Predictors of cardiorespiratory fitness improvement in phase II cardiac rehabilitation

Abstract

Background

Cardiac rehabilitation (CR) improves cardiorespiratory fitness (CRF) and has been shown to reduce cardiovascular events and death. However, data about predictors of fitness improvement during CR are limited and conflicting. The objective of this study was to determine predictors of improvement in metabolic equivalents of task (METs) based on formal exercise testing throughout phase II CR.

Methods

We retrospectively reviewed 20 671 patients enrolled in phase II CR at our center from 2006 to 2016. Patients who completed 36 sessions and had entry and exit exercise stress tests were included for study. The short form‐36 (SF‐36) questionnaire was used to assess quality‐of‐life. Univariate and multivariate regression analyses were performed to determine independent predictors of METs improvement.

Results

Of the full cohort, 827 patients completed 36 sessions and had entry/exit stress test data. The majority of patients (N = 647, 78.2%) had improvement in METs (mean Δ 2.0 ± 1.2 METs), including patients ≥65 and < 65 years old (77% vs 79%, P = 0.46 for difference). METs improvement was negatively associated with body mass index, diabetes, left ventricular dysfunction, and poor baseline fitness; and positively associated with SF‐36 score (P < 0.05 for all). After multivariable adjustment, improvement was no longer affected by age, ejection fraction, or baseline fitness. Patients with poor fitness (≤5 METS) and adequate fitness (> 5 METS) both had improvement, with no statistical difference between the groups (P = 0.36).

Conclusions

In a large cohort of phase II CR patients, improvement in CRF was seen in the majority of patients across all ages, genders, and levels of baseline fitness.

1. INTRODUCTION

Cardiovascular diseases remain the leading cause of death globally, representing 31% of all deaths.1 Most cardiovascular diseases can be prevented by addressing behavioral risk factors. This has led to the emergence of cardiac rehabilitation (CR) services; which are comprehensive, long‐term programs consisting of nutritional counseling, aggressive risk factor management, psychosocial counseling, and exercise training.2 The American Heart Association/American College of Cardiology (AHA/ACC) guidelines on management of coronary artery disease (CAD) and congestive heart failure (CHF) recommend CR for patients with CAD and CHF (class 1 and IIa recommendations, respectively).3, 4, 5

The importance of exercise training and the subsequent improvement in cardiorespiratory fitness (CRF) has been supported by studies showing reduced cardiovascular mortality by up to 36%.6 Despite these benefits, few studies have investigated factors that affect the degree of improvement achieved, and results are conflicting. For example, some studies show that low baseline CRF predicts lower benefit with CR.7, 8 However, earlier data from the Massachusetts Association of Cardiovascular and Pulmonary Rehabilitation showed that those with low baseline CRF and continuing with CR had the greatest reduction in long‐term outcomes.9

Our center treats a large number of CR patients and encompasses a wide geographic area throughout Northeast Ohio. Data from this contemporary cohort could help to answer important questions on the benefits of CRF‐enhancing programs. The purpose of this study was to determine predictors of improvement in metabolic equivalents of task (METs) based on formal exercise testing in patients completing a 36‐session program of phase II CR.

2. METHODS

2.1. Study design

This was a retrospective study of all patients enrolled in phase II CR at the Cleveland Clinic, Ohio between 1 January 2006 and 31 December 2016. Full completion of the phase II CR program was considered to be 36 sessions since that is the maximum number of reimbursable sessions per the Centers for Medicare and Medicaid Services.10 Only patients who completed 36 sessions, and who had an exercise treadmill test at entry and exit were considered for the study.

Patient demographics, comorbidities, and medications were collected by exercise physiologists and nurses at the time of entry. Complementary clinical data and other missing variables were manually derived from the electronic medical records (EMR). All patients were de‐identified at the time of analysis, and the study was approved by the Institutional Review Board at the Cleveland Clinic.

2.2. Study cohort

All patients who started CR at our center underwent exercise treadmill testing for quantification of CRF. The treadmill testing protocol (eg, Bruce, Cornell) was selected at the discretion of the supervising clinician. Patients who completed all 36 sessions also had treadmill testing at the exit of phase II CR. CRF, expressed as METs, was calculated based on achieved speed and incline. The testing was symptom‐limited and heart rate was used to judge test adequacy. Testing was terminated at the discretion of the supervising clinician for reasons that included abnormal hemodynamic responses, ischemic ST‐changes, or exercise‐limiting symptoms.

With the exception of METs on exit test, all variables were collected at the time of entry into the program and included: demographic data (age and sex), body measurements (height, weight, and body mass index [BMI]), comorbidities (known CAD, hypertension, hyperlipidemia, diabetes mellitus, and smoking status), left ventricular ejection fraction (EF), METs achieved, and quality‐of‐life parameters expressed as short form‐36 (SF‐36) mental and physical summary scores.

CAD was defined as previous acute coronary syndrome, percutaneous coronary intervention, or coronary artery bypass grafting. Hypertension and hyperlipidemia were defined by self‐reported history or documentation in the EMR. Diabetes mellitus was defined by HbA1c ≥ 6.5%, self‐reported history, or use of anti‐hyperglycemic medications. Smoking status was defined by current cigarette smoking or last use within 1 year. Poor baseline CRF was defined as entry METs ≤5, as per AHA recommendations.11

The SF‐36 questionnaire was administered to all participants at entry into our program. The SF‐36 score is a generic health status instrument with 36 items, eight subscales that aggregate 2 to 10 items each, and two summary measures that aggregate the subscales—the physical and mental component summary scales. It is one of the most widely used generic health status instruments to assess health‐related quality of life. Among other conditions, it has been used rather extensively with cardiac patient populations.12, 13

2.3. Outcomes

The primary outcome was to assess the degree of association between each of the aforementioned variables with improvement in METs achieved based on a multivariate regression model. Secondary analyses of METs improvement were also performed in subgroups based on entry CRF, age, and any variables noted to be significant in multivariate regression analysis.

2.4. Statistical analysis

Demographic, clinical, and exercise variables were tabulated for all participants, and compared for those with a positive change in METs vs those without. Data were summarized as median and interquartile range (IQR) for continuous data, and number and percentage of non‐missing data for categorical variables.

Univariate analysis was performed to determine variables associated with a positive change in METs. Variables with P < 0.10 in univariate analysis were considered for inclusion in multivariate regression to determine independent associations. The alpha level for significance in the multivariate regression was <0.05.

Statistical analyses were performed using Student's t‐test for continuous variables and χ² test for categorical variables to examine the difference between groups. Linear regression was used to examine the association between change in METs and SF‐36 physical score. All statistical analyses were performed using IBM SPSS 23.

3. RESULTS

3.1. Patient characteristics

We reviewed 20 671 patients enrolled in phase II CR from 2006 to 2016. Of those, 1009 patients completed a full 36‐session program but only 827 had both entry and exit stress tests. Of the final cohort, 647/827 patients (78.2%) had an improvement in CRF at the end of phase II CR based on a positive change in METs. Overall, the mean improvement in METs for the whole cohort was 1.5 (±1.4), and for those with a positive change in METs was 2.0 (±1.2 METs). Our cohort was predominantly male (70.7%), the mean age was 62.5 years (±11.9), and the mean baseline CRF was 7 METs (±2.6) (see Table 1).

Table 1.

Baseline characteristics and univariable associations with metabolic equivalents of tasks improvement

	All (n = 827)	Improvement in METs (n = 647)	No improvement in METs (n = 130)	OR for improvement	P‐value
Age, mean ± SD	62.5 ± 11.9	62.2 (11.7)	63.6 (12.6)	0.99 (0.976‐1.004)	0.156
Male, n (%)	585 (70.7)	463 (71.6)	122 (67.8)	1.2(0.84‐1.71)	0.324
BMI, median (IQR)	27.7 (25‐31.7)	27.6 (24.8‐31.3)	28.2 (25.8‐33.0)	0.96 (0.94‐0.99)	0.015
Diabetes, n (%)	237 (28.8)	162 (25)	75 (41.7)	0.47 (0.33‐0.67)	<0.001
Hypertension, n (%)	572 (69.9)	438 (67.7)	134 (74.4)	0.71 (0.49‐1.04)	0.078
Dyslipidemia, n (%)	661 (80.5)	512 (79.1)	149 (82.8)	0.76 (0.49‐1.18)	0.225
Smoking, n (%)	110 (13.3)	87 (13.4)	23 (12.8)	1.05 (0.65‐1.72)	0.824
CAD, n (%)	606 (73.3)	471 (72.8)	135 (75)	0.89 (0.61‐1.30)	0.555
Entry METs <5, n (%)	194 (23.5%)	134 (20.7%)	60 (33.3%)	0.52 (0.36‐0.75)	<0.001
Entry EF < 50%, n (%)	170 (20.5)	121 (18.7)	49 (27.2)	0.63 (0.43‐0.92)	0.018
SF‐36 mental summary score, median (IQR)	55.5 (49.3‐59.1)	55.8 (50‐59.1)	54.3 (46.3‐59.5)	1.0 (0.99‐1.01)	0.16
SF‐36 physical summary score, median (IQR)	48.4 (38.9‐53.7)	49.5 (40‐53.9)	44.6 (33.6‐52.5)	1.04 (1.03‐1.08)	<0.001

Abbreviations: BMI, body mass index; CAD, coronary artery disease; IQR, interquartile range; EF, ejection fraction; METs, metabolic equivalents of tasks; OR, odds ratio; SF‐36, short form‐36.

Bold entries are for values that are statistically significant with P < 0.05.

The most common indication for enrollment in CR was CAD (N = 566, 68.4%) followed by valvular heart disease (N = 113, 13.7%) and congestive heart failure (N = 45, 5.4%). The rest of the cohort (N = 103, 12.5%) had multiple co‐indications. (Figure S1, Supporting Information).

3.2. Univariable associations

After performing univariate regressions, improvement in METs was negatively associated with BMI ( odds ratio [OR] 0.96 [95% confidence interval (CI) 0.94‐0.99], P = 0.015], diabetes [OR 0.47 95% CI 0.33‐0.67], P < 0.001), low METs at entry (OR 0.52 [95% CI 0.36‐0.75], P < 0.001), and low EF (OR 0.63 [95% CI 0.43‐0.92], P = 0.018), but positively associated with the SF‐36 physical summary score (OR 1.04 [95% CI 1.02‐1.08], P < 0.001) (Table 1).

There was no significant association between degree of improvement in CRF and age or gender, and no association with having CAD, hypertension, dyslipidemia, or smoking status.

3.3. Multivariable associations

A multivariate regression analysis was then performed after adjusting for age, gender, and all univariable associations with P < 0.1 (ie, BMI, diabetes, hypertension, low entry METs, low EF, and SF36 physical summary scores). After multivariable adjustment, improvement in CRF was no longer associated with age, BMI, poor baseline METs, or low EF (P > 0.05 for all). Patients with higher SF‐36 physical scores were statistically more likely (OR 1.03 [95% CI 1.01‐1.05], P = 0.005) and patients with diabetes were less likely (OR 0.62 [95% CI 0.42‐0.91], P = 0.015) to show improvement in METs (Table 2). However, little variability in METs improvement was explained by either of these variables—R² was only 0.05 for SF‐36, and confidence intervals for METs improvement in diabetic vs non‐diabetic patients appeared to overlap (Figures S2 and S3).

Table 2.

Multivariable associations with metabolic equivalents of tasks improvement

	OR (95% CI)	P‐value
Age	0.99 (0.98‐1.01)	0.421
Male	0.95 (0.63‐1.43)	0.808
BMI	0.98 (0.95‐1.01)	0.228
Entry Mets <5	0.80 (0.56‐1.28)	0.355
Diabetes	0.62 (0.42‐0.91)	0.015*
Hypertension	0.95 (0.61‐1.47)	0.821
EF <50	0.76 (0.46‐1.08)	0.112
SF‐36 physical score	1.03 (1.01‐1.05)	0.005*

Abbreviations: BMI, body mass index; CI, confidence interval; EF, ejection fraction; OR, odds ratio; SF‐36, short form‐36.

3.4. Subgroup analyses

Mean age in our cohort was 62.5 years, so we dichotomized the cohort to those ≥65 and < 65 years old. Subjects ≥65 years were more likely to have low METs at entry (31.1% vs 17.4%, P < 0.001) and a lower mean BMI (28.1 vs 29.6, P < 0.001). However, the majority in each group achieved improvement in METs with a non‐statistically significant difference between both groups (77% vs 79.2%, P = 0.46) (Table S1).

We also performed a subgroup analysis according to baseline fitness at entry (≤5 vs > 5 METs) (see Figure 1). We found that patients with poor fitness at entry were significantly different from their counterparts with adequate fitness in almost every aspect (Table 3); they were older, more likely to have diabetes, and had a higher mean BMI and a lower mean SF‐36 physical component score. This explains the initial negative correlation between poor baseline fitness and likelihood of achieving improvement on exit test. However, after adjusting for the variables above, the impact of baseline fitness was no longer statistically significant (P = 0.36).

Figure 1.

The association of A, age, and B, entry metabolic equivalents of tasks with the degree of cardiorespiratory fitness improvement

Table 3.

Baseline characteristics according to baseline fitness category

	Entry METs ≤5 (n = 194)	Entry METs >5 (n = 633)	P‐value
Age, mean (SD)	67.0 (11.6)	61.1 (11.7)	<0.001
Male, n (%)	95 (49.0%)	490 (77.4%)	<0.001
BMI, mean (SD)	30.9 (7.1)	28.3 (5.4)	<0.001
EF > 50%, n (%)	121 (62.4%)	508 (80.3%)	<0.001
Diabetes, n (%)	91 (47.0%)	146 (23.1%)	<0.001
SF‐36 physical score	38.3 (10.1)	47.9 (8.8)	<0.001
Improvement in METs, n (%)	134 (69.1%)	513 (81.0%)	<0.001

Abbreviations: BMI, body mass index; EF, ejection fraction; METs, metabolic equivalents of tasks; SF‐36, short form‐36.

Since SF‐36 score and diabetes were still predictive of change in METs after multivariate adjustment, we also conducted additional subgroup analyses of these variables. With regards to the SF‐36 score, we evaluated the distribution of physical summary scores across the cohort (Figure 2). Median SF‐36 score was 48, so we dichotomized patients with scores above and below the median. Patients with SF‐36 scores above the median were noted to have a particularly significant increase in CRF [OR 1.56 (95% CI 1.05‐2.29), P = 0.026]. Furthermore, a linear regression model showed that regardless of the reference score, there was an associated 1.0 increase in METs on exit stress test for almost every 30‐point increase in the SF‐36 physical summary score although variability was wide ( Figure S2A).

Figure 2.

Distribution of the short form‐36 scores A, across the whole cohort and B, in relation to metabolic equivalents of tasks improvement

With regards to diabetes, patients with diabetes were noted to be older, more obese, more likely to have low METs, low EF, and a lower SF‐36 physical summary score. Characteristics of patients with diabetes in our cohort are represented in Table S2. Although diabetic patients were shown to have a statistically lower likelihood of CRF improvement, the CIs for absolute difference in METs improvement overlapped for patients with and without diabetes (Figure S3A). This suggests that even if there is a small quantitative difference in METs, absolute difference may not be clinically meaningful.

Finally, analysis of patients with low EF at entry showed that a negative association with CRF improvement was only seen in patients with severely reduced EF (<35%), but not in those with mild‐moderate reduced EF (35%‐50%) (Table S3).

4. DISCUSSION

Results from our study showed several important findings: (a) Baseline functional status was not correlated with likelihood of improvement in METs. That is, both patients with baseline METs ≤ and > 5 were able to increase their CRF throughout the CR program. (b) Neither age nor gender appeared to be factors in determining patients who will benefit from CR, with improvement observed in patients of all ages and both genders. (c) Patients with diabetes were noted to have less improvement in CRF throughout the CR program. This may be because of the fact that they were older, with higher BMI, lower EF, and lower baseline fitness compared to those without diabetes. (d) Progressively lower EF correlated with lower likelihood of CRF improvement in subgroup analysis. (e) Self‐perceived high quality‐of‐life positively predicted CRF improvement independent of typical confounders. In fact, patients with an SF‐36 physical score > 50 had a nearly 50% higher odds of improving CRF throughout the program.

CRF has been shown in many studies to be correlated with long‐term cardiovascular outcomes.14, 15, 16 As a result, improving CRF remains the most essential part of CR and the component where most effort should be targeted. Our study showed that the majority of CR patients (78.2%) had CRF improvement by mean 1.5 ± 1.4 METs. This is comparable to prior reports including a Cochrane meta‐analysis of 31 studies that showed a mean improvement of 1.55 METs (95% CI 1.21‐1.89).17

Unlike other studies which showed that the likelihood of benefit was inversely related to age,18, 19 our study showed benefit from CR across all age groups. Even when we stratified the cohort by age < 65 and ≥ 65 years old, there was improvement in CRF in both cohorts. This led to one of our main conclusions; that age does not affect the likelihood of benefit from CR, and as a result it should not be a source of tentativeness for referring elderly patients to CR programs. There was also no difference when it came to gender, since both male and female patients appeared to benefit in our study. Although that is an encouraging finding, women were still under‐represented in our study (29%), which is more of a reflection of poor enrollment of women in our and other CR programs across the country.20, 21, 22

Another unexpected yet encouraging finding was that CRF improvement was not related to baseline functional status. This is contrary to some data including from the APPROACH investigators and a recent study of >10 000 patients that showed mean baseline METs was among the strongest predictors of CRF improvement.7, 14 On the other hand, earlier data from the Massachusetts Association of Cardiovascular and Pulmonary Rehabilitation showed more significant improvement and significantly lower mortality among those who started with low fitness levels (METs ≤5) and improved throughout CR.9 It also showed that much of the variation in baseline METs was related to gender and age. Indeed, our results are consistent with those earlier findings since patients with low entry METs in our cohort were older, had higher BMI, and were more likely to be diabetic (Table 3 and Table S1); but after adjustment they still had significant improvement in CRF with no statistical difference as compared to their higher fitness counterparts.

With regards to weight, after adjusting for confounders our study showed that both obese and non‐obese patients were equally likely to benefit from CR. A similar finding was shown by Lim et al.23 in a study of 359 CAD patients which showed that all parameters of CRF (maximal METs, exercise duration, and maximal oxygen consumption) improved significantly in obese and non‐obese patients with no difference detected between the groups.

Similar to prior studies, diabetes proved again to be a negative factor when it comes to functional benefit from CR.24, 25, 26 This may stem from the fact that patients with diabetes are less likely to be enrolled in CR,21, 27 and less likely to be adherent.20, 28 Another possible explanation is lower baseline fitness. In our cohort, the mean entry METs for patients with diabetes was 5.8 compared to 7.6 METs in those without (P < 0.001). Patients with diabetes also had higher mean BMI (31.1 vs 28.0, P < 0.001). All of these factors in aggregate may have ultimately lead to worse response to CR in patients with diabetes. It is of utmost importance to mention that this should not be a barrier for diabetic patients to participate in CR programs. In fact, this data may suggest that even greater attention should be directed to such patients. Uncontrolled diabetes already doubles the risk of cardiovascular disease29 and CR has been thoroughly shown to reduce hospitalization, cardiovascular death, and all‐cause death in such patients.27, 28, 30 So aggressive efforts should be made to exert improvements in CRF patients with diabetes.

One of the novel findings in our study was the association between the SF‐36 physical summary score and degree of CRF improvement. There is mounting evidence for improvement in quality‐of‐life measures after CR programs,31, 32, 33, 34, 35 as well as evidence of such measures predicting risk of cardiovascular events36 and death.37 SF‐36 is essentially a questionnaire that can be delivered within a few minutes by any member of the CR program and does not require any testing or investigations. It is unique in that it depends only on the subjective feeling of a patient's wellness and functional capacity, but at the same time strongly predicts risk of hard outcomes.36, 37 Prior studies have shown that the SF‐36 predicts earlier death in those with lower scores and achieving no improvement in CRF during CR.12, 38, 39 Our study corroborates that data and adds to it by showing that the score also predicts likelihood of improvement in CRF within the program itself.

Finally, the impact of heart failure on CRF remains an interesting area of research. Prior data shows that those with reduced EF attain benefit from CR40, 41, 42 albeit to a lesser degree than patients with preserved EF. This was corroborated by our findings which showed that the lower the EF the less likely patients were to improve CRF (Online Table 3). It is possible that initial exercise plans in this subgroup were less aggressive, targeting a lower maximal exercise capacity and fewer number of sessions. As is the case for patients with diabetes, greater attention may be needed to exert larger benefit and more aggressive targets for these high‐risk subgroups.

A major limitation of our study was the potential for selection bias at our center which we described previously.22 Our study was also observational in nature over a long duration which could subject it to additional biases due to changes in CR participation and demographic patterns over time. Furthermore, patients who were enrolled and who completed the full CR program may have been a disproportionately more motivated section of the population, since traditionally the CR referral rates are approximately 60%43 and the participation rates are less than 30%.44 Also, women were under‐represented in our population, accounting for less than 30% of the cohort. Unfortunately, that is a reflection of CR program enrollment nationally and requires concerted efforts to improve enrollment and retention. Finally, the study did not assess other outcomes post CR, such as mortality data or ongoing change in other risk factors. Despite these limitations, this study is still one of the largest reports of predictors of CRF to‐date. Further studies are needed to understand the impact of baseline fitness, heart failure, and diabetes on CRF improvements throughout CR.

5. CONCLUSION

Phase II CR was associated with significant improvement in CRF throughout 36 sessions of exercise. Improvement was seen in both younger and older patients, males and females, and patients with poor and adequate baseline fitness. This bolsters the AHA/ACC recommendations for phase II CR, and supports the guideline that CR should be offered to all eligible patients with particular efforts targeting high‐risk and low‐enrollment populations such as those with diabetes, those with heart failure, and women.

CONFLICTS OF INTEREST

The authors declare no potential conflict of interests.

Supporting information

Table S1. Baseline characteristics according to age category

Table S2. Baseline characteristics according to presence of diabetes

Table S3. Stratified analysis of the association of low ejection fraction with metabolic equivalents of tasks improvement

Figure S1. Indications for enrollment in phase II cardiac rehabilitation

Figure S2. Scatterplot for association of A, short form‐36 and B, body mass index with the change in metabolic equivalents of tasks

Figure S3. association of A, Diabetes and B, short form‐36 physical summary score with the degree of improvement

Abu‐Haniyeh A, Shah NP, Wu Y, Cho L, Ahmed HM. Predictors of cardiorespiratory fitness improvement in phase II cardiac rehabilitation. Clin Cardiol. 2018;41:1563–1569. 10.1002/clc.23101