Participants

Seventeen adult patients surgically treated for NFMA and seventeen age-matched healthy controls underwent a single night of ambulatory polysomnography [9]. All patients were otherwise healthy and received yearly follow-up by an endocrinologist. Exclusion criteria were age > 65 years, use of psychotropic or cardiac medication, diagnosis of a sleep disorder, hypertension, dyslipidemia, or diabetes mellitus. HRV data from two patients and two controls were excluded for insufficient data quality for this analysis (detached HR electrode), thus fifteen patients and fifteen healthy controls were studied.

Markers of circadian rhythmicity in these patients and their methods were previously published [9,10], and included intradaily variability, skin temperature, and melatonin secretion profiles. The intradaily variability, derived from 7 days of actigraphy, quantifies how fragmented the rhythm of motor activity is relative to its 24-h amplitude; more frequent alterations between an active and an inactive state lead to a higher intradaily variability [11]. Proximal skin temperature was measured for 24 hours at both infraclavicular areas, supra-umbilical, and on the left mid-thigh. Its diurnal variation is related to increased sleep latency in both narcoleptics and healthy persons [12,13], and to sleep depth [14]. In our previous study differences in proximal skin temperature between NFMA patients and controls were more pronounced than those in distal skin temperature or core body temperature [10]. Melatonin secretion was measured for 36 hours using at least thirteen saliva samples. An altered melatonin profile was defined as the absence of an evening rise or daytime values >3 pg/mL in a 36-hour salivary melatonin collection. We refer to S1 Table for the values of these markers for each patient.

The ethical committee of the Leiden University Medical Center approved the study. All subjects gave written informed consent.

HRV

Methods of polysomnography [9] and calculation of HRV [15,16] have been detailed previously. In short, sleep stages, apnea/hypopnea events and leg movements were manually scored in 30-second epochs by an experienced sleep technician. All HR data from nocturnal sleep onset until morning awakening were selected. A continuous wavelet transform was implemented in Matlab (Version 13.1, Mathworks, MA, USA) to detect R-peaks in the 30-second epochs. Outliers, differing > 25 beats per minute from adjacent samples, were excluded. Heart beat data were resampled at 5 samples per second. The full spectrum of HRV in the region surrounding each sample was calculated with a fast Fourier transform using a window of 512 samples, ranging 256 samples left and right of a particular sample. After averaging these spectra for each 30-second epoch, the LF (0.04–0.15 Hz) and HF (0.15–0.4 Hz) power components per epoch were calculated. The LF band is thought to predominantly reflect the baroreflex-mediated sympathetic activity [16] whereas the HF band is thought to represent an index of vagal activity, and the LF/HF ratio an index of the sympathovagal balance [17]. All measures have their limitation and should be interpreted cautiously. To account for the autonomic effects of arousals, we identified all epochs with apnea, hypopnea, leg movements, or transitions from NREM3 to NREM1/2 or NREM2 to NREM1. The raw data file on HR and HRV is available in the S1 Data File.

Statistical analysis

Patients’ clinical characteristics were compared with controls using the Student’s t-test for numerical data (or in case the assumption of normality was not met (Shapiro-Wilk test) the Mann-Whitney U test), and Pearson’s χ² for categorical data. Mixed effect regression models were used to analyse the effect of disease on HR, LF, HF, and LF/HF. This model included all 30 second epochs. Epochs were classified as rapid eye movement- (REM), non-REM stage 1- (NREM1-), NREM2-, and NREM3-sleep. Disease, wake and sleep stages, leg movements, apnea events, gender, and arousal transitions were selected as fixed effects and participants as random effect. A natural logarithm-transformation was applied to obtain a normal distribution of the model’s residuals. Differences were considered statistically significant at P < 0.05. In a secondary analysis, the interaction between disease and each of the sleep stages was added to the model, to explore sleep stage-dependent differences in HRV. To correct for multiple testing after introducing the interaction terms (five sleep stages), sleep stage-dependent differences were considered statistically significant at P < 0.01.

Participants

Median age was 58 yr (range 26–65 yr) for NFMA patients (8 females), and 52 yr (range 30–63 yr) for the control subjects (5 females) (Table 1). Prior to surgery, most patients had visual field defects (80%) or suprasellar extension on MRI (93%). Hypopituitarism was present in eleven (deficiency of growth hormone in ten, of corticotropin in five, of thyrotropin in ten, and of the gonadotropins in five); all received hormone replacement, except for optional growth hormone replacement, which was left untreated in three patients. Healthy control subjects were not significantly different in terms of gender, age, or BMI. Compared to controls, patients spent more time awake (P = 0.005) and in NREM1-sleep (P < 0.001), and less in REM-sleep (P < 0.001). Mean duration of apnea was shorter in patients (13.4 ± 9.1 seconds vs. 21.8 ± 8.0 seconds, P = 0.008).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Clinical characteristics of patients with NFMA and healthy age-matched control subjects.

https://doi.org/10.1371/journal.pone.0152390.t001

HRV

We included an average total number of 1374 epochs per participant in our model. Patients did not differ from their age-matched controls during the night with respect to HR (P = 0.137), LF (P = 0.205), HF (P = 0.959), or LF/HF (P = 0.184) (Table 2). As differences might be dependent on patients with the most prominent signs of circadian dysrhythmicity, e.g. those with disturbed melatonin secretion, this feature was added as an additional coefficient to the model, but lacked significance. Fig 1 displays the results from the secondary analysis, exploring sleep stage—dependent differences in HR and HRV between patients and controls. LF/HF was slightly lower in patients than controls (difference: -0.44 [standard error: 0.20], P = 0.040), but this difference did not reach statistical significance after correction for multiple testing (P > 0.01).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Effects of mixed effects regression model coefficients on heart rate and heart rate variability.

https://doi.org/10.1371/journal.pone.0152390.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Heart rate variability in patients and controls.

Heart rate, low- and high frequency power and their ratio, in 15 NFMA patients and 15 healthy controls, stratified for wake and sleep stages. Data represent mean ± SD, calculated over averages for each subject. WBSO, wake before sleep onset; REM, rapid eye movement sleep; LF, low frequency power in ln(ms²/Hz); HF, high frequency power in ln(ms²/Hz). None of the differences reached statistical significance after correction for multiple testing (P < 0.01) in our secondary analysis.

https://doi.org/10.1371/journal.pone.0152390.g001