Exercise Reveals the Interrelation of Physical Fitness, Inflammatory Response, Psychopathology, and Autonomic Function in Patients With Schizophrenia

Abstract

Maintaining and improving fitness are associated with a lower risk of premature death from cardiovascular disease. Patients with schizophrenia are known to exercise less and have poorer health behaviors than average. Physical fitness and physiological regulation during exercise tasks have not been investigated to date among patients with schizophrenia. We studied autonomic modulation in a stepwise exhaustion protocol in 23 patients with schizophrenia and in matched controls, using spirometry and lactate diagnostics. Parameters of physical capacity were determined at the aerobic, anaerobic, and vagal thresholds (VT), as well as for peak output. VT was correlated with psychopathology, as assessed by the Positive and Negative Syndrome Scale, with the inflammatory markers IL-1β, IL-6, and TNF-α and with peak output. The MANOVA for heart and breathing rates, as well as for vagal modulation and complexity behavior of heart rate, indicated a profound lack of vagal modulation at all intensity levels, even after the covariate carbon monoxide concentration was introduced as a measure of smoking behavior. Significantly decreased physical capacity was demonstrated at the aerobic, anaerobic, and VT in patients. After the exercise task, reduced vagal modulation in patients correlated negatively with positive symptoms and with levels of IL-6 and TNF-α. This study shows decreased physical capacity in patients with schizophrenia. Upcoming intervention studies need to take into account the autonomic imbalance, which might predispose patients to arrhythmias during exercise. Results of inflammatory parameters are suggestive of a reduced activity of the anti-inflammatory cholinergic pathway in patients, leading to a pro-inflammatory state.

Introduction

Regular physical exercise is known to induce profound effects on physiological and psychological processes in humans. Remarkable results of physical exercise were demonstrated for depressive mood, for anxiety, and for cognitive performance in patients at risk for dementia. ¹ It was even suggested that exercise interventions should be explored as a potential strategy for delaying disease onset of dementia. In addition, cardiorespiratory fitness is a strong and independent mortality predictor for humans. ²

Both one-time aerobic exercise tasks and regular exercise training induce a wide range of biochemical and physiological changes. Apart from the beneficial influence on cardiovascular function and obesity, the influence of exercise on mood, on neurotrophic factor levels, and on stress reactivity is well known. ¹ , ³ In addition, various interventional studies have shown the beneficial effects of regular training as an add-on treatment for patients with major depression and anxiety disorders. ⁴

People living with schizophrenia are known to exercise less and have poorer health behaviors compared with the general population. In addition, these patients report high incidence rates of metabolic syndrome and coronary heart disease. ⁵ Thus, regular physical exercise might be an ideal supplementary treatment provided patients could overcome negative symptoms and a lack of motivation. Recent reviews described 3 randomized control trials in these patients and concluded that regular exercise programs are possible and that they can have beneficial effects both on physical and mental health. ⁶ , ⁷ However, the authors concluded that larger randomized studies were needed before any final conclusion could be drawn.

We have previously described profound changes of physiological regulation in patients with schizophrenia. A substantial body of evidence shows reduced heart rate variability (HRV) in patients with schizophrenia. ⁸ , ⁹ This was described in the acute ¹⁰ , ¹¹ and chronic states. ¹² While a high degree of HRV indicates healthy cardiac regulation, decreased HRV is associated with a high incidence of cardiac morbidity and mortality and may predispose patients to cardiac events such as ventricular fibrillation. In contrast to autonomic dysfunction in primary medical conditions such as diabetes or heart failure, decreased HRV in patients with schizophrenia is most likely caused by reduced interaction between prefrontal brain areas and subcortical structures, leading to arousal and an autonomic imbalance. This is characterized by vagal withdrawal and sympathetic predominance. ¹³ In addition, increased breathing rates at rest associated with positive symptoms and reduced cardiorespiratory coupling were shown in patients with schizophrenia. ¹⁴ , ¹⁵

Thus, in this study we aimed to prepare the basis for upcoming exercise intervention studies by identifying suitable parameters for aerobic exercise intensity training for patients with schizophrenia. In addition, we intended to describe physiological changes in patients with schizophrenia during a physical task. We hypothesized that we would find reduced HRV and increased breathing rates for all intensity levels, reflecting altered physiological regulation in patients. We also controlled for smoking severity by means of carbon monoxide measurements. ¹⁶ In addition, we hypothesized a reduction in physical fitness in relation to various physiological respiratory and metabolic thresholds, such as the aerobic threshold, the individual anaerobic threshold, and the vagal threshold (VT) indicating the point at which a disappearance in vagal activity occurs. ¹⁷ During this exercise task, emphasis was primarily placed on VT. We hypothesized VT to be reduced in patients, correlating with psychopathology of the disease. In addition, based on studies of the cholinergic anti-inflammatory pathway showing that the cholinergic system suppresses inflammation, we hypothesized that reduced VTs would be associated with an augmented inflammatory response as reflected in levels of interleukin-1β (IL-1β), interleukin-6 (IL-6), or tumor necrosis factor-alpha (TNF-α) after a one-time exercise task. ¹⁸

Methods

Subject and Exercise Protocol

Twenty-three patients suffering from paranoid schizophrenia and 23 healthy controls matched with respect to age, sex, weight, smoking habits, and education were included in the study (see table 1). Inpatients receiving a fixed dose of antipsychotic treatment were studied. Serum drug levels were controlled for legal drugs (eg, antidepressants and benzodiazepines) and illegal substances (eg, cannabis). In accordance with our inclusion criteria, only subjects with negative results were included in the study. A clinical electrocardiography was recorded prior to the investigation and evaluated by a cardiologist. Similarly, patients were clinically assessed for the presence of any other somatic diseases (eg, heart diseases). Diagnosis of paranoid schizophrenia was established when patients fulfilled DSM-IV criteria (Diagnostic and Statistical Manual of Mental Disorders, 4th ed.). The semi-structured clinical interview SCID-1 was used to approve the clinical diagnosis. Psychotic symptoms were quantified using the Positive and Negative Syndrome Scale (PANSS) ¹⁹ . Control subjects were recruited from hospital staff (n = 8), medical students (n = 3), and the local community (n = 12). Similarly, an interview (SCID-1) and a clinical investigation were performed on controls to exclude any potential psychiatric or other diseases, as well as any interfering medications. Control subjects were matched for total physical activity per week using the short version of the international physical activity questionnaire (IPAQ-total) in metabolic equivalent (MET) minutes per week. The IPAQ-total included intensive, moderate, and walking physical activities. Furthermore, the sitting time per week (IPAQ-sitting) was measured. ²⁰

Table 1.

Epidemiological Data of Participants

	Controls Mean ± SD	Patients Mean ± SD	P value
Epidemiological Data
Gender (female/male)	6/17	7/16	n.s.
Age (years)	28.2 ± 4.1	28.4 ± 5.3	n.s.
BMI (kg/m2)	23.6 ± 2.7	24 ± 3.7	n.s.
Education
8 years at school	0	2	—
10 years at school	3	5	—
12 years at school (A-Level)	20	16	—
Sport
No sport	11	9	—
<2h/week	5	6	—
2–5h/week	5	6	—
>5h /week	2	2	—
Smoking Behavior
Smokers	19	20	n.s.
<5 Cigarettes/day	5	5	—
5–10 Cigarettes/day	9	6	—
>10 Cigarettes/day	5	9	—
Fagerström test for nicotine dependence	0.61 ± 1.1	3.9 ± 2.9	P <.001
International Physical Activity Questionnaire (IPAQ)
IPAQ-total (MET minutes/week)	2640 ± 2597	1838 ± 2102	n.s.
Vigorous physical activity (h/week)	1.41 ± 1.9	1.1 ± 1.4	n.s.
Moderate physical activity (h/week)	1.9 ± 2.0	1.5 ± 2.2	n.s.
Time spent sitting (min/day)	353 ± 212	316 ± 312	n.s.
Walking distance (METs min/week)	1053 ± 999	1143 ± 1983	n.s.
Disease-related Data
Disease duration (years)	—	4.1 ± 3
Age of onset	—	24.1 ± 6.2
PANNS	—	74 ± 29
Medication
Neuroleptics		19/19
Quetiapine (300–500mg)	—	7
Olanzapine (5–10mg)	—	6
Amisulpride (100–200mg)	—	6

Note: SD, standard deviation; BMI, body mass index; IPAQ, International Physical Activity Questionnaire; PANSS, Positive and Negative Syndrome Scale.

P values indicate results of t tests.

This study complied with the Declaration of Helsinki. All participants gave written informed consent to a protocol approved by the Ethics Committee of the University Hospital, Jena. Patients were informed that refusal to participate in this study would not affect future treatment. Patients were only included after a psychiatrist (S.B.) certified their ability to give full consent to the study protocol. Patients unable to provide informed consent were not included. In accordance with our laboratory procedures, participants were asked to refrain from smoking, exercising, or any eating 2h prior to the exercise assessment. Carbon monoxide was measured in the exhaled breath of all participants (CO-Check, Neomed GmbH, Korschenbroich, Germany) just before the exercise testing was started. ¹⁶

Exercise Testing Protocol

The body mass index (BMI) was calculated prior to each test. Participants received instruction regarding the operational aspects of the exercise test and the tests were begun once individuals indicated that they understood the information.

Exercise testing was performed in the upright position with an electronically braked cycle ergometer (Ergometrics 900, Ergoline, Bitz, Germany) with its seat and handlebar heights correctly set for each subject. After a resting period of 5min, the incremental bicycle protocol started with the subject pedalling at 15W for 3min. The power output was then increased by 15W for every 3min until the subject reached his or her limit of tolerance. Exhaustion levels were determined by measuring maximum lactate levels (LaPeak) and using the respiratory exchange ratio (RER) of carbon dioxide (CO₂) output to oxygen (O₂) uptake that are known to be the most reliable indicators of maximal effort.

The degree of effort exerted by the participants was further determined using the standardized subjective exhaustion Borg Scale. The attainment of RER_max ≥ 1.05 and a BORG rating of perceived exertion >16 were regarded as indicators of sufficient effort. We encouraged all subjects to aim to maintain a pedalling frequency of 70–80 revolutions per minute throughout the test session. After a patient’s individual limit of tolerance was reached, a cooling down phase ensued, consisting of 3min pedalling at a slow rate (<40 revolutions/min) at a power output of 15W and the participant remained seated for another 7min without pedalling.

Physiological Regulation Parameters

During the resting period and the cycling test, ventilatory indices and gas exchanges were measured continuously on a breath-by-breath basis using an automatic mobile wireless ergospirometer (MetaMax 3B, Cortex, Leipzig, Germany). Before each test, the turbine (flow and volume) was calibrated with a syringe (Hans Rudolph Inc, Kansas City). The gas analyzers were calibrated according to the manufacturer’s guidelines with the same certified calibration gas mixture of 5% CO₂ and 15% O₂ (Air Liquide Healthcare America Corporation, Plumsteadville, PA). The breath-by-breath gas exchange and ventilatory data were interpolated to give second-by-second values. For data analysis, the second-by-second values were time averaged for 15 s and the average value was aligned with the center of the time interval.

Peak oxygen consumption (VO_2peak), peak minute ventilation (VE_peak), and peak respiratory exchange ratio (RER_peak) were defined as the mean of the two highest consecutive values of 15-s averages occurring during exercise (see table 2). The peak power output (P _peak) was defined as the highest work rate that was sustained for 3min during the test. When participants were not able to cycle to the end of the last 3-min interval (in two cases), P _peak was linearly interpolated based on the proportion of the time completed during the terminal stage.

Table 2.

Explanation of Applied Parameters

		Parameter Abbreviation	Parameter	Background of Calculation	Significance
Heart rate variability (HRV)		HR	Heart beats per minute
	Time domain	RMSSDHRV	Root mean square of differences of successive heart beats	Standard deviation	Vagal modulation
		RSA	Respiratory sinus arrhythmia	Variation of heart rate that occurs during a breathing cycle	Vagal modulation
	Poincaré plot analyses	SD1	Standard deviation of the instantaneous RR-interval variability	Standard deviation	Vagal modulation
	Poincaré plot analyses	SD2	Standard deviation of long-term RR-interval variability	Standard deviation
	Complexity measure	HCHRV	Compression entropy	Nonlinear data compression	Vagal modulation
Parameters of physical fitness		AerT	Aerobic threshold	Power output during initial increase in blood lactate concentrations above resting values	Reflects physical fitness
		AnT	Anaerobic threshold	Power output at a lactate concentration of 1.0 mmol/l above lactate concentration at AerT	Reflects physical fitness
		P peak	Peak power output	Maximal power output of participants	Reflects physical fitness
		VT	Vagal threshold	Power output at which no subsequent decline in heart rate variability occurs (measure: SD1)	Vagal modulation
		VO2peak	Peak oxygen consumption	Maximum capacity of an individuals body to transport and use oxygen during incremental exercise	Reflects physical fitness
Parameters of spirometry		VE	Minute ventilation or tidal volume	Total volume of gas entering the lungs per minute	Key indicator for lung function
		RER	Respiratory exchange ratio	Ratio between the amount of CO2 exhaled and O2 inhaled in one breath	Index of energy consumption and effort

Capillary blood samples of 20 µl were obtained from the ear lobe when the patient was at rest, and again after the end of each exercise level. The lactate concentration in millimoles per liter were measured by the EBIO basic system analyzer, using an enzymatic-amperometric measuring system (Eppendorf, Hamburg, Germany).

Physical Fitness Parameters: Aerobic and Anaerobic Thresholds

Maximal parameters (ie, maximal oxygen uptake and maximal power output) attained during exercise testing are the most frequently applied indicators of endurance capacity and physical fitness. The determination of these maximal performance parameters requires adequate motivation and maximal effort of participants. Therefore, in addition to the measurement of VO_2peak and P _peak we assessed the aerobic threshold (AerT) and the anaerobic threshold (AnT) to determine aerobic fitness. Both thresholds are common in clinical exercise testing and are known to be reproducible submaximal indicators of endurance and physical fitness. In contrast to the case of peak parameters, the determination of these thresholds offers the great advantage that maximal effort and motivation in subjects are not mandatory.

AerT represents the first increase in blood lactate concentrations above resting values during incremental exercise and demarcates the upper limit of a range of exercise intensities (moderate exercise domain) that can be accomplished almost entirely aerobically (table 2). In contrast, AnT describes the maximal lactate steady state and represents the exercise intensity above which a continuous increase in blood lactate is unavoidable during incremental exercise. The AnT indicates the transition between heavy and very heavy exercise domains.

AerT and AnT were determined according to the method described by Dickhuth and Roecker. ²¹ A special software (ERGONIZER, Freiburg, Germany) was used for investigator-independent calculation and was based on an equalizing SPLINE interpolation procedure. The AnT was defined as the power output at a lactate concentration of 1.0 mmol/l above the lactate concentration at AerT. The power output that corresponds to the occurrence of the AerT was corrected for the exercise time with the following formula:

P AerT_LA = P_com +(t_Th/180×15)

where P _com is the last completed work rate (W) before occurrence of the AerT and t _Th is the time of the incomplete work rate (s) until AerT occurs. We proceeded similarly with the power output at the AnT.

Autonomic Parameters of Heart Rate and Respiration

The LifeShirt (Vivometrics, Inc, Ventura, CA), a multi-function ambulatory device consisting of a Lycra garment, a data recorder a computer-based analysis software (VivoLogic) was used. ²² Respiratory inductive plethysmography, a core method which has been demonstrated to provide an accurate non-invasive assessment of respiratory patterns, was employed. In order to obtain the ventilated volume for further analysis, the device was calibrated in two different successive steps, as recommended by the manufacturer. Initially, a qualitative diagnostic calibration procedure was performed using a 5-min period of calm breathing to compute a calibration factor. A second calibration procedure was conducted based on defined breaths of a fixed volume. The breathing rate was defined as the number of breaths per minute (Br/min).

Heart rate time series consisting of successive beat-to-beat intervals (RRI) were extracted from the raw data records. Afterwards, these time series were filtered by applying an adaptive variance estimation algorithm to remove and interpolate ventricular premature beats and artifacts (eg, movement, electrode noise, and extraordinary peaks).

Parameters of Time and Frequency Domain of Heart Rate

We obtained the basic heart rate (HR) and the RMSSD (root mean of squared successive difference) as a time domain parameter of HRV (table 2). ²³ The low frequency (LF_HRV, 0.04–0.15 Hz) and high frequency parameters (HF_HRV, 0.15–0.40 Hz) of the frequency domain were not calculated, because previous reports have shown that these measures are not conclusive during exercise, due to a break-down of the power spectrum. ²⁴ In addition, a Poincaré plot analysis (PPA), respiratory sinus arrhythmia (RSA), and compression entropy (HC) were calculated.

Poincaré Plot Analyses of Heart Rate

PPA is based on a technique derived from nonlinear dynamics and represents the nature of heart beat time series fluctuations. The Poincaré plots are 2-dimensional visual representations (scatter plots) of each RR-interval value in the time series, plotted against the subsequent RR-interval value. The shape of the plot is assumed to be influenced by changes in vagal and sympathetic modulation. Typically, PPA shows an elongated cloud of points aligned along the characteristic line. To illustrate this graphically, an ellipse showing the shape of the cloud of points can be drawn into the plot, such that the center of the ellipse is the mean RR. The standard deviation of the instantaneous RR-interval variability (minor axis of the cloud—SD1 in ms) and the standard deviation of long-term RR-interval variability (transverse diameter of the ellipse—SD2 in ms) can be calculated (table 2). ²⁵

Respiratory Sinus Arrhythmia

The RSA was quantified as a measure of cardiac vagal activity (table 2). This is a cardiorespiratory phenomenon characterized by HR fluctuations that are in phase with inhalation and exhalation. ²⁶ The LifeShirt uses the peak-to-valley method for assessing each breathing cycle and assesses the HR simultaneously.

Compression Entropy of Heart Rate

Besides linear computational algorithms, nonlinear methods were applied. Here, the entropy (complexity) of a given data set is defined as the smallest algorithm that is capable of generating the data. Although it is theoretically impossible to develop such an algorithm, data compressors represent a sufficient approximation. In this study, we applied the LZ77 algorithm for lossless data compression developed by Lempel and Ziv (1977). They introduced a universal algorithm (ZIP) for lossless data compression using string matching on a sliding window. This algorithm is widely used and implemented in many file compressors such as Winzip. With some modifications, this algorithm can be applied for the analysis of heart beat time series as introduced by Baumert and colleagues. ²⁷ The ratio of the compressed to the original time series length represents an index of entropy and is referred to as HC. Therefore, HC indicates to which degree data from HR time series can be compressed using the detection of recurring sequences. The more frequent certain sequences occur—and therefore the more regular these series are—the higher the compression rate. Thus, reduced variability (reduced short-term fluctuations) of HR results in increased compression and smaller HC values. Entropy reduction appears to reflect a change in sympathetic/parasympathetic HR control. ²⁷

Determination of Vagal Threshold

The RR intervals from the last 2min of rest and each stage of exercise were used for VT determination. For calculation of the VT, values of the instantaneous variability in the RR intervals (SD1) from the Poincaré plot (see above) were plotted in relation to power output using Matlab 2009 (The Math Works Inc, Natick). The deflection point (DP) is the point in time at which no subsequent decline in HRV occurs (see figure 3A for illustration), indicating the moment when parasympathetic activity has decreased and sympathetic activation will increase, in such a way that almost no vagal modulation remains. DP was determined when the HRV dropped below a threshold (mean value of the HRV plus 3 times the standard deviation during the last 30% of the “exercise time”). In this time period (during the last 30% of the test), the decline in HRV was completed and there were almost no variations in HRV for all subjects observed.

Fig. 3.

The calculation of the vagal threshold is depicted in A, using data of one patient and one control subject for illustration. The deflection point (DP) is the point in time at which no subsequent decline in heart rate variability occurs. DP was determined when the heart rate variability dropped below a threshold (mean value of the heart rate variability plus 3 times the standard deviation during the last 30% of the “exercise time”). In this time period (the last 30% of the test), the decline in heart rate variability was completed and there were almost no variations in heart rate variability among all subjects observed. The negative correlation between positive symptoms and vagal threshold is shown in (B), indicating that a reduced vagal threshold is associated with an increased amount of positive symptoms. (C) Shows the association of a reduced VT with increased TNF-α values after the exercise task.

Blood Counts and Tests

Blood samples were collected from an antecubital vein at rest and 10min after the end of the exercise test. For evaluating the cell counts of erythrocytes, platelets and leucocytes, a commercial blood cell counter (Act-Diff; Beckman Coulter GmbH, Krefeld, Germany) was used. The hematocrit was determined before and after exercise to evaluate changes in plasma volume and was used for correcting confounds caused by fluid loss after exercise. After centrifugation, aliquots of plasma were snap-frozen and stored at −80°C until cytokine levels could be determined. Enzymelinked immunosorbent assays (ELISAs) for the cytokines IL-1β (R&D Systems, Minneapolis, MN, USA), IL-6 (Beckman-Coulter, Krefeld, Deutschland), and TNF-α (R&D Systems, Minneapolis, MN) were used. All samples were measured with the same assay for each cytokine. ELISAs for IL-1β and TNF-α were quantified using a Dynatech MR 4000 (Dynex, Denkendorf, Germany). IL-6 was assessed using the Access Immunoassay System (Beckman Coulter, Krefeld, Germany). The intra- and interassay variation coefficients for the ELISA kits were below 5% and 10%, respectively.

Statistical Analyses

For statistical analysis, SPSS for Windows (version 17.0) was used. All parameters were tested for normal distribution with the Kolmogorov-Smirnov test. Collected data were investigated as follows: first, we performed a multivariate analysis of variance (MANOVA) for repeated measures, including obtained physiological parameters. Second, we performed a MANOVA focused on different physiological thresholds with respect to power output of patients and controls. In a last analysis, the relation of fitness parameters to psychopathology and inflammatory values was investigated.

The repeated measures MANOVA was performed to investigate an overall effect for the factors GROUP (patients and controls), POWER OUTPUT (baseline, 15W, 30W, 45W, 60W, 75W, and 90W) and GROUP × POWER OUTPUT interaction for the physiological parameters (HR, breathing rate, RMSSD, SD1, SD2, RSA, and HC). Repeated measures ANOVAs were calculated to analyze the effect of POWER OUTPUT, GROUP, and GROUP × POWER OUTPUT interaction for single parameters. At each workload-step, a planned post hoc t test was calculated to compare patients and controls for descriptive analysis. The post-test resting condition (recovery) was not included in any analysis, because peak power output was significantly different between patients and control subjects. The MANOVA was repeated in terms of a MANCOVA using the carbon monoxide concentration of exhaled air as a covariate to remove the influence of smoking from the results.

A second MANOVA was performed applying the factor GROUP (patients and controls) and including parameters of physical capacity (normalized to weight) at the AerT, AnT, and VTs, as well as for peak output. ANOVAs were calculated for each parameter. The second MANOVA was repeated in terms of a MANCOVA using the carbon monoxide concentration of exhaled air as a covariate to analyze the influence of smoking on the physical capacity of patients and control subjects. In addition, we performed a comparison of fitness parameters (VT, AerT, AnT) with respect to medication by means of a simple t test for descriptive analysis. Blood counts and interleukins were compared by means of independent t tests for descriptive analysis (table 3). Spearman’s rank correlation coefficient was calculated to investigate a potential correlation of VT of patients with PANNS scores, interleukins, and peak-power output. Similarly, we correlated the duration of the disease with parameters of physical fitness.

Table 3.

Physiological Parameters Obtained in the Study

Parameters	Controls Mean ± SD	Patients Mean ± SD	P value
Borg90W	14 ± 2	15±2	n.s.
Carbon monoxidepre (parts/million)	9.3 ± 9.9	26.0±25.0	P < .005
Carbon monoxidepost (parts/million)	7.0 ± 5.2	19.0 ± 15.5	P < .001
Lactatebaseline (mmol/)	1.02 ± 0.61	0.98 ± 0.56	n.s.
Lactate90W (mmol/l)	2.5 ± 1.78	2.42 ± 1.42	n.s.
pHbaseline	7.45±0.02	7.45 ± 0.02	n.s.
pH90W	7.42±0.02	7.43 ± 0.03	n.s.
Blood Gases
pCO2/baseline (mmHg)	38.5±3.4	38.6±3.2	n.s.
pCO2/90W (mmHg)	36.9±4.5	37.6±4.2	n.s.
pO2/baseline (mmHg)	78.5±15.9	75.2±14.0	n.s.
pO2/90W (mmHg)	81.3±6.1	78.2±6.6	n.s.
HCO3 − (mmol/l)	27.6±2.4	27.5±2.1	n.s.
HCO3 – (mmol/l)	24.9±3.1	25.9±2.9	n.s.
Blood Count
Hemoglobinbaseline (mmol/l)	9.3±0.7	9.2±1.3	n.s.
Hemoglobinrecovery (mmol/l)	9.8±0.9	9.5±1.5	n.s.
Hematocritbaseline (%)	44±3	43±5	n.s.
Hematocritrecovery (%)	46±4	45±6	n.s.
White blood cellsbaseline (×109/l)	6.6±1.3	8.1±1.6	n.s.
White blood cellsrecovery (×109/l)	9.6±2.5	9.6±2.1	n.s.
Red blood cellsbaseline (×106/ml)	4.7±0.3	4.7±0.5	n.s.
Red blood cellsrecovery (×106/ml)	5.0±0.4	4.9±0.6	n.s.
Lymphocytesbaseline	30.7±6.0	26.0±7.0	n.s.
Lymphocytesrecovery	31.3±5.7	28.0±7.7	n.s.
Plateletsbaseline (109/l)	250.4±60.7	262.7±48.3	n.s.
Plateletsrecovery (109/l)	290.2±81.7	292.4±58.2	n.s.
Inflammatory Parameters
IL-1βbaseline (pg/ml)	0.08±0.05	0.12±0.08	n.s.
IL-1βrecovery (pg/ml)	0.13±0.09	0.15±0.08	n.s.
IL-6baseline (pg/ml)	1.41±1.44	1.89±2.77	n.s.
IL-6recovery (pg/ml)	2.69±3.74	4.85±11.71	n.s.
TNFαbaseline (pg/ml)	0.69±0.36	0.63±0.28	n.s.
TNFαrecovery (pg/ml)	0.95±0.41	1.0±0.58	n.s.
Parameters Measured in Spirometry
VO2baseline (l/min)	0.45±0.16	0.45±0.15	n.s.
VO2peak (l/min)	2.64±0.61	2.02±0.53	P < .001
VO2rel peak (ml/min/kg)	36.1±8.6	27.1±6.2	P < .001
VEbaseline (l/min)	11.8±4.3	11.9±3.7	n.s.
VEpeak (l/min)	84.6±22.8	59.4±14.8	P < .001
RERbaseline	0.81±0.06	0.91±0.07	P < .001
RERpeak	1.06±0.07	1.09±0.06	n.s.

Note: IL-1β, Interleukin-1β; IL-6, Interleukin-6, TNF-α, tumor necrosis factor-alpha; VO_2, oxygen consumption; VE, ventilation; RER, respiratory exchange ratio.

Results

All parameters were log-transformed before being included in analysis to achieve normal distribution after the Kolmogorov-Smirnov test had indicated non-normal distribution of some parameters.

The repeated measures MANOVA for physiological parameters (HR, breathing rate, RMSSD, SD1, SD2, RSA, and HC) indicated an effect for the factor GROUP [F(38,7) = 2.5, P < .03], a significant overall effect for the factor POWER OUTPUT [F(3,42) = 146.5, P < .001] and a significant effect for the POWER OUTPUT × GROUP interaction [F(3,42) = 31.2, P < .008], which is suggestive of differential changes of physiological parameters during the exercise task in both groups. ANOVAs for specific parameters revealed a significant effect for the factor POWER OUTPUT for HR (F = 52.2, P < .001), breathing rate (F = 4.3, P < .04), RMSSD (F = 123.5, P < .001), SD1 (F = 126.1, P < .001), SD2 (F = 101.9, P < .001), RSA (F = 101.7, P < .001), and for HC (F = 206.6, P < .001). In addition, we observed significant differences for the factor GROUP for HR (F = 5.6, P < .02), RMSSD (F = 11.0, P < .002), SD1 (F = 11.6, P < .001); SD2 (F = 5.9, P < .02), RSA (F = 14.2, P < .001), and HC (F = 17.2, P < .001). Significant POWER OUTPUT × GROUP interactions were observed for RMSSD (F = 6.2, P < .001), SD1 (F = 4.9, P < .001), SD2 (F = 8.5, P < .001), and RSA (F = 2.5, P < .02). Descriptive analyses for patients and controls by means of planned post hoc t tests are indicated in figure 1 and see online supplementary figure 1. Because the carbon monoxide concentration of exhaled air was different in patients and controls (being a measure of smoking behaviour [table 3]), we included this parameter as a covariate in the above described repeated measures MANOVA and calculated a repeated measures MANCOVA. Results showed significant differences for the factor POWER OUTPUT [F(2,42) = 44.8, P < .022] and POWER OUTPUT × GROUP interaction [F(2,42) = 50.8, P < .019], reducing the likelihood that results were profoundly influenced by smoking behavior. The ANCOVAs for specific parameters revealed a significant effect for the factor POWER OUTPUT for HR (F = 17.8, P < .001), RMSSD (F = 39.0, P < .001), SD1 (F = 38.5, P < .001), SD2 (F = 20.1, P < .001), RSA (F = 30.8, P < .001), and HC (F = 74.4, P < .001). No significant effect was found for the breathing rate (P = .8) in this ANCOVA. In addition, we observed significant differences for the factor GROUP for HR (F = 5.5, P < .02), RMSSD (F = 5.9, P < .02), SD1 (F = 6.3, P < .016), RSA (F = 7.9, P < .007), and HC (F = 11.2, P < .002). No significant effect was observed for the breathing rate (P = .6) and SD2 (P = .07). Significant POWER OUTPUT × GROUP interactions were observed for HR (F = 2.1, P < .05), RMSSD (F = 2.2, P < .04), SD2 (F = 5.9,P < .001), and RSA (F = 2.5, P < .02), while no differences were observed for breathing rate (P = .5), SD1 (P = .1), and HC (P = .6).

Fig. 1.

Changes of heart rate (A), breathing rate (B), respiratory sinus arrhythmia (C), and the complexity parameter compression entropy (D) are shown for controls (white boxes) and patients (blue boxes) during an incremental exercise task. At each workload-step, an independent two-tailed t test was calculated to compare patients and controls for descriptive analysis. While differences between heart rate were only shown at rest and during low intensity levels (A), reduced variability (C) and complexity (D) were evident at all intensity levels. Similarly, breathing rates were increased in patients (B) at all levels. Boxes indicate data between the 25th and 75th percentile with the horizontal bar reflecting the median (▯ = mean; – = 1st and 99th percentile; x = minimum and maximum of data). Significant differences of Bonferroni corrected pair-wise comparisons are indicated: *P < .05; **P < .01; ***P < .001.

The MANOVA for parameters of physical capacity (AerT, AnT, peak workload, and VT) revealed a significant effect for the factor GROUP [F(4,36) = 3.0, P < .029]. ANOVAs showed significant differences between patients and controls for AerT (F = 5.4, P < .026; figure 2), AnT (F = 6.8, P < .012; figure 2), peak output (F = 9.1, P < .004; figure 2), and VT (F = 6.2, P < .001; figure 2). Even after including the carbon monoxide concentration of the exhaled air as a covariate in the MANCOVA, we still observed significant differences for the factor GROUP [F(4,34) = 4.5, P < .005] as well as for single parameters (AerT [P < .011], AnT [P < .004], peak output [P < .001], and VT (P < .001]).

Fig. 2.

Significantly reduced physical fitness is shown at the aerobic and anaerobic thresholds for patients compared with controls. Similarly, patients show reduced peak physical capacity and decreased workloads at the vagal threshold. Boxes indicate data between the 25th and 75th percentile with the horizontal bar reflecting the median (▯ = mean; – = 1st and 99th percentile; x = minimum and maximum of data). Significant differences of Bonferroni corrected pair-wise comparisons are indicated: *P < .05; **P < .01; ***P < .001.

In addition, we compared VT, AerT, and AnT with respect to prescribed medication (olanzapine, quetiapine, and amisulpride) by means of a t test. The only observed difference was found for the VT of patients taking olanzapine (n = 6, 68.5±19) when compared with the threshold of patients taking quetiapine (n = 7, 51.5±9.3; P < .01).

Correlation Analysis

To analyze an assumed association between symptoms and VT, we calculated the Spearman’s rank correlation coefficient. An association was observed for PANSS total (r = −.452, P < .03), for PANSS positive (r = −.443, P < .034; figure 3B), and for the general psychopathology scale of PANSS (r = −.439, P < .036).

VT data of patients were correlated with concentrations of the inflammatory markers IL-1β, IL-6, and TNF-α after the exercise test. We observed a negative correlation of VT with IL-6 (r = −.453, P < .03) and with TNF-α (r = −.48, P < .02; figure 3C). No such correlation was observed in healthy subjects.

Peak power output (in W/kg) correlated with VT of patients (r = .403, P < .05), indicating a relationship between physical fitness and vagal modulation in patients. A trend was observed for the correlation of the duration of the disease with peak power output (r = −.398, P < .06) indicating that a longer disease duration might be associated with reduced physical fitness.

Discussion

The one-time physiological exercise task performed by patients with schizophrenia revealed three major findings. First, it supports findings on altered HRV and respiration previously described at rest in these patients in comparison to controls. Second, it demonstrates reduced physical fitness of patients in relation to various physiological thresholds. Third, it underlines the importance of vagal function in relation to clinical state, physical fitness, and inflammatory response.

Previous studies have described reduced HRV and increased breathing rates in patients at rest. ⁸ , ¹⁵ , ²⁸ Although a similar pattern of HRV alterations was described in relatives as well, some criticism was raised with respect either to smoking, agitation caused by the laboratory environment, or the influence of medication. While many authors have described altered HRV patterns in drug-naive or unmedicated patients, ²⁸ , ²⁹ the influence of smoking and investigation-related distress have not been addressed in detail. Here, we have shown that reduced vagal modulation dominates not only the pattern of cardiac regulation at rest but at all intensity levels up to 90W (figure 1C, see online supplementary figure 1A–C). Similarly, the complexity of cardiac regulation is not only decreased at rest, as previously shown, ³⁰ but during physical exercise as well. As previous studies have shown, increased breathing rates were also associated with shallow breathing in patients with schizophrenia. ¹⁴ , ¹⁵ Here, we present evidence that this prevailing pattern remains during physical exercise as well. Because autonomic dysregulation due to anxiety or investigation-induced distress normally settles during physical exercise, we suggest, in accordance with our hypothesis, that the autonomic imbalance described in patients with schizophrenia is a core feature of the disease that does not normalize in the context of a one-time physical exercise. Whether regular exercise has the potential to modify this autonomic imbalance in patients with schizophrenia remains to be investigated. The autonomic imbalance in patients during exercise is moreover reflected in the finding that higher intensity levels were associated with relatively increased breathing rates in patients compared with healthy controls, whereas HRs were not significantly increased during higher workloads (ie, >15W, figure 1A). This finding matches with the significant interaction effect for RSA that likewise indicates an abnormality in vagal modulation and cardiorespiratory coupling across the different exercise intensities in patients. These results are relevant for all exercise-related activities of patients, because they argue for decreased vagal modulation in favor of augmented sympathetic modulation. Sympathetic modulation is associated with an increased risk of arrhythmias due to complex effects on cardiac electrophysiological properties. Apart from direct effects on various repolarizing channels, indirect effects on receptor profile and cell metabolism have been described as pro-arrhythmogenic mechanisms of the sympathetic branch.

Most studies in patients with schizophrenia run the risk of being confounded by the influence of smoking severity. Although we tried to control for this confounding factor, it is rather difficult to find healthy young subjects who smoke the same amount of cigarettes per day as our patients do. We therefore measured the concentration of carbon monoxide (table 3) in the exhaled air of all subjects studied, as it can be used to estimate smoking habits. ¹⁶ As in the Fagerström test for nicotine dependence (table 1), the measured amount of carbon monoxide indicates a significantly stronger nicotine use of patients. We have therefore included the carbon monoxide concentration as a covariate in the analysis of our data. Because overall differences remained between patients and controls, we concluded that results were not caused by differences in smoking habits. However, covarying carbon monoxide in the analysis of physical fitness parameters seemed to improve statistical results between patients and controls. Future studies need to pay close attention to this result to elucidate the underlying cause.

The assessment of physical fitness was the second aim of our study. In accordance with our hypothesis, we found reduced physical fitness in patients compared with healthy controls, which was indicated by reduced workloads at the AerT, the individual AnT, and the VT. The AerT represents the workload associated with the first increase in blood lactate concentrations above resting values. Thus, in contrast to peak workloads, it is less affected by motivation. We conclude, therefore, that decreased workloads of patients at both the AerT and AnT argue for reduced physical fitness of patients. An interesting aspect of this study was that we applied a relatively new measure: the VT. This is defined as the point in time during exercise at which no subsequent decline in HRV occurs (figure 3A) and almost no vagal modulation remains. We found that this point is reached in patients at significantly lower workloads. This illustrates, on the one hand, the previously described reduced activity of the efferent vagal system in these patients, and, on the other, the close relation between the vagal system and physical fitness. ³¹ Moreover, this breaking point during physical activity was associated with general psychopathology corroborating previous results. Assuming that VT constitutes an indicator of vagal modulation in relation to physical fitness this finding supports previous evidence showing a close relationship between physical fitness and psychopathology in patients with schizophrenia. In addition, we observed a correlation between the inflammatory response after this exercise task with low vagal modulation, because VT correlated negatively with TNF-α and IL-6 in patients. There is increasing evidence for a correlation between physical fitness and the degree of exercise-associated inflammatory response in healthy people. ³² Our results suggest similar mechanisms in patients, potentially in connection with a reduced activity of the anti-inflammatory cholinergic pathway ¹⁸ , ³³ in patients as inflammatory parameters (apart from TNF-α) were increased in patients, albeit not significantly but on a descriptive basis (table 3). Longitudinal studies should investigate whether regular physical exercise in patients with schizophrenia influences these alterations in vagal modulation and, maybe as a direct consequence, the association between VT and inflammatory parameters. Future studies need to carefully describe disease-related data such as duration and previous types of medication, age of onset, or the number of episodes because a correlative trend was observed between duration of the disease and peak power output that might be indicative of such an association. Comparable findings were described in patients with major depression. ³⁴

Some limitations need to be addressed. The main limitation is that antipsychotic treatment of patients has influenced our results. However, many studies that have investigated drug-naive patients have shown similar HRV and respiratory findings. ⁸ , ⁹ , ²⁸ , ²⁹ , ³⁵ In addition, we have mainly abstained from comparing peak parameters of power output or oxygen consumption with physiological parameters. We aimed to exclude any motivational bias from being associated with peak parameters. Although patients smoked more cigarettes per day than controls, the inclusion of carbon monoxide as a measure of smoking behavior in our calculation reduced the potential influence of this confounder. Furthermore, differences of the VTs between patients taking olanzapine and quetiapine need to be taken with caution because this analysis included a very small number of patients. Similarly, we tried to match the amount of physical activity between patients and controls by means of the IPAQ. We cannot, however, exclude the possibility that patients overestimated their physical activity and shortcomings due to the assessment interval measuring the last 7 days only. In addition, the methods applied in this study were unable to disentangle the close relationship of decreased physical capacity and autonomic alterations in more detail.

Maintaining or improving fitness is associated with a lower risk of premature deaths from all-causes and cardiovascular disease in men. ³ Preventing fitness loss with increasing age, regardless of whether BMI changes, is important for mortality risk reduction. Our study shows reduced physical capacity, even after adjusting for weight, and decreased vagal modulation in patients with schizophrenia to be associated with psychopathology and a pro-inflammatory response after exercise. This underlines the need for larger randomized, maybe even longitudinal, studies in patient with schizophrenia investigating the effects of repeated, regular physical exercise on physical fitness, psychopathological symptoms as well as autonomic and immunological alterations.

Supplementary Material

Supplementary figure 1 is available at http://schizophreni abulletin.oxfordjournals.org.