What Works
There is no physical therapy treatment for this condition — it is a cancer, and it is managed by a specialist team. What follows grades the clinician and patient actions the evidence actually supports, which is the honest equivalent.
Tier 1 — Strong Evidence
StrongAn inappropriately placed biopsy contaminates the tissue around it and constrains the surgery that follows.
Unplanned removals accounted for 18.2% of all first operations for this cancer nationally — a quarter of them performed outside a hospital. The tumours removed this way were more often 5 cm or smaller and superficial, and imaging beforehand was more often skipped.
Tier 2 — Moderate Evidence
Moderate95.3% sensitivity, 100% specificity for identifying an ordinary fatty lump, and the only assessment domain in which any single sign predicted a non-benign result.
Over 4 cm performs better if you want a number — 89% versus 76% sensitivity in the cohort that tested both.
Doing so added a median 91 days (95% CI 76–106) to the diagnostic journey. This does not mean never image — it means imaging belongs inside the fast pathway, not as a detour ahead of it.
Tier 3 — Emerging / weak evidence
Three patients arrived at physical therapy with confident referral diagnoses — spinal stenosis, heel bursitis, post-surgical weakness — and all three had tumours. The trigger in every case was atypical findings or failure to respond.
What Doesn't Work
- Using pain to decide. 27% sensitivity — the worst of the four standard criteria, with a published recommendation to remove it from the urgent referral form. It persists because "painless lump = sinister" is memorable, teachable, and wrong in both directions.
- Using small size as reassurance. Half the cancers in the largest relevant series were under 5 cm — and the ones removed unknowingly are precisely the small, superficial ones.
- Relying on the fast-track cancer pathway as a safety net. One designated clinic supplied just 13% of its centre's sarcomas, and 51% of what it did find was already over 10 cm.
- Relying on familiarity. A 2026 survey found awareness modest across every specialty and training level, concluding that exposure alone is insufficient. Most family doctors see one case in a career.
- Not a failure mode: there is no evidence that massage, needling or loading spreads a sarcoma, and this page does not claim it. The measured harm is delay and unplanned surgery.
Red Flags
Most lumps are harmless — that is the honest starting point, and it stays true all the way down this page. This list exists so that the rare one gets looked at, not so that you worry about the common one.
- A lump that is getting bigger. The single most important sign there is.
- A lump that sits deep — anchored under the muscle covering rather than sliding about just under the skin. Any size.
- A lump bigger than about 5 cm (roughly the short side of a credit card). But do not read the reverse into this — small does not mean safe.
- A "bruise" or "bleed" that has not gone away, or a lump with no injury that properly explains it.
- The problem you are being treated for has not responded at all to treatment that should have helped.
- Something about the examination does not fit the diagnosis you were given.
- A "cyst" or "ganglion" that feels solid, firm, or fixed in place.
- Children and teenagers — a separate specialist pathway applies.
- Anyone with neurofibromatosis type 1.
- Tissue that has previously had radiotherapy.
- Lumps in the foot, ankle or lower leg — escalate rather than wait.
Return to Training
Return-to-training criteria for the condition itself do not exist — it is managed by an oncology service. What can be made binary is the clearance pathway that gets someone with an unexplained lump back to unrestricted loading.
- Maximum width and depth measured and recorded with a date
- Imaging performed — ultrasound at minimum for a lump you can feel
- Result is a confident benign diagnosis — not "probably", not "likely"
- If not confidently benign: specialist opinion obtained before any biopsy or removal
- No documented increase in size between two dated measurements
- Safety-net understood — any future change in size gets re-imaged
- The original problem, if there was one, is responding as expected
Meanwhile, keep training everything else. There is no evidence that general activity is harmful here, and shutting someone's whole programme down is its own harm in a person who almost certainly has an ordinary fatty lump. Stop the hands-on work over the lump; leave the rest alone.
Find the lump and do two things right now. Measure it across its widest point and write the number down with today's date. Then press gently and work out how deep it sits — does it slide around just under the skin, or does it feel anchored underneath the muscle covering?
Those two facts are exactly what a clinician triages on, and without a first measurement nobody can prove later whether it has grown. If it sits deep, or it is bigger than about 5 cm, or you already know it has changed — book to have it scanned rather than watched.
Most lumps are harmless. The two rules people use to spot the dangerous ones are both wrong.
Think of it like a weed growing in the middle of a flower bed you tend every week. It is not hiding. It is standing in plain sight, in exactly the soil you expect flowers in — which is precisely why your eye slides straight past it.
The danger is not that it is invisible. It is that pulling it out casually, the way you would pull a weed, breaks the roots and scatters them — and then the proper job becomes far harder than it needed to be. That is why the instruction is scan it before anyone touches it, rather than watch it and see.
- Here's what's really happening: this cancer grows in muscle, fat and connective tissue — the exact same tissue behind almost every ache, strain and lump a therapist sees, which is why it blends in so completely.
- What most people get wrong: believing a painful lump is safe and a small lump is safe. Pain turned out to be the worst of the four warning signs in standard use, and in the biggest relevant study half the cancers found were under 5 cm.
- Start here: measure it, write the number down with the date, and get it scanned before you let anyone remove it.
Conviction
Strong on what to do. Weak on the numbers everybody quotes.
HIGH — that this cancer mimics common musculoskeletal problems; that unplanned local removal is a frequent, measurable harm; that specialist contact should precede biopsy.
MODERATE — that ultrasound is the right first test; that an enlarging or deep lump warrants referral.
LOW — that the 5 cm threshold reliably separates what needs escalating. It is contradicted at the small end.
EVIDENCE AGAINST — pain being a useful discriminator. That is absence of discrimination, not absence of evidence.
NO EVIDENCE — how any of these warning signs perform in a first-contact physical therapy clinic. Every study was done in patients already referred somewhere else.
What would change my mind about the 5 cm threshold
A prospective study of at least 20,000 first-contact musculoskeletal visits, scoring the four standard criteria against 24-month cancer-registry linkage, and reporting how many referrals it takes to find one cancer rather than reporting sensitivity. Sensitivity measured in an already-referred cohort is the statistic that exists, and it is the one that has misled the field.
What would change my mind about "delay costs the limb"
The largest analysis available — 8,648 patients — found time from diagnosis to treatment related to survival non-linearly, with the lowest risk at 42 days, which the authors attribute to the time taken to reach a higher-volume specialist centre. A randomised comparison of direct-access ultrasound against usual referral, measured on stage at detection rather than referral yield, would settle whether speed or routing is the active ingredient.
Don't want to guess which lumps matter and which don't? The Verdict breaks down one piece of health evidence a week — what it says, what it doesn't, and what to actually do.
Join The Verdict — freeThe Full Picture — Anatomy, Diagnosis & Evidence
What's Actually Going On
Sarcomas arise from mesenchymal tissue — muscle, fat, fibrous tissue, blood vessels and nerve sheath. That is the same tissue behind almost every complaint in a musculoskeletal caseload, and it is the whole reason the mimicry works so well.
There is no separate compartment for tumours. A mass inside a muscle belly produces a firm swelling, a local ache and lost range — which is the presentation of a resolving bruise or a chronic strain. Sitting against a nerve, it produces a nerve-pattern symptom and gets worked up as an entrapment. In the back of the knee it reads as a Baker's cyst. In the forefoot it reads as a Morton's neuroma.
And the base rate governs everything downstream. Soft tissue sarcoma is under 1% of all cancers, and most family doctors see one case in a working lifetime. Meanwhile 83.5% of low-risk lumps that get removed turn out to be ordinary lipomas, and another 9.1% are a close relative. That is why a warning sign with decent sensitivity can still be nearly useless in practice — and it was measured, not theorised: 147 of 461 lumps carrying at least one "high-risk" feature were all benign.
How to Identify It
There is no special test for this. The assessment is measurement and imaging. What follows is how the criteria and tools actually in use perform — which is the honest equivalent of a test table.
- Ultrasound, for identifying an ordinary fatty lump Sn 95.3% Sp 100% — the highest-value single investigation on this page
- Size over 5 cm Sn 76% — measured in patients already referred
- Size over 4 cm Sn 89% — same cohort, the authors' recommended revision
- Pain Sn 27% — worst of the four; recommended for removal from the referral form
- Depth on ultrasound — the only individually significant predictor of a non-benign result found (p = 0.044)
In a nine-year double-blind study of fatty lesions, not one clinical sign and not one MRI sign reached statistical significance. Only ultrasound depth did. The authors' own words: the work "challenges the traditional perception of red-flags."
A comparison this page deliberately refuses to make: ultrasound's 95.3% is sensitivity for identifying a specific benign thing, in a different population against a different reference standard. Placing it beside pain's 27% would produce a dramatic and completely invalid conclusion. Two numbers that look identical on the page are not always the same measurement.
The Debate
The standard teaching: pain is one of the four criteria for urgent referral. It is on the forms and in every "lumps and bumps" module.
What the data say: pain had 27% sensitivity — the worst of the four tested — and the authors explicitly recommend removing it from the urgent referral form. The best-performing combination actually contains "no pain".
But here is the trap. A 376-patient series found rest pain and swelling were the two commonest presenting complaints. Both findings are true: pain is common in the cancer and overwhelmingly common in what it mimics, so it carries no information either way. Drop it from the form. Do not drop it from your suspicion.
The standard teaching: 5 cm is the threshold that identifies which lumps need escalating.
What the data say: half the cancers in a 552-patient series were under 5 cm, and the single cancer among 461 "low-risk" lumps was under 5 cm too.
The threshold was calibrated on patients already referred to sarcoma units, where tumours are large by the time they arrive — then applied to everyday community lumps, where they are not. Use it to escalate. Never use it to reassure.
The standard teaching: the fast-track cancer pathway is how these get caught early.
What the data say: one designated clinic hit its 10% yield target — while only 15% of the sarcomas it found were under 5 cm, 51% were over 10 cm, and it supplied just 13% of the sarcomas its centre managed. The paper's own title asks whether it is "meeting the target but failing the task."
The pathway is judged on yield — a process target — rather than on stage at detection, which is the outcome that matters. A rule optimised to keep referral volumes manageable will select for the obvious, large tumours.
Honest Limitations
Every study here was done on somebody else's patients
Those sensitivities — 76%, 89%, 27% — were all measured in people already sent to a sarcoma unit or a fast-track clinic. Applied to an unselected first-contact caseload, where the odds of cancer are far lower, the same criteria behave completely differently. No study anywhere measures how these warning signs perform in a physical therapy clinic. This is a population gap, not just an evidence gap: the studies are decent, and they are about a different group of people.
The rules are written for a lump, and many patients don't bring one
Every referral criterion keys off a lump you can measure. The three documented physical therapy cases arrived as spinal stenosis, heel bursitis and post-surgical weakness — and were a pelvic osteosarcoma, a knee chondroma and a thigh liposarcoma. A checklist that starts "measure the lump" cannot fire on a patient who hasn't presented with one.
Rarity makes experience an unreliable teacher
Most family doctors see one case in a career. A 2026 practitioner survey found awareness modest across every specialty and training level and concluded that exposure alone is insufficient. Meanwhile 30–60% of sarcoma patients report three or more visits before referral, against 23.4% for cancers generally — the system keeps seeing these people and keeps not acting. Confidence in spotting this is not calibrated by caseload, because the caseload contains almost none.
The Nuance
The intuitive model is linear: the tumour grows, delay does proportional harm, faster referral saves limbs. The largest analysis available does not support the linear model.
In 8,648 patients with localised high-grade disease, time from diagnosis to treatment related to survival non-linearly, with the lowest risk at 42 days. The authors' own reading: that interval reflects the time taken to be referred into a higher-volume specialist centre — a marker of better care, not a benefit of waiting.
Read carelessly, that sentence says "delay is good." It does not. It is time from diagnosis to treatment, in people who are already diagnosed, and 42 days is not a target. The harmful interval is the one before anybody suspects anything: a median 166 days from first symptom to diagnosis when local imaging happened before referral, and a mean 7 months of symptoms before referral in the foot and ankle series.
So the honest version is this: getting to the right place matters more than getting anywhere fast. Which is also why the single most consequential thing anyone can do here isn't a treatment decision at all — it's making sure the first operation is a planned one.
Sources
- Smolle, Leithner & Grimer 2015, Ann R Coll Surg Engl — N=135, 33% malignant. Size >5 cm 76% sensitivity, pain 27% (worst of four), >4 cm 89%. Recommends removing pain from the urgent referral form.
- Tan et al. 2025, BMC Cancer — N=552. 147/461 low-risk lesions carried a high-risk feature with zero malignancies; 1 of 2 malignancies under 5 cm; ultrasound 95.3%/100% for lipoma.
- Melis et al. 2022, Eur J Surg Oncol — N=2,187 national registry. Unplanned excision 18.2% of all first operations; those tumours more often ≤5 cm and superficial.
- Khan et al. 2020, Sci Rep — nine-year double-blind cohort. No clinical or MRI sign significant; only ultrasound depth (p=0.044).
- Szucs et al. 2016, Sarcoma — N=397. Met the 10% yield target; only 15% of sarcomas under 5 cm, 51% over 10 cm; supplied 13% of the centre's sarcomas.
- Featherall et al. 2019, J Surg Oncol — N=8,648. Non-linear time-to-treatment relation, lowest risk at 42 days. Not a target.
- Dyrop et al. 2017, Acta Orthop — N=545. Pre-referral local imaging added 91 days (95% CI 76–106) to the diagnostic journey.
- Noebauer-Huhmann et al. 2024, Eur Radiol — ESSR consensus, 46 radiologists, 12 countries, 95.9% agreement. Ultrasound as initial triage; contact the specialist centre before biopsy, without exception.
- Peterson, Denninger & Porter 2017, JOSPT — 3 cases, level 5. All referred with confident musculoskeletal diagnoses; trigger was atypical findings or failure to respond.
- Jenkins et al. 2025, J Foot Ankle Surg — N=376. Rest pain and swelling the commonest presenting complaints; mean 7 months to referral.
- van Kouswijk et al. 2023, BMC Musculoskelet Disord — 7 studies, N=214. No exercise programme after limb-preserving surgery was therapeutically valid; very low certainty.
- Fernández et al. 2025, Cancers — review. Existing referral guidelines are mostly clinical and "their results have been disappointing."