweight’ (including both overweight and obese with SDS equivalent >1.31 for boys and >1.24 for girls) (Saari et al. 2011). Based on BMI-SDS information, we created a dichotomy as follows: children with obesity or overweight versus normal weight or underweight children.

Polysomnography

The PSGs of the children in the follow-up study were recorded at home, whereas those of the case–control study were recorded at the sleep laboratory. The home PSGs were recorded with an Embla Titanium device and consisted of six electroencephalography channels (F4-A1, C4-A1, O2-A1; F3-A2, C3-A2, O1-A2), right and left electro-oculography, submental EMG, oxygen saturation (pulse oximeter, Nonin), thoracoabdominal inductance plethysmography, diaphragmatic and abdominal EMG, Emfit mattress sensor, ECG, oronasal thermistor (Dymedix), and snore sensor (piezo). The nasal pressure transducer was omitted from the protocol to minimize the sleep-disturbing effects of the recording equipment (Goodwin et al. 2001).

The in-laboratory PSGs of the snoring children and their controls were recorded with an Embla N7000 device, and comprised, in addition to the signals recorded in the home PSGs, the recordings of the frontopolar EEG channels (Fp1-A2; Fp2-A1), nasal pressure transducer signal, oxygen saturation by two pulse oximeters (Nonin and Masimo), end-tidal partial pressure of carbon dioxide, sleeping position, and video. The contemporary pediatric guidelines (Troester et al. 2023) were used in the visual sleep staging and scoring of respiratory events and arousals. All of the recordings were manually analyzed by a clinical neurophysiologist.

Snoring was quantified from a PSG piezo channel. The percentage of time with snoring, referred to total sleep time, was calculated for each child. Snoring episodes lasting more than three consecutive breathing cycles were selected, based on the piezo and nasal pressure signal, and verified by listening in the in-laboratory recordings. In home PSGs, snoring was quantified based on piezo sensor alone, because they did not comprise the nasal pressure transducer or video.

Metabolic blood samples

The families were offered the option to join the study without any invasive research methods. As a consequence of this, the majority of the families refused blood samples. In conclusion, metabolic blood samples were drawn from 31 children from the polysomnography studies (PSG non-snorers N = 22, PSG snorers N= 9). When snoring times in children with and without blood samples were compared, found no statistically significant differences were found in age or snoring time between children with and without blood samples. The blood analysis was provided by the laboratory of the Finnish Institute for Health and Welfare by using enzymatic assays for measuring total cholesterol, triglycerides, and glucose; the homogenous method for direct measurement of HDL cholesterol; the ultrasensitive immunoturbidimetric assay for hs-CRP; and a chemiluminescent microparticle immunoassay (CMIA) for insulin. LDL cholesterol was calculated by the Friedewald formula. We also measured lipid and lipoprotein traits by Nuclear Magnetic Resonance (Ala-Korpela 2008). There were seven blood samples that were insufficient for the metabolic profile, and due to this, metabolic profile was available only for 24 children (PSG non-snorersN = 18, PSG snorers N = 6). The missing data was analyzed, and it showed no statistical difference.

Statistical analysis

The statistical analysis was performed using IBM SPSS statistics (IBM Corp). Dichotomized background factor variables were compared with PSG snoring with chi-squared and t-tests. The evaluation of associations between snoring and growth variables at each time point was performed by t-tests and multivariate ANCOVA. The comparisons between PSG snoring and other sleep parameters and metabolic blood samples were carried out by the Mann–Whitney U test, considering the relatively small size of the recruited population. All reported p-values were two-tailed with the statistical significance set at < 0.05. The positive and negative predictive values were based on crosstabulation of parental reported snoring and snoring based on PSG recordings and calculating confidence intervals for prevalences. Finally, multiple linear regression models were performed for further associations of metabolic test results and snoring time in PSG. The distributions of the studied values were evaluated before performing statistical modelling and also the assumptions of the models were evaluated during the modelling procedure. The statistical analyses were performed under supervision of statistician. The outcome variables were normally distributed and there were now other violations regarding the assumptions of linear regression models. In linear regression models, the residuals (errors) of the regression line were approximately normally distributed.

Results

The characteristics of the study population are shown in Table 1.

Table 1 Demographic characteristics of the children

Parental reports of snoring at least weekly were significantly related to a long snoring time in PSG at the age of 24 months (p < 0.001). The positive predictive value of the parents’ report of snoring in PSG was 57.6% (27.2–80.7%). Children who snored less than once a week based on the parents’ report had a low risk for detected snoring in PSG, as the negative predictive value was 79.7% (74.9–83.8%). The parents’ report of the child’s snoring had a fairly low sensitivity (26.3%, 9.2–51.2) but high specificity (93.2%, 83.5–98.1). Besides the parents’ report of snoring at the age of two years, there were no significant differences in the demographic characteristics between the children in the non-snorer and the PSG snorer groups.

Growth

There were no significant differences between the growth parameters among PSG snorers and PSG non-snorers excluding the boys’ weight at the age of 24 months after adjustment (see Figs. 1 and 2, Table A1). After adjusting for birth weight, the weight of snoring boys was significantly lower compared to PSG non-snorers (p = 0.041) at 24 months (Table A1).

Fig. 1

Weight for height SDS and height SDS, comparison between PSG non-snorers, PSG snorers and CHILD SLEEP cohort

Fig. 2

Height SDS, comparison between PSG non-snorers, PSG snorers and CHILD SLEEP cohort

Polysomnography

The PSG parameters of 78 children are shown in Table 2.

Table 2 Comparisons (Mann–Whitney U) between PSG non-snorers and PSG snorers for PSG parameters at 24 months of age

Sleep stage N1 was decreased in snorers (p = 0.043). Moreover, significant differences were expected to be found in snoring, and OAHI between PSG snorers and PSG non-snorers (p < 0.001).

In the polysomnography studies, the total sleep time was approximately 8 h 45 min (SD 56 min). There were no statistically significant differences in total sleep time between PSG non-snorers and PSG snorers (p = 0.213).

Metabolic blood samples

The results of the blood samples are shown in Table 3.

Table 3 Comparison (Mann–Whitney U) between PSG non-snorers and PSG snorers for metabolic blood samples at 24 months of age

We compared the median values of metabolic blood sample parameters (HDL-C, Serum-TG, ApoA1, ApoB, Glc, Insulin and HS-CRP) using the Mann–Whitney test and found no statistically significant differences between the two groups, although the difference in HDL-C was almost significant (p = 0.064).

Linear regression models

Multiple linear regression was used to test if snoring time in PSG, age, and gender significantly predicted unfavorable levels of metabolic markers (see Table 4).

Table 4 Multiple linear regression analysis with serum hs-CRP level, serum HDL level and serum ApoA1 as dependent variable

We found that snoring time in PSG was significantly associated with lower levels of HDL (β = -0.484, p = 0.007) and ApoA1 (β = -0.451, p = 0.049). In addition, snoring time was associated with higher levels of hs-CRP (β = 0.410, p = 0.019). Furthermore, the boys’ HDL levels were higher (β = 0.367, p = 0.038) and hs-CRP levels lower (β = -0.379, p = 0.031) compared to the girls’ levels. Age was not related significantly to unfavorable levels of metabolic markers.

Discussion

Although the evidence of an unfavorable effect of SDB on metabolic health has been disclosed in epidemiologic studies among adults, previous reports concerning the association of OSA and serum hs-CRP (Tauman et al. 2004; Gozal et al. 2008; Tauman et al. 2007) and lipids (Quist et al. 2016; Alexopoulos et al. 2011; Zong et al. 2013) in childhood are somewhat scarce. This present study adds to the existing knowledge covering the association of early-onset SDB and the burden of adverse metabolic factors. Consistently with previous reports, in this study, snoring duration observed in PSG showed a potential link with lower HDL and ApoA1 levels, as well as higher hs-CRP levels, among Finnish toddlers.

Previous reports covering SDB and unfavorable effects on the metabolic profile have concentrated on children with OSAS but not snoring alone (Tauman et al. 2004; Gozal et al. 2008; Quist et al. 2016; Alexopoulos et al. 2011; Zong et al. 2013; Tauman et al. 2007). The diagnostic criteria for OSA are defined as one or more obstructive events per hour of sleep or hypoventilation together with snoring, paradoxical thoracoabdominal movement, or flattening of the nasal airway pressure waveform, implying flow limitation (International 2014). In our study cohort, there were seven children with OAHI > 1 and only one child with OAHI > 5, indicating a fairly mild disease in the majority of children. All of the children with OAHI > 1 were in the group of PSG snorers. In addition to OSAS, primary snoring has been associated with dyslipidemia among obese adults (Zhang et al. 2017). To the best of our knowledge, this is the first study to report the association of snoring and disadvantageous changes in markers of metabolic health among young children.

The prevalence of obesity has increased worldwide in recent decades (Ng et al. 2014). Previously in the Finnish population, the same trend of increasing body-mass index (BMI) was reported in school-aged children and teens, but not among toddlers (Vuorela et al. 2011). In a moderately recent study, the prevalence of overweight was 23.7% among Finnish 2–6-year-old boys and 13.9% among girls of the same age group based on the age and sex adjusted body mass index (ISO-BMI) (Mäki et al. 2018). However, this stable trend in overweight among toddlers seems to be changing in Finland as well. In the latest report of the Finnish Institute for Health and Welfare, the proportion of overweight among 2–6-year-old children increased and was 27% among boys and 16% among girls (Jääskeläinen et al. 2020). In our cohort, 18.6% of boys and 14.3% of girls were overweight or obese. These numbers are somewhat lower than those reported two years ago in Finland (Jääskeläinen et al. 2020) and relatively close to the level reported in 2018 (Mäki et al. 2018). This was rather expected since our study population was born between 2011 and 2013. There was no association between the snoring time and overweight in our study population. This is mainly due to the moderately short follow-up time and relatively small number of children in our study. It would be interesting to examine the association of overweight and SDB repeatedly at present in a larger sample, as the proportion of toddlers with overweight is presumably higher.

In this present study, there were no significant differences between the growth parameters among PSG snorers and PSG non-snorers. Growth restriction has been previously reported mainly among children with markedly greater OAHI values and thereby a more serious disease (Nieminen et al. 2002). As the children in our study in the group of PSG snorers were suffering from a fairly mild disease, the missing association of SDB and growth failure was not completely unexpected. Furthermore, as growth during infancy and early childhood is largely nutrition-dependent and growth hormone plays a role somewhat later (Res Paediatr et al. 2017), the effect of abnormal nocturnal growth hormone secretion could presumably be seen among older children.

In the PSG parameters, significant differences were expected to be found in snoring and OAHI between PSG snorers and PSG non-snorers. Furthermore, sleep stage N1, which is the transition period from being awake to falling asleep, was decreased in snorers. This is somewhat unforeseen, since in previous studies primary snoring has been associated with increase in sleep stage N1 (Zhu et al. 2014). Snoring causes awakenings and restless sleep. Due to this, the sleep pressure might increase, causing a decrease in stages with light sleep. However, we did not find a statistically significant increase respectively in the deeper stages of sleep (sleep stage N2 and N3). In conclusion, the association with snoring and changes in the architecture of sleep needs further study.

Exposure to various environmental risk factors for cardio-metabolic health during early childhood can increase the possibility of metabolic disease in adulthood (Berenso 2002). Cardiovascular and metabolic risk in young children consists of components such as elevated glucose and blood pressure, obesity, and dyslipidemia (Kamel et al. 2018). The metabolic syndrome definition consists of the presence of at least three of the following measures: excess body fat around the waist, elevated plasma triglycerides, elevated fasting blood glucose, high blood pressure, and decreased HDL cholesterol (Grundy et al. 2005). Definitions of metabolic syndrome with cut-points are defined for children aged 10 years and older (Zimmet et al. 2007) but not for younger children. The National Cholesterol Education Program (NCEP) Expert Panel has defined the cut-off values for pediatric lipid concentrations (de Jesus 2011). An acceptable concentration of HDL is defined as 1.2 mmol/l or more and the borderline level is 1.0–1.2 mmol/l. Levels below 1.0 mmol/l are considered abnormal values. In our study, among the 31 children with blood samples, there were five children (16%) whose serum HDL concentration was lower than 1.0 (0.92–0.98). Of these five children, three were in the PSG snorer group. Eight children (26%) had HDL concentrations at the borderline level and half of these were PSG snorers. Some 58% (N = 18) of the children had HDL concentrations at the normal level.

In our study, at the age of 24 months the overall levels of HDL cholesterol were higher (girls’ mean 1.17, SD 0.19, boys’ mean 1.31, SD 0.23) compared to children aged 36 months in the Special Turku Coronary Risk Factor Intervention Project (STRIP) study, which is a longitudinal Finnish prevention trial (girls’ mean 1.07, SD 0.20, boys’ mean 1.12, SD 0.22) (Kaitosaari et al. 2003). In addition, interestingly in the linear regression models, males had a healthier profile of metabolic markers. Boys’ HDL levels were higher and hs-CRP levels lower compared to those of girls. The amount of physical activity and dietary aspects have effects on children’s serum lipid concentrations (Kallio et al. 2021). One reason for this healthier metabolic profile among boys could be explained by the higher amount of physical activity. Unfortunately, in our study cohort, neither the comprehensive food records nor information on the children’s daily physical activity were available.

Despite the lacking statistically significant differences in HDL concentrations between the PSG snorers and non-PSG snorers in our study, snoring time in PSG was associated with lower levels of HDL and ApoA1 in the multiple linear analysis. It is remarkable that this kind of phenomenon can be seen already early in childhood and arouses interest as to whether the timely management of SDB in childhood could offer an efficient tool to prevent metabolic disorders in adulthood.

Strengths and limitations

Our study population was composed of 78 children from the CHILD-SLEEP birth cohort. Snoring was detected by PSG, which is the gold standard to assess SDB. One limitation of this present study is that the number of children in our study was somewhat small. PSG is an overnight study requiring plenty of resources, and it is rather difficult to accomplish in large study populations. In addition, due to ethical aspects, families were offered the option to join the study without any invasive research methods, and several families chose to decline providing blood samples. This explains the small number of blood samples in our study. It is noteworthy that in this present study, the association of snoring and unfavorable effects on the metabolic profile was established regardless of our small sample size. However, these results are preliminary, and in further studies, larger samples are needed in order to verify the findings in our study.

We have previously reported that in the CS cohort, children with a recurrent cycle of infection snored more often habitually compared to children without cyclic infections (Katila et al. 2021). However, in this study, we did not statistically control for respiratory infections because our only available data was a subjective estimation based on parental reports. Additionally, to our knowledge, there is no strong evidence to suggest that respiratory infections are associated with children's metabolic profiles, which was the main focus of this study—specifically, analyzing how snoring is related to metabolism in children.

As there is currently no consensus on the objective measurement or scoring of snoring, variability in scoring criteria, equipment selection, and placement based on local protocols are common. In our study, we used a piezo snoring sensor, with validation through listening when feasible. The piezo signal was selected because nasal prongs were excluded from the home recording protocol to minimize potential sleep disturbances. Given the lack of universally accepted scoring rules for children's snoring, the scoring of snoring periods was largely guided by the considerable clinical experience of our scorers, who are trained clinical neurophysiologists.

Conclusion

In conclusion, there were no significant variances observed in the growth parameters among children with longer snoring durations in PSG recordings when compared to those with shorter snoring durations. The duration of snoring identified in the PSG recordings predicted lower levels of HDL and ApoA1, as well as higher levels of hs-CRP among Finnish toddlers. Accordingly, snoring during early childhood may have the potential to negatively impact the serum metabolic profile; consequently, it could contribute to the risk of cardiovascular diseases in adulthood and needs further research. Early recognition and treatment of childhood SDB might provide a tool for preventing subsequent metabolic disease.