Psychosis risk individuals show poor fitness and discrepancies with objective and subjective measures

Exercise is a promising intervention for individuals at clinical high-risk for psychosis (CHR). However, these youth may not be reliable reporters on fitness. There have been no investigations that utilized objective fitness assessment in this population. The present study objectively characterizes the level of fitness in CHR youth, compares the accuracy of self-report measures to objective fitness indices, and explores clinical factors that may influence the accuracy of self-reported measures of fitness. Forty CHR individuals completed an exercise survey and objective indices of fitness (i.e., VO2max and BMI). Forty healthy volunteers completed objective indices of fitness and a structured clinical interview ruling out the presence of psychiatric illness. CHR youth showed greater BMI and lowered VO2max compared to healthy volunteers. In the CHR group, self-report items (perceived fitness) did not reflect objective indices of fitness, whereas specific exercise behaviors (intensity of exercise) showed stronger correlations with objective fitness measurements. Exploratory analyses suggested that symptoms (grandiosity and avolition) related to errors in self-perception. Results indicate that CHR individuals are less fit than controls as indexed by objective measures of fitness and that it is important to consider unique population clinical characteristics when employing self-report data.

www.nature.com/scientificreports/ Clinical assessment of symptoms. All subjects completed the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID) to rule out any psychosis diagnosis for both the CHR and healthy volunteer groups. CHR participants completed the Structured Interview for Psychosis Risk Syndromes (SIPS) to assess the presence of a CHR syndrome and track attenuated symptoms. Exploratory analyses included positive and negative domain totals and specific symptoms. The SIPS is made up of 19 items that are rated on a scale from 0 to 6 by an interviewer. These items are grouped into larger symptom dimensions of psychosis (e.g., positive and negative). The SIPS scale contains an instrument referred to as the Scale of Prodromal Symptoms (SOPS). On the SOPS an interviewer rates the severity of symptoms along a 7-point scale ranging from absent (0) to severe and psychotic (> 6). Advanced doctoral students served as the study interviewers were trained over a 2-month period after their inter-rater reliabilities exceeded the minimum criterion of Kappa ≥ 0.80. In the current study, individuals were grouped as clinical high risk if they met the criteria for attenuated positive symptoms syndrome (i.e., recently emerging symptoms that occur at a weekly frequency or long-standing symptoms that have recently escalated).
Self-reported fitness scale. CHR participants completed a self-report survey comprised of items from many validated measures. Subscales of this self-report have been previously reported in assessing physical activity in CHR individuals 11,13 . The items selected for the current study include Perceived Fitness, Frequency of Exercise, Time Spent Exercising, and Intensity of Exercise. Perceived fitness was rated by participants on a Likert-type scale that ranged from 0 (Poor) to 3 (Excellent). The frequency of exercise, where exercise included any activity that resulted in sweating or rapid heart rate, was rated based on a typical week and ranged from 0 (rarely or never) to 3 (five or more times). Time spent exercising was rated on a scale spanning 0 (< 30 min) to 3 (> 60 min). The intensity of exercise assessed the frequency that these exercise sessions resulted in sweating or rapid heart rate on a scale from 0 (never) to 3 (always; every time).

Objective indices of aerobic and biometric fitness. Objective indices of fitness included VO 2 max and
Body Mass Index (BMI). VO 2 max indexes an individual's ability to transport and use oxygen at the maximum capacity during aerobic exercise. At both sites, body mass index (BMI) was calculated based on height and weight measurements collected in the lab using the U.S. Department of Health and Human Services‫׳‬ National Heart Lung and Blood Institute (http:// nhlbi suppo rt. com/ bmi/ bmino js. html) BMI online calculator. The error of fitness estimation was calculated by relating objective fitness metrics to self-reported perceived fitness. The standard model error was computed for each subject, reflecting the difference between the actual values of selfreported perceived fitness and the expected values of fitness according to objective fitness metrics. Individuals were then categorized into inaccurate and accurate reporter groups consistent with their quartile distribution of standardized errors. Those individuals who were inaccurate in their estimates (either highly over-or underestimating their fitness) were grouped as inaccurate reporters; those with more typical amounts of error in their estimates (near the 50 th percentile of standard model error) were grouped as accurate reporters for follow-up analyses. At Northwestern, an expert exercise physiologist conducted a modified Balke max-exercise protocol 41 under the supervision of a physician. In this modified Balke max-exercise protocol, the treadmill speed was set to elicit 70% of the age-predicted max heart rate and an RPE rating of around 13 ("somewhat hard"). This target was achieved by maintaining the speed of the treadmill remained the same throughout the test, but adjusting the incline of the treadmill belt increased 2% every 2 min (or 2.5% for speeds 6 mph or greater). The participants' heart rate and ratings of perceived exertion (RPE) determined the final treadmill speed. Staying within these parameters generally yields an 8-12 min test. During the protocol, the speed of the treadmill remained the same, but the incline of the treadmill belt increased 2% every 2 min (or 2.5% for speeds 6 mph or greater). Tests generally lasted 8-12 min, the recommended target for VO 2 max testing 42 .
The Human Performance Laboratory at Columbia University Medical Center assessed the VO 2 max of participants during cycling. The tests were performed on an electronically-braked cycle ergometer that was calibrated prior to every test (Ergoline 800S electronic-braked cycle ergometer (SensorMedics Corp., Anaheim, CA). Participants completed a four-stage protocol: 1) 5-min resting baseline, 2) 3-min no-resistance warm-up, 3) increasing speed and resistance until 10-15 W with maximum exercise norms (total duration of approximately 12-min), 4) active recovery of 3-min. During the test, the workload was increased 10-15 W every 1-min until one of four exercise criteria was reached. To generate the VO 2 max (mL/kg/min) variable, participants achieved one of the following four criteria during exercise: VO 2 plateau, 85% of maximal heart rate (220-age), respiratory quotient ≥ 1.1, or self-reported exhaustion as indexed by the Borg Scale 43 to generate the VO 2 max (mL/kg/min) (MAX-1, PHYSIO-DYNE Instrument Corp., Quogue, NY) . For more detail on the full exercise procedure, refer to Kimhy et al., 2014. Exercise methods, gender, and age-appropriate z-scores were created for all VO 2 max scores separately for each exercise type (treadmill and cycling) according to the American Heart Association 44 national norms to ensure comparability across sites. By using national normalization guidelines, we aim to reduce the impact of local site features-local fitness opportunities and habits-as well as account for the difference in exercise intervention type (i.e., treadmill and cycling), which are expected to elicit distinct levels of VO 2 max. This approach is consistent with recommendations put forth in extant population normalization aerobic fitness measures 45 . Analytical approach. Participant demographic comparisons across groups were conducted on sample features using chi-square analyses to characterize categorical sample features and t-tests to examine the continuous features of the sample. Group analyses were designed to contextualize the objective fitness to self-reported fitness perception and are treated as exploratory. For multiple comparisons, the group comparisons (CHR vs. healthy www.nature.com/scientificreports/ volunteers) will be treated as independent analyses from the within CHR group analyses. As a result, the withingroup (CHR only) analyses will be corrected for the two comparisons (objective fitness to self-report fitness; objective fitness to symptoms), and findings were treated as significant if the p < 0.025, a Bonferroni correction. Any follow-up analyses with particular symptoms will be interpreted cautiously and treated as exploratory. For all significant analyses, follow-up analyses will be conducted to examine if sex or age contributes significantly to the models. Variables that significantly contribute to the model or impact the direction or magnitude of the reported findings will be included as a nuisance variable, in line with Miller and Chapman (2001) 46 . Separate simultaneous general linear models compared groups (i.e., CHR and healthy volunteers) on objective metrics of health, where group membership predicted fitness (VO 2 max or BMI, respectively). Analyses within the CHR group examined whether self-report indices of fitness related to objective indices; to limit the total number of analyses, self-report subscales were entered simultaneously into separate general linear models to predict VO 2 max and BMI in separate analyses. This approach has the added benefit of accounting for multiple self-reported features of exercise. Any significant subscale provided added insight into the relevance of particular items over and above other items.
Similarly, a repeated-measure general linear model examined the relationship between clinical symptoms (positive and negative) as the within-subject measure. The subscales of self-reported indices of fitness were entered simultaneously as the between-subjects measure in a single model to reduce the number of total comparisons. In this approach, a significant relationship to symptoms would indicate that the self-reported indices of fitness related to symptom severity, and interaction by scale would indicate that the relationship varies by symptom type (positive or negative). Finally, in a set of exploratory analyses, CHR individuals were grouped by error quintiles as accurate or inaccurate (based on the relationship of their perceived fitness to each objective fitness) and were compared to symptoms of grandiosity and avolition in separate t-tests: grandiosity by VO 2 max accuracy quintile, grandiosity by BMI accuracy quintile, avolition by VO 2 max accuracy quintile, avolition by BMI accuracy quintile.

CHR errors in perceived fitness related to symptoms.
In separate t-tests, BMI related to errors in perception (how far the actual fitness deviated from the predicted fitness) for both grandiosity, t(38) = 2.28, p = 0.04, Fig. 3a, and avolition, t(38) = 2.55, p = 0.02 Fig. 3b. VO 2 max related to errors in perception (how far the actual fitness deviated from the predicted fitness) for both grandiosity t(38) = 2.29, p = 0.04, Fig. 3c, and avolition, t(38) = 2.55, p = 0.02, Fig. 3d. Follow-up analyses were conducted examining positive and negative symptom totals to examine specificity, which did not relate to errors in perception (how far the actual fitness deviated from the predicted fitness) p's > 0.12. Follow-up analyses examined the potential contribution of age and sex to the models above and found that they did not contribute to the model significantly, p's > 0.51. www.nature.com/scientificreports/

Discussion
The current paper was the first to take a comprehensive assessment of fitness in CHR individuals. CHR individuals were significantly less fit than peers in terms of objective physiological (VO 2 max; Fletcher et al., 2001) and biometric (BMI) measures of fitness. Self-reported perceptions of fitness 11,13,20 did not reflect either objective metric of fitness (VO 2 max, BMI), but objective items regarding fitness behaviors (i.e., intensity and time spent exercising) related to the objective metrics of fitness. Finally, errors in perceived fitness 38 were related to a distorted perception of self (i.e., grandiosity) and motivation (i.e., avolition). CHR individuals showed lower levels of fitness on objective physiological (VO 2 max; 1,41,44 and biometric (BMI) measures in separate analyses, compared to healthy volunteers. This lower objective fitness emphasized the potential benefit of exercise as an early, non-invasive intervention in CHR individuals 6,7,9 . This finding extended the psychosis literature, suggesting that a psychosis spectrum diagnosis is associated with lower fitness 1,6,15,21,22 including individuals with attenuated psychotic symptoms. Taken together, the emergence of attenuated symptoms co-occurred with a decline in physical fitness may have reflected an overall deterioration in neurological fitness. This deterioration of symptoms might have led to a decline in both engagement in fitness activities and health 1,2,6 . These findings emphasized the importance of objective metrics of fitness 38 . Future studies should evaluate fitness longitudinally as these markers of health may confer additional risk for conversion and may be a therapeutic target for individuals at greatest risk.
Self-report perceived level of fitness 13 did not reflect objective metrics of fitness overall 1,40,41,44 . Self-reported intensity of exercise 13 did reflect physiological fitness as expected; increased self-reported exercise intensity related to an increased VO 2 max capacity 1,41 . In contrast, self-reported time spent exercising and intensity of exercise 13 corresponded to an increased BMI. This inconsistency with BMI highlights may be a limitation of BMI; BMI does not distinguish between the content of body mass in terms of whether the weight reflects increased muscle mass or body fat [47][48][49] . Individuals with high BMI may have reflected a heterogeneous fitness group comprised of individuals with high body fat percentage and elevated weight due to muscle mass 47,49 . Alternatively, self-reported intensity and time spent exercising may reflect the individual's subjective experience 38,43,50 . Individuals with a higher BMI may experience more difficulty exercising and report objectively less intense exercise as more intense and as lasting a longer period of time. In addition to biases of subjective experience, self-report of fitness behaviors may be further distorted by the presence of attenuated symptoms 38,50 .
Self-reports of perceived fitness 13 reflected the overall severity of positive and negative symptoms, even when accounting for other features of health behavior (e.g., frequency and intensity of exercise). Exploratory analyses modeled the difference in perceived fitness to actual fitness; the errors of perceived fitness were related to specific symptoms of self-perception (i.e., grandiosity) and motivation (i.e., avolition). Higher symptoms of grandiosity www.nature.com/scientificreports/ and avolition are related to more distortion in self-perception of fitness. Errors in self-perception of fitness were defined as the discrepancy between self-reported perceived fitness and the scores that we would predict based on objective indices of fitness. These findings build on extant findings of clinical influence on some self-report measures 38,[51][52][53][54] . Collectively these symptom analyses support the possibility that symptomatology may distort the ability of CHR individuals to accurately self-assess fitness. Future studies should also examine the possibility that these errors may be affected by deficits in memory. Despite the many strengths and novelty of this study, there are some relevant limitations. First, the healthy controls did not complete the same self-reported fitness measure, as a result it is unclear if these distortions in self-reported fitness are unique to the CHR group or a larger problem with self-reported fitness data 20,[35][36][37] . It is notable, however, that self-report items did relate to the symptoms that uniquely define individuals at CHR, which somewhat mitigates concerns regarding the relevance of these self-reported discrepancies to features of the CHR group. Although BMI is a useful metric in distinguishing CHR from healthy controls, it was not entirely clear if increased BMI reflected increased muscle mass or body fat percentage [47][48][49]55 . Future studies on this topic should consider including additional metrics of fitness such as waist to hip ratio 22 or skinfold thickness 55 to estimate body fat percentage in addition to body mass index. It is also notable that the current group comparisons were exploratory analyses of available samples and were less than ideal comparisons. Future work should collect both CHR and healthy control subjects from the same location and complete the same exercise protocol, which should be repeated at least once. Additionally, future studies should match groups on critical features, such as BMI, age, sex, race, and ethnicity. Finally, future studies should also include questionnaires that are more common to www.nature.com/scientificreports/ the larger exercise literature, including the Simple Physical Activity Questionnaire (SIMPAQ; 56 this would help integrate future findings into a larger exercise literature. The current paper attempted to address these potential differences from the two contributing sites varied in the specific exercise approach (e.g., treadmill and cycling) by normalizing the data using a national population standard 44,57 . Although this correction is not perfect, the concern about site differences may be somewhat mitigated by the equivalent peak heart rates across exercise types, which suggests that the exercises were of roughly equivalent rigor. Finally, the sites varied according to age; although the model accounted for variability related to age, it remains possible that age impacts the current findings. It is notable that the difference in age between samples was quite small (3 years); as such, the current study is not well suited to interpret the impact of age on fitness.
The current study sample was similar in size to comparable extant literature 1,9,11,32 . Nevertheless, the field would benefit from larger sample sizes to examine additional variables that may affect fitness, including local access to exercise, culture, race, income, and ethnicity. Additionally, this smaller sample size restricted our sensitivity to symptoms within the CHR group. As a result, analyses were restricted to strong candidate subscales. Notably, the presence of significant relationships, despite this reduced power, suggests that this area is a promising line of inquiry for future studies. However, these follow-up analyses should be interpreted with appropriate caution, given their exploratory nature.
In conclusion, despite CHR individuals being objectively less fit than their peers, their self-reported perceived fitness related to symptoms and not objective levels of fitness. Additionally, the errors in perceived fitness may reflect grandiosity and avolition. These findings add to a larger literature that suggests attenuated positive and negative symptoms contribute to a disturbed perception of self 54 . Future studies should use caution in depending on self-report measures of fitness alone. Additionally, exercise interventions may benefit from emphasizing objective self-report items (e.g., time spent exercising) and objective metrics (VO 2 max) or concrete features of exercise behavior (time and intensity of exercise) rather than relying on subject recall 38 or subjective intensity 50 .