What a prescribing limit actually does

The intuition behind a numeric prescribing limit is reasonable: too much of this drug is being prescribed, so cap it. The reason it does not work as intended is worth spelling out, because the same reasoning applies to most population-level instruments.

A limit acts on everyone the same way

Clinical guidance says: reassess at this point, take extra precautions. A limit says: no. The first produces different actions for different patients. The second produces the same action for the patient whose metabolism makes a high number modest and the patient for whom a low number is already dangerous. Given roughly 13-fold variability in the dose people require, that is a lot of information being discarded.

It moves the decision away from the person with the information

A limit implemented in a payer edit or a pharmacy system is enforced by someone who cannot see the diagnosis. The clinician who can see it is now arguing with a system rather than treating a patient. Whatever the merits of the number, the decision has relocated to the party with the least context.

It gets applied outside its scope, reliably

This is not speculative. The 2016 CDC guideline was written for primary care clinicians treating chronic pain outside cancer, palliative and end-of-life care. Between 2016 and 2018, 527 federal and state opioid-related policies were approved; 170 imposed prescribing limits, and 35 explicitly invoked the CDC guideline. It was applied to specialists, to established patients and to populations it excluded by name.

CDC said so itself in 2022, listing among the misapplications “extension to patient populations not covered,” “rigid application of opioid dosage thresholds,” “duration limits by insurers and pharmacies,” and “patient dismissal and abandonment.” It then removed the number from the recommendation entirely, stating that this was to discourage exactly that.

And it succeeds at its own metric while the outcome moves elsewhere

Prescribing fell. Opioid-involved overdose deaths rose from 42,435 in 2016 to 82,851 in 2022 — a 95% increase — before falling to 55,007 in 2024. Whatever was driving mortality across that period was substantially not the thing being regulated.

The general lesson

Any number issued as guidance will be implemented as a rule by every system downstream that needs something machine-readable. If you publish a threshold, assume it will be enforced as a ceiling by parties you did not address, against patients you excluded, and design accordingly — or do not publish a number.

More: dose thresholds and variability · the guidelines themselves.