Design: Systematic review and random-effects meta-analysis of randomised controlled trials
Population: 13 trials, 958 adults with cardiovascular disease
Primary outcome: Peak oxygen uptake, systolic and diastolic blood pressure
Effect: VO₂ peak +2.58 mL/kg/min (95% CI 1.16 to 4.00, I² 74%); systolic BP −1.80 mmHg (−7.42 to 3.81); diastolic BP −2.00 mmHg (−4.76 to 0.75)

Thirteen randomised trials totalling 958 adults with cardiovascular disease compared exercise-based telerehabilitation with usual care. Peak oxygen uptake improved by 2.58 mL/kg/min (95% CI 1.16 to 4.00, P = .003), a difference of the size that maps onto meaningful functional gain. Neither systolic blood pressure (−1.80 mmHg, 95% CI −7.42 to 3.81) nor diastolic (−2.00 mmHg, −4.76 to 0.75) moved convincingly.

The review is unusually candid about its own limits. Heterogeneity for the fitness outcome was I² = 74%, the 95% prediction interval ran from −1.28 to 6.44 and so includes no benefit at all, the Egger test suggested small-study effects at P = .03, and GRADE certainty was rated very low for every outcome. Smartphone delivery and professional-led supervision produced larger and more consistent gains, but no moderator reached significance.

Read that way, the finding is about access rather than equivalence. Telerehabilitation is a reasonable offer to a patient who cannot reach a centre-based programme — which in Indian practice is most of them — and should be presented as such, not as a like-for-like substitute.

Offer telerehabilitation to patients who would otherwise do no rehabilitation at all, which is the comparison that matters.
Prefer professionally supervised, smartphone-delivered programmes where there is a choice.
Do not count telerehabilitation towards blood pressure targets — treat hypertension on its own terms.
Measure something: a six-minute walk distance before and after gives you a local answer the literature cannot.

Why it matters

The case for telerehabilitation is access, and stating it as equivalence oversells what these trials measured.

The statistics, in plain English

The prediction interval is the number to read here. The confidence interval (1.16 to 4.00) describes the average effect across these 13 trials; the prediction interval (−1.28 to 6.44) describes what a new programme might achieve, and it includes zero. Combined with very low GRADE certainty and a positive Egger test for small-study effects, that means the direction is plausible and the magnitude is not dependable.