Refining Maximal Heart Rate Estimation to Enhance Exercise Recommendations for Persons With Parkinson Disease

Abstract

Objective:

To derive and evaluate an alternative equation to estimate maximal heart rate in persons with Parkinson disease (PD) in the absence of structured exercise testing using observed maximal heart rate data from a maximal cardiopulmonary exercise test (CPET) and basic demographic and clinical data.

Design:

Baseline data from a randomized controlled trial.

Setting:

Academic Medical Center.

Participants:

Eighty-two persons with mild-to-moderate PD who completed a CPET.

Interventions:

Not applicable.

Main Outcome Measures:

A linear regression model was fit to maximal heart rate from CPET using the relaxed least absolute shrinkage and selection operator (lasso) and 7 readily clinically accessible candidate covariables. Model fit was assessed by leave-one-out cross-validation. Maximal heart rates from the CPET were compared with estimates from the regression model and from 2 traditional age-based maximal heart rate estimators: (220 − age) and [208 – (0.7 × age)].

Results:

The regression-based heart rate estimator was [166 – (1.15 × age) + (0.60 × resting heart rate)] and most closely fit the observed maximal heart rate from the CPET. The (220 – age) and [208 – (0.7 × age)] equations overestimated maximal heart rate for 88% and 94% of the participants, respectively. The mean square error of the regression-based estimator was 63% and 75% lower than those of the 2 traditional age-based estimators, respectively.

Conclusions:

Overestimating maximal heart rate generates prescribed target heart rate zones that are likely unachievable during aerobic exercise. The proposed regression-based maximal heart rate estimator most closely fit observed maximal heart rates from the CPET. Adoption of this estimator, based on both age and resting heart rate, may improve estimated maximal heart rate accuracy and thus provide more appropriate and achievable exercise heart rate zones for persons with PD in the absence of a CPET.

Parkinson disease (PD) is a progressive neurologic condition anticipated to affect 12+ million people worldwide by 2040.¹ Aerobic exercise has been touted as a universal prescription for persons with PD (PwPD)^2–4 and is recommended as part of comprehensive PD management to increase mobility and fitness and mitigate disease symptoms.⁵ Guidelines from the Parkinson’s Foundation and the American College of Sports Medicine Exercise recommend 90–150 minutes of aerobic exercise per week at moderate to vigorous intensities for PwPD.⁶ Intensity is most commonly prescribed as a percentage of maximal heart rate or heart rate reserve.⁷ A maximal cardiopulmonary exercise test (CPET) is the criterion standard for establishing maximal heart rate. However, for most of the general population and those with PD, a CPET is not medically indicated⁸, access to testing is limited, nor is it fiscally justified prior to initiating an exercise program. Considering the challenges in obtaining a CPET, maximal heart rate estimation is necessary to aid in exercise prescription. Well-used maximal heart rate estimators such as (220 – age)⁹ and [208 – (0.7 × age)]¹⁰ were created using data from healthy young and older adults. The equations rely on age due to a strong, inverse correlation (−0.9) between age and maximal heart rate.¹⁰ Notably, these equations have not been systematically evaluated in individuals with neurologic disease.

Chronotropic incompetence is the inability to increase heart rate commensurate with increasing physical activitydemands.¹¹ Chronotropic incompetence is a manifestation of cardiovascular autonomic dysfunction that frequently accompanies PD. Kanegusuku and colleagues¹² reported that the mean peak heart rate of PwPD during a CPET was only 84% of their healthy peers, and only 8% of individuals with PD reached their estimated age-predicted maximal heart rate. Another cohort of 90 PwPD reached an average of 86% of age-predicted maximal heart rate during CPET.¹³ Blunted heart rate appears to be present regardless of medication status, suggesting that the autonomic dysfunction is related to the disease process itself.¹⁴ Together, these studies and others^15–17 suggest that chronotropic incompetence is common in PwPD and prevents many from achieving their estimated maximal heart rate as estimated from conventional equations based solely on age.

The primary aim of this project was to use observed maximal heart rate data from a CPET to derive an alternative equation to estimate maximal heart rate in PwPD in the absence of structured exercise testing, using only basic demographic and clinical data. To address the aim, a regression model for maximal heart rate was created from participants enrolled in the CYClical Lower Extremity Exercise Trial (CYCLE) for PD,¹⁸ who completed a CPET as part of the clinical trial. Observed maximal heart rate during a CPET was compared with 2 traditional maximal heart rate estimators, (220 – age) and [208 – (0.7 × age)], and with a new regression estimator derived from the CYCLE data. It was hypothesized that the traditional estimators (220 – age) and [208 – (0.7 × age)] would overestimate maximal heart rate in individuals with PD because these estimators do not account for PD-related autonomic nervous system dysfunction and that the derived regression estimator would be better calibrated and most closely approximate the observed CPET values as assessed by mean squared error (MSE).

Methods

Participants enrolled in the CYCLE randomized controlled trial for PD¹⁸ were required to complete a CPET prior to initiating the intervention. Inclusion criteria for the CYCLE project were as follows: (1) individuals with a clinical diagnosis of PD classified as stage I-III Hoehn and Yahr, (2) between the ages of 30 and 75 years, and (3) not currently engaged in physical therapy or another interventional clinical study. Primary exclusion criteria included: (1) dementia, (2) deep brain stimulation, (3) previous stroke, (4) medical or musculoskeletal contraindications to exercise, and (5) uncontrolled cardiorespiratory risk factors as determined by the American Heart Association/American College of Sports Medicine exercise preparticipation questionnaire.¹⁹ For the current analysis, CYCLE participants receiving concurrent β-blocker treatment were excluded from the analysis. The decision to exclude those individuals stemmed from the traditional equations being developed on healthy, nonmedicated individuals¹⁰ and the well-documented effect of β-blockers decreasing heart rate at rest and with exertion.²⁰ The CYCLE trial was approved by the institutional review board of the Cleveland Clinic. Participants completed the informed consent process prior to initiating the study protocol.

Cardiopulmonary exercise test

Congruence between the mode of CPET testing and the exercise activity is most appropriate for calibrating exercise intensity.²¹ The CYCLE exercise intervention was performed on a stationary cycle; thus, the CPET was also completed on an upright stationary cycle ergometer (Lode Excalibur Sport with Pedal Force Measurement, Lode B.V.). Participants underwent the CPET in the on-medication state, operationally defined as taking their prescribed antiparkinsonian medication 1 hour prior to testing; non-PD medication was taken as prescribed. Participants refrained from food and drink for 4 hours, except clear liquids, and abstained from caffeine for 12 hours prior to the test. Resting heart rate and blood pressure were obtained during a 5-minute supine period prior to exercise testing.

Participants were fitted with a mouthpiece and nose clip for gas analysis, and cardiopulmonary data were assessed via calibrated open-circuit spirometry (PreVent flow sensor pneumotach and BreezeSuite cardiorespiratory diagnostic software). Expiratory gases were continuously collected throughout the test and reported as breath averages over 30-second intervals (Medgraphics Ultima Cardio). The peak volume of oxygen was defined as the highest averaged sample obtained during the CPET.

A stepwise protocol for the CPET was used. Participants were instructed to maintain a self-selected cadence throughout the entire test. The first stage began at a resistance of 25 W, and wattage increased in 25 W increments every 2 minutes until reaching 100 W at minute 8. After the 100 W stage, each stage increased by 50 W increments every 2 minutes until the participant reached volitional exhaustion, or until the American Heart Association/American College of Sports Medicine test termination criteria were achieved.²² A 12-lead electrocardiogram was continuously monitored, and manual blood pressure was taken during the last 30 seconds of each stage. Maximal heart rate was recorded as the greatest heart rate achieved during the CPET. Detailed results from the CPET have been previously published.¹³

Statistical analysis

Maximal heart rate was modeled using fully relaxed least absolute shrinkage and selection operator (lasso) regularization.^23,24 Candidate linear model predictors were age (y), sex, disease duration (y), levodopa equivalent daily dosage (mg), resting heart rate (beats per minute; bpm), and resting systolic and diastolic blood pressures (mmHg), which were selected for consideration because of their potential effect on maximal heart rate and clinical ease of collection. From all candidate variables, the fully relaxed lasso first selects a subset for inclusion in a model by minimizing a lack-of-fit criterion that balances the dual aims of precision and simplicity, penalizing model complexity. Then, the penalty is fully relaxed by fitting a final linear regression in the usual way, by ordinary least-squares, to only the variables previously selected. The lasso “tuning parameter” λ, which gives relative weights to the aims of precision and simplicity in the first step, was set conventionally by minimizing model MSE (ie, the mean of the squared differences between observed and model-estimated maximal heart rates) in 10-fold cross-validation. Pearson correlations were calculated between each of the included predictors and the observed maximal heart rate from CPET. The final model fit was assessed using the MSE from leave-one-out cross-validation, in which the full lasso process mentioned earlier was repeated on every possible dataset formed by omitting a single participant, and the omitted participant’s maximal heart rate was then estimated. The MSE was then calculated from the leave-one-out cross-validation estimated and actual values. A higher MSE indicates larger departures of the model estimates from the maximal heart rates observed during CPET.

The performance of the regression estimator was compared with those of the 2 current estimators from studies of healthy adults: (1) 220 – age and (2) [208 – 0.7 × age]. The respecive MSEs for estimating the observed maximal heart rates using equations (1) and (2) were calculated, and fractional reductions in MSE were calculated to separately compare the new regression estimator to each of the current estimators. All statistical analysis was done using RStudio 2024.09.1, R version 4.4.2.^a

Results

Ninety-four participants completed the baseline stress test. Of these, 12 were on β-blockers and were not included in the analysis. The remaining 82 participants were included in the analysis. Sixty-nine (84%) participants were prescribed antiparkinsonian medication. Participant demographics and CPET results are summarized in table 1.

Table 1.

Participant demographics and cardiopulmary exercise test results (N=82)

Variable	Value
Sex
Female	35 (42.7%)
Male	47 (57.3%)
Age (y)	62.7 (8.0)
PD duration (y)	3.3 (1.3, 4.5)
Levodopa equivalent daily dosage (mg)	450 (200, 638)
Height (cm)	171 (9.8)
Weight (kg)	82.0 (18.2)
Body mass index (kg/m2)	28.0 (5.5)
Resting heart rate (bpm)	70.0 (9.5)
CPET maximal heart rate (bpm)	137 (19.3)
Resting systolic blood pressure (mmHg)	128 (17.0)
Resting diastolic blood pressure (mmHg)	80.5 (9.6)
MDS-UPDRS III total score (on meds)*	34.3 (9.8)
Hoehn and Yahr stage (on meds)*
I	11 (13.4%)
II	58 (70.7%)
III	13 (15.9%)
Peak volume of oxygen (VO2peak) (mL/kg/min)	22.3 (6.0)
Peak respiratory exchange ratio	1.17 (0.1)

NOTE. Data presented as n (%), mean (SD), or median (Q1, Q3). Abbreviations: CPET, cardiopulmonary exercise test; MDS-UPDRS, Movement Disorder Society - Unified Parkinson’s Disease Rating Scale; PD, Parkinson disease

^*Includes 13 participants not taking PD medication.

The lasso algorithm selected 2 of the 7 candidate variables, age and resting heart rate, with respective Pearson correlations with maximum heart rate of −0.50 and 0.34, for inclusion in the regression model. The resulting maximal heart rate estimator was:

$${\text{estimated heart rate}}_{\max }=\left[166-\left(1.15\times \text{age}\right)+\left(0.60\times \text{resting heart rate}\right)\right].$$

For each of the 82 participants, figure 1 plots the observed maximal heart rate determined by CPET (y axis) against each of the three maximal heart rate estimators (x axis), with corresponding fitted simple linear regressions and a reference line of equality. As reflected in the rightwards shifts of their regression lines, the estimators (220 – age) and [208 – (0.7 × age)] overestimated maximal heart rate in 88% and 94% of the sample, respectively, whereas the regression estimator was well calibrated. The MSEs were 711, 1054, and 256 for (220 – age), [208 – (0.7 × age)], and the regression estimator, respectively. Thus, the regression estimator MSE was 63% and 75% smaller than those of the former 2 age-based estimators.

Fig 1.

Observed maximal heart rate during a cardiopulmonary exercise test (CPET) vs. each of 3 maximal heart rate estimators from 82 participants with Parkinson disease. The black reference line indicates when the estimated value equaled the observed value. For the (220 – age) and the [208 – (0.7 × age)] equations, the predicted maximal heart rate overestimated the observed value for 88% and 94% of participants, respectively. The regression estimator [166 – (1.15 × age) + (0.60 × resting heart rate)] showed the best calibration and least MSE with 63% and 75% reductions in MSE compared with the (220 – age) and [208 – (0.7 × age)] estimators, respectively.

Discussion

Maximal heart rate is challenging to estimate with accuracy in PwPD because of a high prevalence of chronotropic incompetence.^12,13,17 Accurate maximal heart rate estimation is critical in PD exercise prescription because a substantial body of literature supports the efficacy of moderate to vigorous intensity aerobic exercise in mitigating disease symptoms.²⁵ In addition to the importance of prescribing accurate exercise intensities in PD for symptom mitigation, overestimating maximal heart rate results in target heart rate zones that are unachievable and/or unsustainable for the recommended exercise duration. Difficulty achieving and sustaining a target heart rate is likely discouraging and presents yet another barrier to exercise in this population.²⁶ As hypothesized, the most accurate heart rate prediction estimator in this project was the relaxed lasso regression estimator incorporating both age and resting heart rate: [166 – (1.15 × age) + (0.60 × resting heart rate)]. The developed equation resulted in better calibration (fig 1) and lower mean square estimation error than either of the traditional heart rate estimators, (220 – age) and [208 – (0.7 × age)], which overestimated the observed maximal heart rate in 88% and 94% of participants, respectively.

In individuals without neurologic disease, the correlation between age and maximal heart rate is excellent (−0.9).¹⁰ In this study, a substantially weaker correlation between age and maximal heart rate in PD (−0.5) was found; moreover, the traditional estimators relying solely on age led to a near global overprediction of maximal heart rate in PwPD. The striking difference between the relationships ofof maximal heart rate to age for those with and without PD emphasizes the influence of autonomic dysfunction in this population. Stankovic et al²⁷ reported that 71% of people with early PD (Hoehn and Yahr stage I) reported at least 1 symptom of autonomic dysfunction; within a 3-year follow-up, all reported 1 or more autonomic symptom(s). While non-motor signs of some kind (eg, gastrointestinal, genitourinary, cardiovascular) are ubiquitous in PD, cardiovascular autonomic dysfunction is particularly relevant to exercise prescription. Because cardiovascular autonomic dysfunction occurs early and frequently in PD, the inclusion of resting heart rate to improve maximal heart rate prediction is justified. Resting heart rate, recorded as the lowest heart rate observed during a 5-minute supine resting period,⁷ is simple to capture in a clinical or nonclinical setting without undue burden on the patient or clinician. Given the 63% to 75% reduction in error compared with traditional prediction equations, the 5–10 minute effort is warranted. Such a recommendation is also consistent with best practice recommendations for clinical examination of cardiovascular autonomic dysfunction in PD.²⁸

Study limitations

Although the proposed PD-specific regression estimator appears a substantial improvement, it remains, expectedly, imperfect. For example, there were 4 PwPD with observed maximal heart rates below 100 bpm; the extremely low observed heart rates of all 4 were substantially overestimated by the new estimator. Nevertheless, its markedly better calibration and internally cross-validated MSE suggest that the regression estimator using both age and resting heart rate is markedly superior to the existing methods for exercise prescription in PD based solely on age.

The proposed equation should be externally validated on additional cohorts of individuals with PD, including those with various disease durations and severities. This can readily be accomplished at centers treating substantial numbers of PwPD by compiling, whether prospectively or retrospectively, the ages and resting and maximal heart rates of PwPD undergoing CPET who are not on heart rate-modifying medication and plotting the deviations of each of the estimators from the maximal heart rates, as shown in figure 1. Particularly of interest are such data from those early in their disease process, for whom understanding of cardiovascular autonomic dysfunction and chronotropic incompetence is less clear. Data from a newly diagnosed, unmedicated PD cohort reported only 13% were unable to reach 85% of their age-predicted maximal heart rate.²⁹ However, a preliminary report from Palma et al³⁰ reported that chronotropic insufficiency may be used as a prodromal sign of PD because heart rate response was blunted by approximately 15% in those who went on to be diagnosed with PD compared with those who were not diagnosed. Further work examining the effect of chronotropic incompetence on aerobic exercise in PD is warranted.

Conclusions

Traditional heart rate estimators overpredict maximal heart rate in PwPD because of the prevalence of chronotropic incompetence in this population. The derived equation based on age and resting heart rate, [166 – (1.15 × age) + (0.60 × resting heart rate)], substantially improved calibration and reduced the error in maximum heart rate prediction compared with traditional estimators in the CYCLE dataset. Adoption of the proposed estimator has the potential to yield more accurate exercise target heart rate zones for individuals with PD in the absence of a CPET.

List of abbreviations:

bpm

beats per minute

CPET

cardiopulmonary exercise test

CYCLE

CYClical Lower Extremity Exercise Trial

lasso

least absolute shrinkage and selection operator

LEDD

levodopa equivalent daily dosage

MDS-UPDRS

Movement Disorder Society - Unified Parkinson’s Disease Rating Scale

MSE

mean square error

PD

Parkinson disease

PwPD

persons with Parkinson disease

Data statements

The datasets analyzed are available upon request from the corresponding author.