Knee Osteoarthritis: A Fifth Quarterly Corticosteroid Injection, Weighed Against Its Own Trial's Findings
A patient who feels her injections working asks for a fifth. The trial that established the every-three-months schedule found the drug did no better than saline for pain — and measurably worse for her cartilage.
Carol T., 61, keeps the books for a small landscaping company two mornings a week and spends most of the rest of her time in her own yard, where a half-acre of raised vegetable beds has become the thing she organizes her calendar around. Two years ago a bad spring — too much kneeling in wet soil, a year into a diagnosis she was still mostly working around rather than treating — cost her most of the planting season, and it was that lost season, more than the diagnosis itself, that pushed her toward her first injection. Her right knee has been the more troublesome of the two since then, and for the past eleven months she has been getting a triamcinolone injection into it roughly every twelve weeks — four so far. Each one, by her own account, buys her about ten good weeks before the last two or three turn achy again, at which point she starts counting down to the next appointment. She is here today, on schedule and with seed catalogs already on her kitchen table, asking for a fifth.
Her regimen sits almost exactly inside the protocol McAlindon and colleagues used in their 2017 JAMA trial to test this exact question: 40mg of intra-articular triamcinolone every three months, compared against saline, over two years, in 140 patients whose knee osteoarthritis looked much like hers. That trial found no significant difference in pain relief between the two arms — a WOMAC change of roughly ‑1.2 units on steroid versus ‑1.9 on saline, not a meaningful gap either way — while the steroid group lost significantly more cartilage thickness on MRI over the same period. Carol isn't an extrapolation from that trial's findings; her injection interval, dose, and duration place her inside the population it actually studied, not adjacent to it.
The tension the trial raises isn't abstract for her specifically. A regimen she experiences as clearly working — real function regained each time, real decline each time it wears off — produced, in a rigorously blinded comparison, no advantage over an inert injection, while doing something measurable and one-directional to the joint underneath it. That the trial's own blinding held up reasonably well (only 45% of its patients correctly guessed which arm they were in) makes it harder, not easier, to wave off the finding as a fluke of an unblinded design.
In clinic, requesting the fifth injection
Carol has gotten real relief from this four times in a row, and I don't think we should take that away from her on the strength of a single trial's aggregate result. Ten good weeks out of twelve, reliably, for over a year — that's not nothing, and it's not obviously explained away by expectation alone when it's held up this consistently.
I'd also point out what McAlindon's own limitations section concedes: pain was assessed every three months, never within the four weeks after an injection, which is exactly the window where a corticosteroid's benefit is expected to sit. A trial designed that way could not have detected the pattern Carol is describing — good weeks early, breakthrough late — even if it were entirely real and entirely pharmacologic.
I take the cartilage finding seriously, but she is sixty-one, not headed toward a knee replacement conversation anytime soon by her own report, and a millimeter-scale MRI difference over two years hasn't been shown to translate into a faster path to surgery in someone at her stage of disease.
The problem is that the trial testing exactly this regimen found saline did just as well for pain. Forty-five percent of that trial's patients correctly guessed which injection they'd gotten — blinding held reasonably well — and the steroid group still didn't outperform placebo. What Carol is crediting to triamcinolone is more likely the injection itself: the needle, the fluid, the appointment, the attention. Saline would very plausibly give her the same ten weeks.
The measurement-window objection is fair, and I'll grant it limits what the trial can rule out — but it cuts both ways. Carol doesn't report four good weeks. She reports ten, then decline, on a twelve-week cycle. That pattern is squarely inside the interval the trial did measure, and across two years of it the steroid arm's WOMAC pain fell 1.2 points against saline's 1.9 — neither arm anywhere near the 3.94-point change the trial itself defined as a minimal clinically important improvement.
So the thing we're actually weighing isn't meaningful relief against a small structural cost. It's a real, measured acceleration in cartilage loss against a benefit the best evidence says probably isn't coming from the drug.
There's a version of this that doesn't require deciding whether the drug or the ritual is doing the work. The trial tested injecting every twelve weeks on a fixed calendar, whether or not she was flaring. That's not how most of us actually use this drug in practice — we reserve it for a real flare, not a standing appointment.
Move her from scheduled quarterly dosing to as-needed dosing, triggered by her own pain crossing a defined threshold rather than the calendar. If the effect really is mostly ritual, she'll do just as well with fewer total injections. If there's a real pharmacologic component we're not giving full credit to, she still gets it when she actually needs it — just with meaningfully less cumulative steroid exposure to the joint she's trying to keep functional for years to come.
Agreed: Carol's next injection will be triggered by her own pain crossing a threshold she and her physician define together, not by the twelve-week calendar mark. Weight-bearing quadriceps strengthening and continued weight management were discussed as adjuncts that don't carry the same structural tradeoff.
She likely receives an injection close to every ten to twelve weeks anyway, in which case little has actually changed except that the timing now reflects her symptoms rather than the calendar.
That would itself be informative — evidence the earlier fixed schedule was treating time, not need, and a reason to trust the new approach going forward.
Not agreed: whether to test a full corticosteroid holiday at some point to see whether her symptom pattern actually changes without it. The rheumatologist would like to try it; the primary care physician and orthopedic surgeon were unwilling to risk losing a real-world functional gain without a clearer sense of what would replace it.