Why So Many Treatments Look Better Than They Are

What regression to the mean actually is

Take any measurement that varies — blood pressure, mood, a golf score, the pain in your back. Measure it when it is unusually high, and the next measurement will probably be lower. Not because anything was done, but because unusually high was unusual. Extremes are, by definition, not where things normally sit, and things return to where they normally sit.

Regression to the mean is the tendency of an extreme measurement to be followed by one closer to average, with no cause involved beyond the fact that it was extreme.

Your pain over a year is not one number. It is a jagged line. Some weeks you barely notice it; some weeks you cannot put your socks on. Plotted, it wanders around an average with peaks above and troughs below — and the peaks are the parts you remember, because the peaks are the parts that hurt.

Now add the one detail that turns a statistical curiosity into a problem: nobody starts a treatment at a random point on that line.

Why this hits pain harder than almost anything else

Two things have to be true for regression to the mean to fool you: the measurement has to fluctuate, and you have to act at the extreme. Pain does both, harder than almost any other symptom. It swings week to week for reasons nobody can fully name, and nobody books an appointment on a good day. The selection is built into the behaviour.

Think about what prompts the call. Not the average Tuesday — the morning you could not lift your kid. By the time you have found a clinic, waited, and started treatment, you are days or weeks past the worst point, and the line was already heading back toward its average before anyone touched you.

The treatment did not choose you at random. You chose the treatment at your worst. Everything that follows has to be read through that.

None of which is a criticism of anyone. Seeking help when things are bad is the correct response to things being bad. It just means the moment of maximum motivation is also the moment of maximum statistical trap, and the two are inseparable.

The conditions that were going to improve anyway

Regression to the mean is only half the problem. The other half is natural history: a great many musculoskeletal conditions resolve on their own, on timetables long enough that people try several treatments while waiting. Whatever they happened to be doing when the curve turned gets the credit. The published timetables are the clearest way to see this.

• Plantar fasciitis. More than 90 percent of people improve within about ten months of simple nonsurgical treatment 1Ref 1American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Plantar Fasciitis and Bone Spurs.The prognosis claim that more than 90% of people with plantar fasciitis improve within about ten months of simple nonsurgical treatment, used here as a worked example of natural history long enough to absorb several treatment attempts..
• Frozen shoulder. It moves through freezing, frozen, and thawing stages and usually resolves over one to three years, with range-of-motion physical therapy as the primary treatment 2Ref 2American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Frozen Shoulder (Adhesive Capsulitis).That frozen shoulder progresses through freezing, frozen, and thawing stages and usually resolves over one to three years, with range-of-motion physical therapy as the primary treatment..
• A lumbar herniated disc. It can compress a nerve root and cause sciatica, but most people improve within weeks to months without surgery, and only a small percentage require microdiscectomy for sciatica 3Ref 3American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Herniated Disk in the Lower Back.That a lumbar herniated disk can compress a nerve root and cause sciatica, that most people improve within weeks to months without surgery, and that only a small percentage require microdiscectomy..

More than 90 percent of people with plantar fasciitis improve within about ten months of simple nonsurgical treatment 1Ref 1American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Plantar Fasciitis and Bone Spurs.The prognosis claim that more than 90% of people with plantar fasciitis improve within about ten months of simple nonsurgical treatment, used here as a worked example of natural history long enough to absorb several treatment attempts..

Read that timetable again and think about what fits inside it. Ten months is long enough to try a night splint, an insole, a stretching routine, a new pair of shoes, a supplement, and an injection. One of them will be the last thing you tried before it stopped hurting. It will feel like the answer. It is at least as likely to be whatever happened to be in your hand when month ten arrived.

The same arithmetic runs through low back pain, through sciatica natural history, and through disc reabsorption. Long recoveries are what fill a market with confident testimonials.

How anyone tells the difference: the control group

The only way to know whether a treatment did something is to compare it against what would have happened otherwise. That is the entire purpose of a control group — not bureaucracy, not caution, but the only available answer to the question "compared to what?" For surgery, the honest control is a sham: an operation that goes through every step except the one being tested.

That sounds extreme until you consider what it controls for. The consultation. The scan. The anticipation. The anaesthetic. The scar. The months of being someone who is being treated. All of it happens in both arms. Only the intervention differs.

In people aged 35 to 65 with a degenerative medial meniscus tear and no osteoarthritis, arthroscopic partial meniscectomy was no better than sham surgery for relieving symptoms 4Ref 4Sihvonen R, Paavola M, Malmivaara A, et al. (FIDELITY) (2013).Arthroscopic Partial Meniscectomy versus Sham Surgery for a Degenerative Meniscal Tear.That in patients aged 35-65 with a degenerative medial meniscus tear and no osteoarthritis, arthroscopic partial meniscectomy was no better than sham surgery for symptom relief — the worked example of what a sham control is for.. The knees that were actually operated on and the knees that were opened and closed reported the same relief.

Almost everything beats nothing. Far less beats a sham. The gap between those two sentences is where regression to the mean, natural history, and expectation have been living.

This is not an argument that nothing works

It would be an easy overcorrection to read all this as proof that treatment is theatre. It is not. The whole reason controlled comparisons matter is that some things survive them — and knowing which ones did is far more useful than either believing everything or believing nothing. Several treatments in this field have been compared against a placebo and come out ahead.

Exercise therapy probably reduces pain and improves function in chronic non-specific low back pain compared with no treatment, usual care, or placebo, with small-to-moderate effects 5Ref 5Hayden JA, Ellis J, Ogilvie R, Malmivaara A, van Tulder MW (2021).Exercise therapy for chronic low back pain.That exercise therapy probably reduces pain and improves function in chronic non-specific low back pain compared with no treatment, usual care, or placebo, with small-to-moderate effects — the counter-example of a treatment that survives a controlled comparison.. Small-to-moderate is an honest description, not a disappointing one; it is what a real effect looks like when it is measured properly rather than sold.

Plantar fasciitis makes the point most cleanly, because both facts about it are true at the same time. More than 90 percent of people improve within about ten months anyway 1Ref 1American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Plantar Fasciitis and Bone Spurs.The prognosis claim that more than 90% of people with plantar fasciitis improve within about ten months of simple nonsurgical treatment, used here as a worked example of natural history long enough to absorb several treatment attempts. — and there is strong evidence that manual therapy, stretching of the plantar fascia and calf, and foot orthoses reduce pain and improve function 6Ref 6Koc TA Jr, Bise CG, Neville C, et al. (2023).Heel Pain — Plantar Fasciitis: Revision 2023 (Clinical Practice Guidelines Linked to the ICF).That strong evidence supports manual therapy, plantar fascia and calf stretching, and foot orthoses to reduce pain and improve function in plantar heel pain — paired with the natural-history figure to show both can be true at once..

Both of those can be true without contradiction. Ten months is a long time to hurt. Shortening the road is worth something even when the destination was always going to arrive.

That is the useful frame. Regression to the mean does not say treatment is pointless. It says your own improvement is weak evidence, and the strong evidence has to come from a comparison somebody else already ran against a control.

How to read your own recovery

You cannot run a controlled trial on yourself — you only get one arm, and you are not blinded. But you can be considerably more careful about what you conclude. The main move is to notice when you started: if you began a treatment during the worst week you have had in a year, the deck was already stacked toward it looking good.

• Ask what the condition's own timetable is. If the expected course is improvement over months, then improvement over months is not news. The timetable is the control group you were never given.
• Count how many things changed at once. A new insole, a new routine, better sleep, a less brutal work month, and an injection all landing in the same fortnight means the fortnight has five candidate explanations and no way to separate them.
• Read testimonials for what they are. A testimonial is a peak followed by a trough, narrated by the person in the middle of it. Conditions that fluctuate most — tennis elbow, low back pain — reliably generate the most confident ones.
• Ask the clinician a sharper question. Not "does this work?" but "what has this been compared against, and what happened?" Sometimes the answer is "nothing rigorous, and we do not know" — which is legitimate, and better than a confident answer that is not true.

When waiting is the wrong answer

Everything above is about the difference between a treatment and time. None of it applies when time is the thing you do not have. Some presentations are not fluctuating pain waiting to regress toward a mean — they are a structural problem getting worse, and the improvement curve this page describes does not exist for them.

Regression to the mean explains why improvement is unreliable evidence. It says nothing about deterioration. A symptom that is steadily getting worse is not regressing toward anything.

The situations where the statistical argument stops applying are recognisable. Weakness that is deepening rather than fluctuating. Numbness spreading into the groin or inner thighs, or a change in bladder or bowel control. Pain with fever, night sweats, or unexplained weight loss. A joint that has mechanically locked rather than one that hurts to move. These are not peaks on a jagged line; they are a line heading in one direction.

The lumbar disc is the honest illustration. Most people improve within weeks to months without surgery, and only a small percentage require an operation 3Ref 3American Academy of Orthopaedic Surgeons (OrthoInfo) (2024).Herniated Disk in the Lower Back.That a lumbar herniated disk can compress a nerve root and cause sciatica, that most people improve within weeks to months without surgery, and that only a small percentage require microdiscectomy. — but a small percentage is not zero, and progressive weakness is exactly what moves someone into it. Knowing that improvement is unreliable evidence is a tool for reading the good news carefully. It was never meant as a reason to sit still while something fails.