The short and long-term effects of aerobic, strength, or mixed exercise programs on schizophrenia symptomatology

The purpose of this study was to compare the effects of three different physical exercise programs on the symptomatology, body composition, physical activity, physical fitness, and quality of life of individuals with schizophrenia. A total of 432 patients were assessed for eligibility and 86 were randomized into the aerobic (n = 28), strength (n = 29) or mixed (n = 29) groups. Positive, negative, and general symptoms of psychosis, body mass index (BMI), physical activity (IPAQ-SF), physical fitness (6-min walk test [6MWT] and hand-grip strength [HGS]), and quality of life (WHOQUOL-BREF) were assessed at baseline, post-intervention (16 weeks), and at 10-months. Our results at 16 weeks showed significant improvements in all three groups in the negative, general, and total symptoms with moderate to large effect sizes (P < 0.01, ηp2 > 0.11), no change in the BMI, 6MWT or IPAQ-SF, and a significant improvement in the HGS test in the strength and mixed groups (P ≤ 0.05, ηp2 > 0.08). Nonetheless, all the improvements had disappeared at 10 months. We concluded that 3 weekly sessions of a moderate to vigorous progressive exercise program for 16 weeks improved the symptomatology of individuals with schizophrenia in all three groups, with no differences between them. However, the effects had declined to baseline levels by the 10-month follow-up, suggesting that exercise interventions should be maintained over time.

Procedure. This study took place at six psychosocial care centers for people with severe mental illness (SMI), from January 2020 to January 2021. Before the start of the trial, an independent researcher unaware of the study characteristics generated a random sequence using a computerized random number generator; this was concealed from all the other study investigators throughout the entire study period. Randomization was performed with stratification for age, sex, and body mass index (BMI). Upon enrolment in the study and after completing the primary and secondary outcome measures, the participants were randomly assigned either to the aerobic (n = 28), strength (n = 29), or the mixed (n = 29) group.
As shown in the participant flowchart in Fig. 1, all the outcome measures were assessed at baseline, 16 weeks post-baseline (post-intervention), and at a 10-month follow-up. It was impossible to mask the group allocation to the physical therapists or the participants; however, the outcome evaluators and data analysts were blinded to the treatment allocations. To avoid inter-observer variability bias, the measurements in each of the groups were always completed by the same investigator.
Interventions. Informed consent was obtained from the subject represented in the pictures. The intervention consisted of a total of 48 sessions (3 weekly sessions lasting 1 h each for 16 weeks). To make the comparison fair, the total number of training sessions and their duration were the same for all three training groups. The exercise training was conducted in groups at each psychosocial care center by a professional physical education instructor. Each session began and ended with 10 min of stretching of the major muscle groups.
Aerobic training. Participants completed 4 series of brisk walking (in the playground of the psychosocial care centers) for 10 min followed by 1 min of recovery. To ensure that the intensity of the exercise progressed from moderate to vigorous, we monitored the heart rate (HR) of each participant using Sport testers (model 610si, Polar, Finland). Progression in exercise intensity was achieved by increasing the participant's target HR every 2 weeks. Thus, using the formula published by Tanaka et al. to calculate the maximum HR (MHR) (208-0.7 × age) 28 , the intensity of the exercise was progressively increased as follows: weeks 1-2: 55% MHR; weeks 3-4: 58% MHR; weeks 5-6: 61% MHR; weeks 7-8: 64% MHR; weeks 9-10: 67% MHR; weeks 11-12: 70% MHR; weeks 13-14: 73% MHR; and weeks 15-16: 76% MHR (Fig. 2). The physical education instructor continuously checked the HR to adjust the walking speed of each participant in each session.
Strength training. Participants completed two sets of 8 strength training exercises with 1 min of recovery programmed between each one ( Fig. 3; informed consent was obtained from the subject represented in the picture). An elastic resistance band (Thera-band) was used in four of the eight strength exercises. The training intensity increased over the 16 weeks of this intervention; the intensity of exercises completed without an elastic band www.nature.com/scientificreports/ was amplified by increasing the number of repetitions the participants perform. For exercises performed with an elastic band, the intensity increase was achieved by using the Borg Scale. This scale measures the effort an individual perceives when exercising and creates criteria to adjust the intensity of the programmed exercise. In order to adequately use the Borg scale, the participants learnt to use Theraband resistance bands on the first day and to easily identify, for each exercise, which gripping point on the band was equivalent to an effort that was moderate, intermediate, hard or very hard according to the Borg scale. Figure 3 shows the progression (from 3 to 8) in the rate of perceived exertion (RPE) during the 16 weeks of the program.

Follow-Up (10 months)
Lost to post-intervention (n = 2) (Reasons: 2 withdrew from study) Lost to post-intervention (n = 5) (Reasons: 5 withdrew from study) Lost to follow-up (n = 7) (Reasons: 5 withdrew from study, 2 were unreachable to contact) Lost to follow-up (n = 6) (Reasons: 4 withdrew from study, 2 were unreachable to contact) Lost to follow-up (n = 4) (Reasons: 3 withdrew from study, 1 was unreachable to contact) Analyzed (n = 28) Excluded from the analysis (n = 0) Analyzed (n = 29) Excluded from the analysis (n = 0) Analyzed (n = 29) Excluded from the analysis (n = 0)  www.nature.com/scientificreports/ Mixed training. The main part of each mixed session was divided into two stages. First, like the strength training group, the participants performed 1 set of the same 8 strength exercises with the same intensity and progression, interspersed with 1 min of recovery time per strength exercise. Second, as in the aerobic training group, the participants performed 2 sets of brisk walking for 10 min followed by 1 min of recovery, according to the same exercise intensities and progressions described for the aerobic training group. All participants were recommended to keep a physically active lifestyle during the 10-month follow-up period.
Primary outcome. The PANSS is a semi-structured interview which assesses the positive (PANSS-P), negative (PANSS-N), and general (PANSS-G) symptoms of psychosis experienced by patients in the week prior on a 7-point Likert-type scale (from 1, 'none' , to 7, 'extreme') 31 . We separately analyzed the 3 subscales as well as the overall scores (PANSS-T).
Secondary outcomes. The BMI was calculated using a SECA® 780 electronic balance scale with a mechanical telescopic stadiometer. Body fat mass (BFM) was determined using a TBF-300A body-fat analyzer. A submaximal exercise test, the 6-min walk test (6MWT), was used to assess the participants' functional capacity for aerobic exercise. This test has been shown to be a reliable measure of exercise capacity in people with SMI 32 . Hand grip strength (HGS) was assessed using a Jamar hydraulic hand dynamometer, as described elsewhere 33 . The 30-s sit-to-stand (30-s STS) test was used to assess lower limb strength. It consisted of standing-up from a chair and sitting-down again as many times as possible during a 60 s period, with the researcher registering the number of repetitions 34 . PA level was be assessed using the International Physical Activity Questionnaire-Short Form (IPAQ-SF) 35 . Using seven items, this self-reported questionnaire collects data on the patients' PA in the 7 days prior to the test. The total number of days and minutes of PA were calculated by adding all PA category scores performed over the seven days. Data from the IPAQ-SF were converted into metabolic equivalent minutes per week (METs-min/week), using the formula published by Ainsworth et al. 36 . Specifically, the IPAQ-SF questionnaire records activity at four intensity levels: (1) vigorous activity such as aerobics; (2) moderate activity such as leisure cycling; (3) walking; and (4) sitting. This makes it possible to classify the PA levels of the participants as 'high' (> 1500 METs), 'moderate' (600-1500 METs), or 'low' (< 600 METs). The IPAQ has been validated in 12 countries 37 and showed adequate psychometric properties and the short version (the IPAQ-SF) has shown acceptable validity in an adult Spanish population 38 . Finally, quality of Life was assessed using the abbreviated WHO Quality of Life Assessment-BREF 39 : This survey comprises 26 items with five Likert-type responses each, and is a standard questionnaire used to measure patient quality of life. It assesses patients under four health domains: physical, psychological, social and environmental. In this study, we will analyze the sum of the four dimensions, with higher scores indicating a better quality of life. This scale has been validated for Spanish and the instrument has a good internal consistency with a Cronbach alpha of 0.88 for the overall scale and a range of 0.70-0.79 for its dimensions 40 .
Sample size and statistical analysis. The desired sample size was calculated after undertaking a pilot study in 15 participants 41 , which indicated a partial eta 2 effect size (η p 2 ) of 0.038 for the primary outcome (PANSS-N). Considering this and using an α value of 0.05 and a desired power of 90%, we used the G*Power 10 minutes of brisk walking. The exercise intensity should be moderate. We will control it by monitoring the heart rate (HR). Any paƟents who cannot use their HR as a reference because of their medical treatments should perform the exercise at an intensity which makes it difficult but not impossible for them to speak while performing it.

HR MEASUREMENT
See the proposed HR table created based on the paƟent age and the training program exercise week.

AEROBIC TRAINING
10 minutes of brisk walking. The exercise intensity should be moderate. We will control it by monitoring the heart rate (HR). Any paƟents who cannot use their HR as a reference because of their medical treatments should perform the exercise at an intensity which makes it difficult but not impossible for them to speak while performing it.

HR MEASUREMENT
See the proposed HR table created based on the paƟent age and the training program exercise week.

AEROBIC TRAINING
10 minutes of brisk walking. The exercise intensity should be moderate. We will control it by monitoring the heart rate (HR). Any paƟents who cannot use their HR as a reference because of their medical treatments should perform the exercise at an intensity which makes it difficult but not impossible for them to speak while performing it.

HR MEASUREMENT
See the proposed HR table created based on the paƟent age and the training program exercise week.

AEROBIC TRAINING
10 minutes of brisk walking. The exercise intensity should be moderate. We will control it by monitoring the heart rate (HR). Any paƟents who cannot use their HR as a reference because of their medical treatments should perform the exercise at an intensity which makes it difficult but not impossible for them to speak while performing it.

COOL-DOWN
At the end of the second set: 10 mins of stretching the major muscle groups  Possible confounding factors were assessed by testing baseline differences between the groups using chisquared (sex, educational level, marital status, employment, and institutionalization regime), Kruskal Wallis (age), and one-way ANOVA (illness duration, BMI, antipsychotic drugs consumption, and ACB scores) tests. ANOVA tests were also used to compare the percentage of attendance rates between the intervention groups.
Two-way mixed ANOVA tests were used to compare how the study interventions affected the primary and secondary outcomes, using time (baseline, post-intervention, and 10-month follow-up) as the within-group factor and group (aerobic, strength, or mixed) as the between-group factor. Analogous ANOVA tests but with two  www.nature.com/scientificreports/ levels in the time factor (baseline and 10-month follow-up) were used to compare how the study interventions affected the consumption of antipsychotic drugs. Bonferroni post-hoc tests were applied following the ANOVAs. The η p 2 effect sizes were calculated such that 0.01 < 0.06, 0.06 < 0.14, and 0.14 or higher, respectively corresponded to a small, medium, and large effect size 43 . The statistical analyses were performed as an intention-to-treat basis using SPSS software (v.18.0) for Windows (IBM Corp., Armonk, NY). All the statistical tests were two-tailed with the critical P value for significance set at < 0.05.

Results
The general characteristics of the study population are shown in Table 1. The statistical analysis did not show between-group differences at baseline for any variable, except for the number of years of illness, for which there were differences between the aerobic and mixed intervention groups. In addition, the percentages of attendance rates did not show statistically significant differences between the aerobic, strength, and mixed intervention groups (88 ± 15%, 88 ± 12%, and 79 ± 15%, respectively; P = 0.084). Likewise, the results of two-way mixed ANOVA did not show intra-group or between-group differences in the comparisons of the consumption of antipsychotic drugs. None of the participants reported any adverse events to the interventions.
The results of the Bonferroni post-hoc tests are shown in Table 2. The PANSS-N, PANSS-G, and PANSS-C scores showed statistically significant improvements from baseline to 16 weeks in the 3 groups (P < 0.01, η p 2 > 0.11); however, all these significant improvements disappeared at the 10-month follow-up. The PANSS-P scores also improved in the aerobic and strength groups (P ≤ 0.005, η p 2 > 0.07), and showed a trend towards improvement in the mixed group (η p 2 > 0.04), but also got worse at the 10-month follow-up. No significant differences were shown in the intra-group comparisons of body composition (BMI and BFM), 30-s STS test, 6MWT, or IPAQ-SF. Of note, there were significant improvements in the HGS test results from baseline to 16 weeks both in the strength and mixed groups (η p 2 > 0.08). However, these improvements also disappeared at the 10-month follow-up. Similarly, the quality of life showed improvements (significantly in the aerobic group) from baseline to 16 weeks that disappeared at the 10-month follow-up. Interestingly, the intra-group comparisons between 16 weeks and the 10-month follow-up did not show statistically significant differences in symptomatology, body composition, 6MWT, or IPAQ-SF in any group. In contrast, the HGS and 30-s STS scores significantly decreased in the strength group, the HGS significantly decreased in the mixed group, and the quality of life significantly    www.nature.com/scientificreports/ decreased in the aerobic group. Finally, no between-group differences were found in any of the primary or secondary outcome comparisons (P > 0.05).

Discussion
Our results showed significant improvements in the negative, general, and overall score for the symptoms of psychosis at 16 weeks in the aerobic, strength, and mixed training groups, with moderate to large effect sizes. The mean difference in PANSS-T scores nearly reached the level of the minimal clinically important difference, estimated as a 16% to 24% change from baseline 27 . These findings are consistent with several meta-analyses that showed that PA (i.e., aerobic or mixed exercise) significantly reduced the negative 16,18,26 and total 16,18 score symptoms for schizophrenia. To date, only two studies have examined the impact of an isolated strength training program on patients with schizophrenia 23,24 . The pilot study by Heggelund et al. 23 , which consisted of an 8-week intervention (training 3 days per week) with a single exercise (4 × 4 repetitions at 85-90% of the 1 repetition maximum, performed with a leg press apparatus), did not report improvements in the negative symptoms. However, this study only included one exercise type completed for short durations, had a small sample size, and had methodological limitations (it was not randomized, assessment was not blinded, the statistical analysis did not study time × group interactions, and the data were not analyzed on an intention-to treat basis).
The most recent study by Silva et al. 24 compared the effects of 20 weeks of strength and mixed training programs (2 days per week) on psychotic symptoms. Although the authors did not find between-group differences in the PANSS-N, the intragroup analysis revealed significant improvements in the strength training group. Nevertheless, this study was also limited by some methodological issues (the sample size was small and not calculated a priori, no allocation concealment was reported, and there was no intention-to treat analysis). Therefore, our results fill a gap in literature because we consistently demonstrated that a strength training program alone was also an effective treatment for the negative and general symptoms of schizophrenia. Indeed, this type of exercise has already obtained good results in diseases such as anxiety and depression 25 and has been shown to increase cognitive efficiency in older adults 44 and patients with early dementia 45 .
Our findings showed that there were significant improvements in the positive symptoms at 16 weeks in the aerobic and strength intervention groups (with small to moderate effect sizes) and showed a positive trend in the mixed intervention group. These findings agree with other studies that analyzed the effects of aerobic interventions on positive symptoms. For example, Firth et al. 16 conducted a meta-analysis of exercise interventions in individuals with schizophrenia and, after performing a sensitivity analysis to exclude low-intensity exercise interventions, reported significant reductions in the PANSS-P scores, with a pooled standardized mean difference (SMD) of − 0.54 (95% CI − 0.95 to − 0.13). The most recent meta-analysis 26 , which included 17 RCTs, conducted a subgroup analysis that differentiated between aerobic interventions (12 RCTs) and non-aerobic interventions (5 RCTs). In this study, aerobic exercise reduced positive symptoms (SMD = − 0.27; [95% CI − 0.46 to − 0.09]; P = 0.004), while non-aerobic interventions did not reduce these symptoms (SMD = − 0.03; [95% CI − 0.29 to 0.23]; P = 0.82). Nonetheless, it must be noted that, rather than training strength, most non-aerobic intervention studies used other types of low-intensity non-aerobic exercises such as yoga 46 .
Contrary to our expectations, in this current work the intragroup analysis failed to show significant improvements at 16 weeks in the mixed training intervention group, and only showed a slight (non-significant) trend. Unlike other mixed intervention studies 21 that performed a per-protocol analysis and found significant improvements in the positive symptoms, we performed an intention-to-treat data analysis. Although the challenges of this type of analysis include possible loss to follow-up and varying compliance levels, this type of analysis maintains the original group composition achieved by randomization (avoiding potential bias due to exclusion of patients) and pragmatically estimates the benefit of the intervention by estimating how it might perform in a real clinical setting 47 .
In another vein, the fact that our results did not reach statistical significance in the mixed group could perhaps be explained both by the higher attrition rates and the lower compliance of the patients of this group. Hence, even though our 3 exercise interventions were delivered considering 2 major factors known to increase adherence and compliance in schizophrenic patients, i.e., supervision 10 and a group training setting 16 , the percentages of dropouts were higher and the attendance rates were lower in the mixed intervention group (aerobic = 17.9% and 88%; strength = 13.8% and 88%; and mixed = 34.5% and 79%, respectively).
While the mechanisms by which the different exercise interventions may have influenced the positive and negative symptoms of our patients extends beyond the scope of this study, several mechanisms have been proposed in the scientific literature. The most frequently cited are neuroprotective mechanisms such as decreased inflammation, increased neurogenesis and neuroplasticity via brain-derived neurotrophic factor, and remyelination of white matter tracts 48 .
A recent large-scale meta-analysis confirmed that SMI patients have a significantly increased risk of CVD and CVD-related mortality and that elevated BMI in these patients, often related to antipsychotic drug use, among other factors, requires urgent clinical attention 6 . In particular, weight gain induced by second-generation antipsychotic drugs is a well-documented and prevalent side effect of antipsychotic treatment 49 . Thus, physical exercise is particularly recommended in this population and is included in clinical guidelines for these patients. In agreement with the literature, most of the patients in our study had baseline BMIs corresponding to excess weight (38%) or obesity (46%). In addition, all our patients were on stable antipsychotic medication, with doses that can be considered 'high' 50 corresponding to an equivalent of approximately 850 mg/day of chlorpromazine.
Consistent with most previous work 16,20,21,24 , none of our three groups showed significant improvements in body composition at 16 weeks. This agrees with the finding of the meta-analysis by Firth et al. 16 , which included studies with different types of exercise programs, that found no significant post-intervention changes in BMI www.nature.com/scientificreports/ in these patients. While achieving reductions in BMI and BFM are important exercise goals that should not be abandoned, a more realistic goal may be the attenuation of expected weight gain, to which schizophrenic patients are particularly susceptible 16 . Thus, perhaps our finding of no BMI or BFM gain should be considered noteworthy in itself. Moreover, active but obese individuals have a reduced risk for all-cause and CVD mortality compared to individuals who have a normal BMI but are physically inactive 51 . Therefore, as suggested by Vancampfort et al. 52 , exercise interventions in schizophrenic patients should be primarily developed to improve physical (cardiorespiratory) fitness, with BMI and weight reduction considered as secondary outcomes. We measured the change in functional exercise capacity using the 6MWT. This test corresponds more to the demands of everyday activities than the maximal oxygen uptake test and is less likely to evoke nervousness or anxiety 53 . Like our results, the studies by Marzolini et al. 20 and Beebe et al. 54 which implemented 12-week mixed and 16-week aerobic interventions, respectively, did not find significant improvements in the 6MWT. In contrast, Korman et al. 19 did report a significant improvement in 6MWT results in their single-arm study (n = 10) with a 10-week mixed intervention. Of these four studies, our study started from the highest basal levels [> 560 m vs. 535 20 , 452 21 , and 430 54 ] and therefore had the lowest room for improvement. On the other hand, the systematic review by Bernat et al. 55 analyzed sixteen intervention studies and found, like our results, that none of them reported a significant increase in the distance walked by adults with schizophrenia.
We measured the levels of physical activity using the IPAQ-SF. In line with the results of the 6MWT, we did not find significant changes in the levels of physical activity in any group-we only found a slight tendency towards improvement-. Although none of the exercise intervention studies discussed here [19][20][21][22][23][24]54 assessed physical activity levels (IPAQ-SF) and it is not possible to make direct comparisons, a recent study suggested that the minutes of sitting reported by the IPAQ do not reflect objective sedentary behaviour measurements 56 . These authors concluded that the IPAQ may be unsuitable for the population level assessment of sitting time among individuals with schizophrenia.
As expected, there were only significant intra-group improvements in strength in the strength and mixed training groups. Thus, HGS significantly increased (≥ 3 kg) in both groups. In addition, the 30-s STS test narrowly missed achieving significance in the strength group. Our results agree with other work that also found improvements in strength measured with one-repetition maximum tests after a period of isolated 23,24 or mixed 20,22 strength training-i.e., the most recent well-designed randomized controlled trial by Nygård et al. 22 concluded that twelve weeks of maximal strength training restored patients' lower extremity force-generating capacity to a level similar to healthy references and improved 30-s STS test performance-. A cross-sectional study concluded that patients with schizophrenia showed lower HGS scores compared to healthy controls, and that HGS scores correlated positively with cognitive functions 57 . A more recent study concluded that higher HGS was associated with greater left and right hippocampal volume and reduced white matter hyperintensities in major depressive disorder (MDD). These authors considered that interventions targeting strength fitness could improve brain health and reduce the neurocognitive abnormalities associated with MDD 58 . Moreover, HGS has been suggested as a risk indicator for cancer mortality 59 and fatal cardiovascular and all-cause mortality events 60 . Furthermore, isolated strength training has been associated with a lower risk of all-cause mortality, regardless of participation in aerobic PA 61 .
To the best of our knowledge, this current study is the first RCT to compare the long-term effects of 3 different types of exercise programs on the symptomatology, body composition, physical fitness, level of physical activity, and quality of life of patients suffering from schizophrenia. Although none of the 3 programs maintained their significant effects for any variable at the 10-month follow-up, it is remarkable that, without any changes in antipsychotic medication, the variables did not worsen with respect to baseline levels.
This study has limitations that must be considered. First, the patients we enrolled were (a) stable on antipsychotic medication; (b) had PANSS-T scores ≥ 58; and (c) had demonstrated an initial level of motivation to engage in the exercise programs. Therefore, our findings may only be generalizable to individuals with similar characteristics. Second, even though we based our sample size calculations on potential losses of 25%, there were 17 losses to the 10-month follow-up, which meant that our statistical analysis was underpowered by the last time point. Therefore, the long-term effects of these interventions should be interpreted with caution. Finally, the lack of a control group should be considered when interpreting the results. Controls eliminate the alternate explanations of experimental results, especially confounding variables and experimenter bias, enabling the investigator to control for threats to validity.
Despite the availability of comprehensive and evidence-based treatment guidelines, not all patients benefit from standard care (e.g., medication) 62 . While anti-psychotic medication is effective for the positive symptoms of schizophrenia 63 , it is less effective for the treatment of its negative symptoms and cognitive deficits 64 . Unfortunately, these deficits constitute major determinants of poor functioning and disability 3 and negative symptoms are an important predictor of an unfavorable disease course 26 . Thus, adjunctive therapies like physical exercise that can reduce negative symptoms, cognitive deficits, and improve functional outcomes are vital.
Perspective. This study concluded that 3 weekly sessions of a moderate to vigorous progressive exercise program for 16 weeks improved the symptomatology of schizophrenic patients in all three groups (aerobic, strength, or mixed), with no differences between them. Additionally, the greatest improvements in strength were achieved with interventions that included strength training exercises. However, their positive effects declined to baseline levels at the 10-month follow-up, suggesting that PA programs should be maintained for longer. Future studies should be deigned to elucidate strategies to keep patients with schizophrenia physically active over time.

Data availability
Data available on request from the authors.