The sentence arrives in the consultation in a standard form: "I tried physical therapy. It didn't work." Two questions follow it. How long did you go? And what did the therapy consist of? The usual answers — a few weeks, and a mixture of heat, stretching, and some exercises at the end — describe something real, but they do not describe a completed trial of exercise therapy. They describe an exposure to it. The distinction matters because the chart entry "failed PT" closes doors: it is quoted in referrals, in authorization decisions, and in the patient's own reasoning about what is left. An entry that consequential should be accurate.
The correction runs in both directions. Two weeks is too early to judge a therapy whose evidence was measured after programmes of weeks to months. And the completed trial, judged fairly, delivers a modest average effect — real, clinically meaningful for pain, smaller for function — not the cure that disappointed expectations were set against. Therapy fails patients in the chart far more often than it fails them in fact; it also helps them less completely than its reputation promises. Both are reasons to measure it properly.
The Guideline Clocks: Two Weeks Is Inside the Acute Window
The American College of Physicians guideline divides low back pain by duration: acute under four weeks, subacute four to twelve weeks, chronic beyond twelve weeks (Qaseem et al., Annals of Internal Medicine, 2017). Exercise therapy's strongest recommendation in that guideline attaches to chronic pain, where it is a first-line non-drug treatment. A verdict delivered at two weeks is delivered inside the acute window — before the category in which the evidence lives has even begun.
The acute window has a second property that corrupts early verdicts. The guideline's first recommendation rests on the observation that most patients with acute or subacute low back pain improve over time regardless of treatment. Early on, the natural course dominates the signal: improvement at two weeks is frequently the episode resolving on its own schedule, and non-improvement at two weeks is frequently the episode not having resolved yet. Neither reading measures the therapy. A trial judged inside that window is a measurement of the calendar, in whichever direction it flatters.
What the Trials Actually Measured
The exercise-therapy evidence is summarized in the Cochrane review by Hayden and colleagues (Cochrane Database of Systematic Reviews, 2021): 249 randomized trials, 24,486 participants, all with chronic non-specific low back pain. Outcomes were measured at the trials' earliest follow-up — after programmes that ran for weeks to months, because that is what a programme is. Against no treatment, usual care, or placebo, exercise reduced pain by an average of 15.2 points on a 0–100 scale, on moderate-certainty evidence — meeting the review's own 15-point threshold for a clinically important difference. The same comparison improved functional limitations by 6.8 points, against a 10-point threshold for clinical importance: the function effect, on average, did not reach it.
Both halves of that result are the expectation a completed trial is set against, and they belong in the same sentence. Pain relief that crosses the importance threshold; functional gain that is real but modest; and against other conservative treatments — rather than against doing nothing — the advantages shrank to 9.1 points for pain and 4.1 for function, neither clinically important. Exercise therapy's standing is not that it outperforms every alternative. It is that it outperforms inactivity, measurably, with the commonest adverse effect in the trials being increased muscle soreness. A patient promised a cure by week two has been set up to record a failure at exactly the moment the evidence would have begun measuring.
What a Dose Actually Is
A trial of exercise therapy consists of a prescribed programme: selected exercises, at a progressing load, performed in sessions and — this is the part most often skipped — as a home programme between them. The home programme is not homework appended to the treatment; in most programmes it is the majority of the dose. Attendance at six sessions with no home practice is a fraction of the intervention the trials delivered, and its result, good or bad, is not the intervention's result.
The intervention also has a definition, and the NICE guideline draws it: the recommendation is for an exercise programme tailored to the person's needs, preferences, and capabilities, and manual therapy is considered only as part of a package that includes exercise (NICE NG59, as amended). A course consisting of heat, ultrasound, and a table — passive modalities only — may be pleasant and may even settle a flare, but it is not exercise therapy, and its failure is not recorded as exercise therapy's failure in an accurate chart. When the two questions are asked — how long, and what did it consist of — the second is the one that most often vacates the verdict.
The Three False Failures
Under-dosed: too few sessions, no home programme, no progression of load. The dose makes the trial; a partial dose produces a partial test. Premature: judged at two weeks, inside the acute window, before the first point at which the evidence measures anything — the verdict describes the calendar, not the treatment. And wrong intervention: passive modalities delivered under therapy's name, judged as therapy's result. Each false failure has the same correction — run the actual trial — and each is preferable, as a chart entry, to the false conclusion it replaces, because a genuine failure changes the plan and a false one merely hardens it.
One finding from the trials belongs here because it rescues a common early dropout: the adverse effect reported most often across the Cochrane trials was increased muscle soreness. New exercise, loaded into deconditioned tissue, produces soreness on a training schedule — worst in the first days of a new load, settling as the tissue adapts. Soreness after sessions is the expected texture of a dose being delivered. It is distinct from the worsening that should be reported promptly: new or progressive leg weakness, radiating pain newly extending below the knee and intensifying, or any bladder or bowel change. The first is the programme working on the tissue. The second list is a checkpoint event.
What "Working" Looks Like at the Checkpoint
A dosed trial is judged at its checkpoint on direction across several measures, compared against the first visit: walking distance, sitting tolerance, the frequency and duration of flares, analgesic use, sleep, and a functional score. Pain intensity is included and deliberately not privileged — it is the slowest of the measures to move and the least reliable alone. The checkpoint pattern of a trial in progress is commonly function first: the patient walks further, sits longer, recovers from flares faster, while the pain number lags behind. Read at two weeks on the pain number alone, that trial looks failed. Read at its checkpoint on the full set, it is the Cochrane result arriving on schedule — function and pain improving on average, at the pace the trials measured.
What a Genuine Failure Looks Like
It exists, and it is specific: a prescribed programme, delivered at its full dose — sessions attended, home programme performed, load progressed — over its full course, with no functional movement at the checkpoint. That is a result. It closes the exercise question for that presentation, and it moves the plan: re-examination, imaging if it has not been done, and the next step in the sequence. Genuine non-response is also not a verdict on the patient. The Cochrane averages conceal a distribution: some participants responded substantially, some did not respond at all, and adherence explains only part of that spread. The chart should record what was delivered and what it changed. A record that complete is what the next decision — conservative or surgical — deserves to be built on, which is more than can be said for "failed PT, two weeks, heat and stretching," the entry this piece exists to retire.
Summary
Two weeks of physical therapy is not a trial of physical therapy. The guidelines place exercise therapy's evidence in chronic pain, measured after programmes of weeks to months. The Cochrane review of 249 trials found an average pain reduction that met its own importance threshold and a functional gain below its threshold, against no treatment or usual care. It is a modest, real effect that a fortnight cannot reproduce or refute. Judged on dose delivered and function measured, most early "failures" resolve into under-dosed, premature, or mislabeled trials. The genuine failure — a full dose, a full course, no movement — is rarer, more useful, and the only version that should follow a patient through the chart. The full menu of options sits in the practice's guide to non-surgical options before spine surgery; this piece has been about how the central one is dosed, timed, and judged.
Frequently asked questions
How long should I give physical therapy before judging it?
Judge it at its checkpoint, not on a calendar impulse — and the checkpoint for a structured exercise programme is set in weeks to months, because that is the course the evidence measured. In this practice a structured trial is typically assessed over six to twelve weeks, with the home programme performed throughout — a practice standard for running the trial, not a number quoted from a single study. For chronic low back pain — the presentation exercise therapy's evidence addresses (Hayden et al., Cochrane, 2021) — a programme judged at two weeks has been judged before its first scheduled measurement. Acute pain is assessed differently, on the guideline's shorter windows, where early improvement is so often the natural course.
I went twice a week for two weeks. Isn't that a fair trial?
It is four to six sessions — the opening of a trial, not a trial. The interventions in the 249 trials behind the Cochrane review (Hayden et al., 2021) were programmes running weeks to months, and their outcomes were measured afterward. Dose matters as much as the calendar: the home programme between sessions constitutes most of the delivered dose, so two weeks of attendance with the home programme performed is a larger dose than two weeks without it — and still an early fraction of a course. A fair trial has three properties: the full course, the full dose including home practice, and a checkpoint read against baseline function. Yours has had time to acquire the first property only.
Therapy made me sore. Does that mean it is harming me?
Ordinarily, no — soreness is the expected texture of a dose being delivered. Across the trials in the Cochrane review, the adverse effect reported most often was increased muscle soreness; new loading of deconditioned muscle and connective tissue produces soreness that peaks early and settles as the tissue adapts across sessions. It should trend down as the same load repeats, and your therapist should be adjusting load against it. What is not ordinary soreness: new or progressive leg weakness, pain newly radiating below the knee and intensifying session over session, numbness in the saddle region, or any change in bladder or bowel control. Those are reported to the office promptly — they change the assessment, not just the exercise list.
Show all questionsShow fewer questions7 more ▾
My therapy is mostly heat and a treatment table. Is that exercise therapy?
No — and the distinction decides what your chart should say. The guidelines describe an active intervention: NICE NG59 recommends an exercise programme tailored to you and considers manual therapy only as part of a package that includes exercise; the ACP recommendations are likewise built on active treatments (Qaseem et al., Annals of Internal Medicine, 2017). Passive modalities — heat, ultrasound, electrical stimulation, time on a table — can settle a flare, but delivered alone they are not the intervention the evidence tests. If your course has no progressive exercise programme at its center, the accurate chart entry is not "failed physical therapy." It is "exercise therapy not yet delivered," and the correction is a referral that specifies it.
The exercises look like ones I could find online. Why attend sessions at all?
Because the programme is the intervention, and a programme is more than a list of movements. It is selection — which exercises, for your diagnosis and the movements your examination showed to be limited or provocative. It is dosing — load, repetitions, and progression adjusted against your response, which is the variable a video cannot measure. It is correction, of the compensations that let the strong structures do the weak structures' work. And it is measurement — baseline and checkpoint function, recorded, which converts attendance into a trial whose result means something. The home programme then delivers most of the dose between sessions. The list without the programme is ingredients without the prescription.
What if I do the full programme and still don't improve?
Then you have a genuine result, and it is a useful one: a prescribed programme, delivered at full dose over its full course, with no functional movement at the checkpoint. That result closes the exercise question for your presentation and moves the plan — re-examination, imaging if it has not been done, and the next step in the conservative sequence or a surgical discussion. It is not a verdict on your effort. The Cochrane averages combine substantial responders and non-responders; completing the dose and not responding places you in a documented category of the evidence, not outside it. What your chart gains is a failure that was actually measured, which every later decision can rely on.
Does physical therapy work better than other conservative treatments?
Against doing nothing, usual care, or placebo — yes, measurably: the Cochrane review (Hayden et al., 2021) found average pain improvement of 15.2 points on a 0–100 scale, meeting the review's threshold for clinical importance, with a smaller functional gain of 6.8 points that did not meet its 10-point threshold. Against other conservative treatments, the advantage narrowed to 9.1 points for pain and 4.1 for function — neither clinically important. Exercise therapy outperforms inactivity and performs comparably to the reasonable alternatives. Among reasonable options, the deciding factors become fit: your diagnosis, your preferences, and — decisively, for an intervention whose dose is mostly self-delivered — which programme you will actually perform.
If it helps a little, should I just keep going indefinitely?
A trial and a maintenance programme are different arrangements, and confusing them serves neither. A trial has a checkpoint: at the end of the course, function is compared against baseline — continue in maintenance, progress the programme, or record non-response and change the plan. Maintenance is the home programme continued on your own; it does not require open-ended supervised attendance, and indefinite attendance without progression or re-measurement is not a trial anymore — it is a subscription. If supervised sessions are still producing measurable gains at the checkpoint, extending the course is a stated conclusion. If the gains have plateaued at a level you can live and work with, maintenance is the success state: the trial worked.
Will doing a full course of therapy delay surgery I actually need?
Not when the trial is run with its carve-outs intact. Progressive weakness, instability, and loss of bladder or bowel control never enter a waiting trial — they are assessed as surgical problems immediately, including mid-course. Inside the trial, the checkpoint system is the safeguard: function and the neurologic examination are re-measured on schedule, and deterioration found at a checkpoint ends the trial that day. The full course governs pain without progressive neurologic loss in a stable spine — the presentations where the evidence, including the sciatica timing trial (Peul et al., New England Journal of Medicine, 2007), shows waiting with surgery in reserve preserves the later outcome. Delay is a property of an unmonitored wait; a checkpointed trial is monitoring.
My pain score is the same, but I can walk further and sit longer. Is that progress?
Yes — by the trial's own measures, it is the characteristic pattern of early progress. Function and pain do not move on the same schedule: walking distance, sitting tolerance, and flare recovery typically improve first, while the pain number lags — one reason the checkpoint weighs the full set of measures, not the pain score alone. The Cochrane result has the same shape — pain and function as related but distinct outcomes (Hayden et al., 2021). Record both at the checkpoint. A trial read on the pain number alone, at two weeks, would call your course a failure; read on function at the proper checkpoint, it is a trial in progress, trending as responding trials trend.