Vancomycin Dosing While the Kidneys Are Still Deciding
A recovering kidney can’t be dosed around by a single trough level. The disagreement is whether the better method is worth adopting mid-admission, or whether the current tool, used more carefully, gets close enough.
E.V., a 70-year-old retired librarian who still volunteers shelving donated books twice a week, was admitted five days ago with MRSA pneumonia complicated by bacteremia, and spent the first thirty-six hours in the ICU on vasopressor support before stabilizing. Acute tubular necrosis developed early in her course, her creatinine peaking at 2.1 from a baseline of 0.9, and she has been on vancomycin throughout — dosed initially by the hospital’s standard trough-based nomogram, the same protocol used for essentially every other patient on the unit. Her kidney function has now begun recovering on its own, creatinine down to 1.5 and falling, but not in a straight line: two consecutive daily draws have shown different rates of improvement, and her most recent trough came back lower than the dose that produced it would have predicted twenty-four hours earlier.
She has no history of kidney disease before this admission, and her team expects a real return toward her own baseline over the coming week, which is part of what makes her vancomycin dosing harder rather than easier — a stable, chronically reduced eGFR would be simple to dose around; a kidney actively changing day to day undercuts the entire premise of a single trough level standing in for her true drug exposure. The pharmacy team flagged her case specifically because two consecutive trough-guided dose adjustments have overshot in opposite directions, and the question in front of the group is whether her renal trajectory itself is the reason a trough-only approach keeps missing rather than a dosing error to correct within the same method.
A trough measures the tail end of a dosing interval, drawn just before the next dose, which means it reflects renal clearance as it stood over the previous eight to twelve hours — useful when that clearance is roughly constant, misleading when it isn’t. Her creatinine fell nearly twenty percent between two consecutive daily draws this week alone, a rate of change the standard nomogram was never built to track, since it assumes the patient behind the numbers is dosing into a steady state rather than a moving target. That gap between what the tool assumes and what her kidneys are actually doing is the entire reason two adjustments in a row landed on opposite sides of her true exposure instead of converging toward it.
ICU step-down, pharmacy rounds
This is exactly the scenario the 2020 vancomycin consensus guideline — Rybak et al., jointly issued by ASHP, IDSA, PIDS and SIDP — moved away from trough-only monitoring for. A trough reflects the renal function that produced it, not her renal function today — and hers is changing daily. AUC-guided, Bayesian-modeled dosing uses her actual concentration-time data to estimate exposure directly, which is the only approach that keeps pace with a kidney that won’t sit still.
I don’t disagree with the rationale, but I want to name the real constraint before we commit to it — we don’t have a validated Bayesian software license on this unit. Building an AUC estimate by hand, with two-level sampling nobody here does routinely, risks introducing a calculation error into a case that’s already moving fast, trading one kind of imprecision for another kind our staff isn’t practiced at.
I think this resolves without either of you giving up your actual point. We use the trough method we have, but we stop treating one trough as sufficient — two levels per day through the unstable window, tracked as a trend rather than a single number feeding the nomogram in isolation. It won’t model her exposure as precisely as true AUC dosing would, but it recovers most of the benefit without requiring software this unit doesn’t have tonight.
Agreed: vancomycin continues on intensified twice-daily trough monitoring through the remainder of her renal recovery, with dose adjustments made against the two-level trend rather than any single value.
Also agreed, separately from tonight’s dosing decision: the pharmacy team will bring a formal proposal for Bayesian AUC-dosing software to the P&T committee, using this case as the concrete example of where the current method’s limits actually show up.