Heart Rate Score versus the programmed lower rate limit

The thing to hold onto is that Heart Rate Score is a histogram-shape statistic, not a rate. It is the percentage of all atrial beats — paced and sensed — that fall in the single tallest 10 bpm-wide bin of the device’s rate histogram. It quantifies rate monotony: how much of the rhythm is concentrated in one narrow band. And the lower rate limit acts as a hard left wall on exactly that histogram. That mechanical relationship is the whole story, and it runs deeper than “higher floor leads to more atrial pacing.”

01The mechanical chain: the floor as a hard left wall

Walk the chain concretely. HRSc is dominated by its mode, and the mode is wherever the largest contiguous mass of beats sits. In most patients that mass is the resting and nocturnal rhythm — sleep alone is 6–9 hours of the lowest-rate beats of the day. When you raise the floor from 50 to 60 bpm, every beat that would naturally have fallen in the 50–59 range gets censored upward and pinned at the floor, and that nocturnal block — roughly a third of the 24-hour beat count — collapses into the 60–69 bin. The mode balloons.

A patient at LRL 50 whose intrinsic rhythm disperses across 50–80 during the day with genuine nocturnal dipping might have a tallest bin of about 25% and an HRSc in the low stratum. Move the floor to 60, pin the nights and sedentary daytime to the floor, and the 60–69 bin can absorb 60–70% of all beats. HRSc jumps from roughly 25% to roughly 65% — across two of the risk strata used in the validation work (the <30% / 30–69% / ≥70% cut points from the pacemaker atrial fibrillation analysis) — without a single thing changing about the myocardium or the autonomic nervous system. The elevation is an artifact of where you placed the wall.

02Phenotype-dependence: loudest where least justified

The phenotype-dependence is the part that is genuinely counterintuitive and clinically useful. The magnitude of the HRSc penalty from raising the floor is inversely related to how much the patient needs the higher floor. In a high-vagal-tone, high-HRV individual, the intrinsic rhythm has wide beat-to-beat dispersion — respiratory sinus arrhythmia, nocturnal vagal surges, a deep resting trough. That dispersion is precisely what keeps HRSc low at LRL 50, and it is precisely what a 60 floor obliterates, so the HRSc swing is large. In a patient with an intrinsically monotonous rate — sick sinus, low HRV, blunted autonomics — the histogram is already a narrow spike, HRSc is already high, and moving the floor barely shifts it. So the people who pay the biggest HRSc penalty for a higher floor are the ones with the least indication for it, and vice versa. The floor change is loudest on the metric exactly where it is least justified physiologically.

03The inferential trap: a mediator mistaken for a cause

Now the inferential trap, which is the same shape as the cardiac resynchronization registry confound — HRSc is the mediator that makes that confound legible. The HRSc-to-outcome associations (mortality in ICD and CRT-D cohorts, new-onset atrial fibrillation in pacemaker cohorts) were established in populations where a high HRSc reflected intrinsic chronotropic incompetence and autonomic disease. There, HRSc is a marker of the sick substrate. When you manufacture a high HRSc iatrogenically by raising the floor in an autonomically intact person, you have reproduced the statistic without the underlying pathology — reverse causation applied to a programmable parameter. The arrow from the lower rate limit to HRSc is mechanical and strong; the arrow from HRSc to outcome is, in the validation cohorts, largely disease-marking. Moving the mediator without touching the disease should not be expected to transfer the risk.

04DANPACE II as the empirical check

DANPACE II is the closest empirical check on that inference, and it is reassuring in the right direction. The DDDR-60 arm necessarily ran higher atrial pacing percentages, and therefore higher HRSc, than the DDD-40 arm, yet it had no excess atrial fibrillation — if anything the lower-pacing arm gained nothing on atrial fibrillation and lost on syncope. So the trial that most directly varied the floor produced exactly the dissociation the confounding argument predicts: more pacing and higher implied HRSc in the 60 arm, without the adverse arrhythmic signal the observational HRSc literature would have led you to fear. It is not a clean HRSc endpoint — they did not report HRSc as such — but the mechanism and the outcome point the same way.

05Consequence for longitudinal monitoring

The consequence for any longitudinal monitoring is sharp and easy to get wrong: a lower-rate-limit change is a structural break in the HRSc time series, not a data point within it. You cannot trend pre- and post-change HRSc as though the difference reflected an autonomic shift, because most of the delta is the wall moving. The same contamination hits the metrics derived from the same rhythm. PRSA-based deceleration capacity integrates over rate decelerations, and the deepest decelerations are nocturnal; a hard floor censors exactly those excursions, biasing deceleration capacity downward — toward the worse direction — as a pure artifact. DFA α1 and the time-domain HRV indices are similarly truncated, because you have clipped the low-rate tail that carries much of their variance. So if the floor moves mid-series, every one of those autonomic surveillance metrics inherits a discontinuity that has nothing to do with the patient’s physiology, and interpreting the step as a change in autonomic health would be a category error.

06Decision-theoretic compression

The programmed floor and HRSc are not independent variables, so HRSc cannot serve simultaneously as a tuning target and an outcome surrogate. Lowering the floor to “improve” HRSc is partly just relabeling the histogram. The defensible use is the reverse — hold the floor fixed at whatever the chronotropic indication actually requires, and read HRSc and the autonomic metrics only within epochs of constant programming, treating any reprogramming as a hard segmentation boundary. Across such a boundary, the only honest comparison is within-rate-band sensed-beat dispersion, not the whole-histogram mode.