Stem Cells vs. Surgery: Understanding Your Options

"You Need Surgery": The Words No One Wants to Hear

You have been managing joint pain for months, maybe years. You have tried physical therapy, anti-inflammatory medication, perhaps a cortisone injection. Finally you see an orthopaedic surgeon, hoping for answers, and after reviewing your X-rays they deliver a verdict:

"You're bone on bone. You need a knee replacement."

For many people that moment feels like a door closing. It is worth knowing two things before you walk through it — and they pull in opposite directions.

The first is that joint replacement is genuinely one of the most effective operations modern medicine performs, and the evidence for it is far stronger than the evidence for anything offered as an alternative.

The second is that "you need surgery" is a clinical judgement about timing as much as it is a diagnosis, and for a substantial minority of patients that timing turns out to be negotiable.

Both are true. This page tries to hold both.

The Surgical Reality: What the Evidence Actually Shows

How well does joint replacement work?

The honest answer is: better than the alternatives that have been tested against it.

In 2015, Skou and colleagues published the first randomised controlled trial of total knee replacement in the New England Journal of Medicine. One hundred patients with moderate-to-severe knee osteoarthritis, all of them eligible for surgery, were randomised either to knee replacement followed by 12 weeks of non-surgical treatment, or to the same 12 weeks of non-surgical treatment alone — supervised exercise, education, dietary advice, insoles and pain medication.

At 12 months, the surgical group improved by 32.5 points on the KOOS4 score against 16.0 for non-surgical treatment alone: an adjusted mean difference of 15.8 points (95% CI 10.0 to 21.5) ^[1]. At two years the gap had widened slightly, to 18.3 points (95% CI 11.3 to 25.3) ^[2].

That is a large, durable, unambiguous benefit, and it is the single strongest efficacy result anywhere in this article — including everything in the regenerative section below. Hip replacement has an even longer track record; it has been described in The Lancet as "the operation of the century" ^[44].

What it costs to get that benefit

The same trial recorded the price. The knee-replacement group had 24 serious adverse events against 6 in the non-surgical group (P = 0.005) ^[1]. That is the most reliable head-to-head safety figure available for joint replacement, because it comes from randomisation rather than from comparing separate populations.

Beyond the trial setting, large registry and cohort data fill in the rest:

Infection. Linking the Australian national joint registry to hospital admission data across 2002–2017, 90-day periprosthetic joint infection rates were 1.3% (1.1–1.7) after knee replacement and 1.1% (0.8–1.3) after hip replacement using diagnostic coding, rising to 1.9% and 1.5% under a broader definition. Revision-for-infection alone captured only 0.1–0.3%, which means registry figures substantially understate the true rate ^[8].

Persistent pain. A systematic review of prospective studies in unselected patients found an unfavourable long-term pain outcome in roughly 7–23% after hip and 10–34% after knee replacement; in the best-quality studies, 9% or more after hip and about 20% after knee ^[9]. Roughly one knee-replacement patient in five still has meaningful pain in that joint.

Satisfaction. Satisfaction after knee replacement is usually reported as high, but a systematic review found the estimates vary widely and most of the instruments used to measure it have poor content validity — the number depends heavily on what question was asked ^[10].

How long implants last

This is where patients are most often given numbers that nobody can source. The best available answers come from two 2019 Lancet meta-analyses that pooled national registry data specifically to answer the question.

For knees, pooled 25-year survival across 14 registries and 299,291 replacements was 82.3% (95% CI 81.3–83.2). Unicompartmental replacements did worse, at 69.8% ^[11].

For hips, pooled 25-year survival across 215,676 replacements was 57.9% (95% CI 57.1–58.7). The authors' own summary: patients and surgeons can expect a hip replacement to last 25 years in around 58% of cases ^[12].

Those figures are averages across all ages. Age at surgery changes them dramatically.

The younger patient question

A population cohort of 63,158 hip and 54,276 knee replacements followed for up to 20 years found that lifetime risk of revision was about 5% for patients operated on after age 70 — but rose to 35% (95% CI 30.9–39.1) for men having surgery in their early fifties, with women in the same age band at roughly 15% lower risk. For patients operated on before 60, the median time to revision was 4.4 years ^[13].

This is the most important single statistic on this page for anyone under 60, and it is a genuine argument for not rushing. It is not, by itself, an argument for any particular alternative.

Does earlier surgery help everyone?

Not equally. A cost-effectiveness analysis using nine years of Osteoarthritis Initiative data (4,498 participants) modelled knee replacement as it is actually practised in the United States. Averaged across the whole population, the quality-of-life gains were small — SF-12 physical component summary 1.70 (0.26 to 3.57), WOMAC −10.69 (−13.39 to −8.01), KOOS quality-of-life subscale 9.16 (6.35 to 12.49) — and the improvements grew larger the worse a patient's baseline function was. Current practice, extended to patients with relatively preserved function, had minimal effect on quality-adjusted life years at the group level ^[14].

Read carefully, this is not an argument against knee replacement. It is an argument that the benefit is concentrated in people who are genuinely badly affected — which is the same thing your surgeon is trying to judge.

When Surgery Is Clearly the Right Choice

Surgery is the appropriate answer for:

1. Severe, end-stage disease with substantial functional loss and failure of conservative care
2. Structural deformity or instability requiring correction
3. Fracture and other trauma requiring surgical repair
4. Tumour or infection
5. Pain that is no longer controlled by any non-operative means

The evidence above supports rather than undercuts this. If your function is poor, the modelled benefit of replacement is at its largest ^[14], and the trial evidence for it is Level 1 ^[1].

The Regenerative Option: What Is Actually Known

Mesenchymal stromal cell (MSC) therapy is an injection-based, joint-preserving treatment. It is investigational for osteoarthritis in most jurisdictions, and its evidence base is real but much younger and much thinner than the surgical one. Here is what it shows.

Mechanism

MSCs are not believed to work by turning into cartilage. Current understanding is that they act largely through paracrine signalling — secreting anti-inflammatory cytokines and growth factors that modulate the joint environment ^[15][16]— and through extracellular vesicles, though the vesicle evidence for cartilage repair remains preclinical: a systematic review identified 13 controlled animal studies totalling 434 animals, and no human trials ^[17].

They are also immune evasive rather than immune privileged — a distinction that matters. Allogeneic MSCs do carry immunomodulatory features that allow them to be given without donor matching ^[18], but they are cleared by the recipient, and their evasiveness is conditional rather than absolute ^[19]. Safety data also does not automatically transfer between tissue sources: bone-marrow-derived safety findings cannot be assumed to apply to adipose or perinatal cells ^[20].

Efficacy in knee osteoarthritis

This is where the evidence is strongest, and it is genuinely positive — within limits worth stating precisely.

Against placebo. The largest randomised trial is a Phase III, multicentre, double-blind study of 261 patients with Kellgren-Lawrence grade 3 knee osteoarthritis, given a single injection of autologous culture-expanded adipose-derived MSCs or placebo. At six months the treated group improved significantly more on a 100-mm visual analogue pain scale (25.2 vs 15.5, P = 0.004) and on total WOMAC (21.7 vs 14.3, P = 0.002), with significantly more patients exceeding the minimal clinically important difference. No serious treatment-related adverse events occurred ^[4].

Against hyaluronic acid. A meta-analysis of 10 randomised trials and 818 patients with KL grade I–III found MSC injection superior to hyaluronic acid at 12 months, with a WOMAC total mean difference of −10.22 (95% CI −14.86 to −5.59), reaching the minimal clinically important difference for function at both 6 and 12 months, and with comparable safety ^[5].

Against controls generally. A 2025 meta-analysis restricted to randomised trials in patients who had not had surgery — 8 trials, 502 patients — found significant WOMAC improvement at 6 months (MD 7.44, 95% CI 1.45–13.42) and 12 months (MD 10.31, 95% CI 0.96–19.67), with VAS and KOOS also improved and no significant excess of adverse events ^[6].

Those are three independent Level-1 syntheses pointing the same way. On symptoms, at 6–12 months, in mild-to-moderate disease, the effect is real.

Where the evidence stops

Structure. The Phase III trial found no significant difference in MRI cartilage defect change between treated and placebo groups at six months ^[4]. A 2026 systematic review concluded that MSC therapy for knee osteoarthritis is symptom-modifying rather than structure-modifying ^[7]. Earlier reviews reported encouraging imaging findings across a heterogeneous, mostly low-quality body of studies ^[16]— the newer and better-controlled the trial, the weaker the structural signal has been.

Durability. Reported five-year outcomes diverge. A 60-month follow-up found scores drifting back toward baseline by year five ^[21], while a separate five-year series reported maintained benefit — in 11 patients, uncontrolled ^[22]. A ten-year safety follow-up of 72 patients recorded no tumours and no deep infection ^[23]. The honest summary is that durability beyond two years is not established in controlled data.

Not every joint or tissue responds. In a randomised trial of MSC injection for partial-thickness supraspinatus (rotator cuff) tears, 24 patients received cells in fibrin glue, fibrin glue with saline, or saline alone. The primary outcome did not differ between groups (P = 0.35), and the authors concluded the injection was not more effective than control injections ^[24]. For meniscal tears, a first-in-human study implanted MSC-seeded collagen scaffolds in five patients: three were asymptomatic at 24 months with no MRI evidence of recurrent tear, and two required subsequent meniscectomy at around 15 months ^[25].

Adverse effects are not zero. Joint-specific reactions are the common signal: a 2026 review found increased post-injection joint pain (RR 2.04) and swelling (RR 3.39) relative to controls ^[7]. A 2025 meta-analysis of adipose-derived and stromal vascular fraction injections found stromal vascular fraction inferior to corticosteroid at three months, adipose MSCs not superior to hyaluronic acid at any timepoint, and complications arising from the liposuction harvest — which led those authors to conclude that these treatments cannot yet replace existing minimally invasive options ^[26].

Context effects are large. Placebo and contextual response in knee osteoarthritis trials is substantial — on the order of 60% of the observed treatment response at six months and around 50% at twelve ^[27]. Any uncontrolled series, including any clinic's own results, is measuring that as well as the treatment.

Product quality is not a given. Unregulated preparations have caused real harm: a 2021 investigation traced 20 infections to a single umbilical-cord product, with 65% of tested vials contaminated ^[28]. Where cells come from, and under what manufacturing standard, is a legitimate question to ask any provider.

More cells is not better

This is worth stating plainly, because it runs against intuition and against a great deal of marketing.

In the only dose-escalation trial of umbilical cord-derived MSCs in knee osteoarthritis, patients received 2 million, 20 million or 80 million cells. All doses were safe — but 100% of the high-dose group experienced injection-related knee swelling, and clinical improvement was greater in the low- and medium-dose groups than in the high-dose group ^[29].

That is not an isolated finding. A dose-ranging randomised trial of bone-marrow MSCs found 25 million the most effective of the doses tested, with adverse events predominant in the higher-dose groups ^[30]. A review of 914 registered MSC trials concluded that both lower and higher doses were less effective than an intermediate range ^[31]. A 2025 meta-analysis subgroup analysis pointed the other way, favouring higher doses ^[6]— but a subgroup comparison across heterogeneous trials is hypothesis-generating, while a within-trial dose escalation is a direct test.

Repeat dosing is a different question from a bigger single dose, and it has better support. In a randomised Phase I/II trial, patients received hyaluronic acid, a single 20-million umbilical-cord MSC dose, or 20 million repeated at six months. At 12 months the repeated-dose group had significantly lower pain than the hyaluronic acid group on WOMAC-A (1.1 ± 1.3 vs 4.3 ± 3.5, P = 0.04) and VAS (2.4 ± 2.1 vs 22.1 ± 9.8, P = 0.03), with no severe adverse events — in nine patients per arm ^[32].

Cell source

Umbilical cord tissue is a common allogeneic source: it requires no harvesting procedure from the patient, and neonatal-tissue MSCs differ from adult bone-marrow and adipose MSCs in proliferative capacity and secretory profile ^[33]. Direct clinical evidence for umbilical cord cells specifically is more limited than for adipose or bone marrow.

A meta-analysis of umbilical cord MSCs in knee osteoarthritis found improved function and reduced pain (WOMAC MD −25.85, 95% CI −41.50 to −10.20) — from just three randomised trials, which the authors flag as too few to be confident about ^[34]. A 2025 randomised, double-blind trial of 55 patients comparing umbilical cord MSCs, hyaluronic acid and placebo found no significant difference in adverse events and no serious treatment-related events; overall pain and quality-of-life improvements favoured the cells but did not reach statistical significance. Significant benefit appeared in the subgroup with accompanying synovitis (VAS P = 0.011, WOMAC P = 0.014), and MRI showed no significant structural change except in the synovium. The authors call for Phase III trials ^[35].

The Substitution Question

This is the question the page exists to answer, so it deserves a direct answer rather than an implication.

Has anyone compared cell therapy to surgery?

No. A 2026 umbrella analysis set out to do exactly this comparison for KL grade 2–3 knee osteoarthritis. It searched seven databases across 2016–2026, screened 1,211 records and assessed 242 papers in full text. Its headline finding was a profound evidence gap: zero studies directly compared MSC therapy to total knee arthroplasty in this population ^[3].

Everything else in this section is inference from separate bodies of evidence, and should be read that way.

Is there any signal on avoiding surgery?

There is one quantitative estimate, and it comes with a large caveat attached by its own authors. The same umbrella analysis pooled non-randomised comparisons and reported a reduced hazard of conversion to knee replacement (HR 0.52), with 36–60 month conversion rates of 12.4% for MSC-treated patients against 22.7% for controls ^[3].

That is a real signal and it should not be dismissed. It is also derived from studies where patients were not randomised — meaning the people who chose cell therapy may simply have had less severe disease, more resources or more motivation than the comparison groups. The authors' own conclusion is that comparative effectiveness and long-term durability relative to arthroplasty remain uncharacterised, and that randomised trials benchmarking the two are urgently required ^[3].

What is well evidenced about deferring surgery

Something is — but it is not cell therapy.

In the two-year outcome of the Skou trials, among patients who were eligible for knee replacement and randomised to 12 weeks of structured non-surgical treatment alone, 16 of 50 (32%) had undergone knee replacement within two years. The authors' summary: two out of three patients with moderate-to-severe knee osteoarthritis eligible for surgery delayed it for at least two years following non-surgical treatment ^[2]. At 12 months in the original trial the figure was 26% ^[1].

Structured non-surgical treatment also outperformed written advice alone by 7.0 KOOS4 points (95% CI 0.4–13.5) ^[2], which is why supervised exercise, education and weight management are the core recommended treatments for knee and hip osteoarthritis in both the OARSI guidelines ^[36]and the AAOS non-arthroplasty guideline ^[37]— and why cell therapy appears in neither as a recommended intervention.

If your goal is to postpone a replacement, the intervention with randomised evidence behind it is a supervised exercise, education and weight-management programme. That is an uncomfortable finding for a page like this one, and it is the finding.

Does delaying actually pay off?

Not automatically, and this cuts against the intuitive case. A Markov cost-effectiveness model of a 55-year-old with KL grade 2–3 knee osteoarthritis compared three platelet-rich plasma injections with a one-year delay to replacement against replacement from the outset. From a payer perspective, replacement from the outset dominated — more quality-adjusted life years (14.63 vs 14.55) at lower cost ($26,235 vs $26,619) ^[38].

That model is about platelet-rich plasma, not MSCs, and a one-year delay is short. But it demonstrates that "delay is obviously better" is a claim requiring evidence rather than a self-evident truth. The counterweight is the revision data: for a man in his early fifties, a 35% lifetime revision risk and a 4.4-year median time to first revision are substantial ^[13], and they argue for waiting where waiting is tolerable.

What about severe disease?

The European consensus on orthobiologics for knee osteoarthritis positions cell-based therapy as a second-line option in KL grade 1–3, with inferior results in KL grade 4, and notes that 22 of its 27 statements rested on grade C or D evidence ^[39]. Its companion part on blood-derived products was able to issue three grade-A recommendations for platelet-rich plasma in KL grade ≤3 ^[43]— a useful indication of how much further ahead the blood-derived evidence currently is. An earlier systematic review of stem cell injections for knee osteoarthritis concluded against recommending the treatment on the evidence then available ^[40].

If your joint is genuinely bone-on-bone throughout, cell therapy is where the evidence is weakest and replacement is where it is strongest. Nothing on this page should be read as encouraging you to defer an indicated operation on that basis.

Comparing the Two: What Can and Cannot Be Put Side by Side

The table below is not a head-to-head comparison, because no head-to-head trial exists ^[3]. Each column reports what has been measured for that treatment, against its own comparators, in its own populations. The two columns are not commensurable, and the differences between them should not be read as effect sizes.

The equivalent comparison for hip osteoarthritis cannot responsibly be drawn at all. The published human evidence for MSC therapy in hip osteoarthritis consists of small uncontrolled series — for instance ten patients given three weekly doses of autologous bone-marrow MSCs, with radiographic scores unchanged in nine of them ^[41]— with no randomised trial and no controlled comparison. A registry analysis of 840 knee procedures using bone marrow concentrate reported modest improvements (lower extremity functional scale +7.9 to +9.8 out of 80; pain 4.0 → 2.6) with adverse event rates of 6–8.9% ^[42], but that is knee data, and bone marrow concentrate is not culture-expanded cell therapy. Anyone offering hip injections should be asked directly what evidence they are relying on.

Who the Evidence Suggests Is a Reasonable Candidate

Candidacy is a clinical judgement your own physician has to make with your imaging in front of them. What the evidence supports, generally:

Where cell therapy has been studied and shown benefit

• Kellgren-Lawrence grade 2–3 knee osteoarthritis — the population in the Phase III trial ^[4]and the consensus-supported range ^[39]
• Symptomatic pain and function as the goal, over a 6–12 month horizon ^[5][6]
• Patients for whom surgery is not yet indicated, or who wish to attempt deferral with realistic expectations

Where the evidence is weak or absent

• KL grade 4 / true bone-on-bone — inferior results, and the population where replacement performs best ^[39]
• Hip osteoarthritis — no controlled human evidence ^[41]
• Rotator cuff tears — the randomised trial was negative ^[24]
• Any expectation of cartilage regrowth, or of avoiding a future replacement ^[4][7]

Where surgery is the better answer

• End-stage disease with poor function — where the modelled benefit of replacement is largest ^[14]
• Deformity, instability, infection, tumour or fracture
• Pain no longer controlled by any non-operative means

What Sterling-Certified Partner Clinics Offer

For transparency, and separately from the evidence above: Sterling-certified partner clinics in Thailand provide allogeneic umbilical cord-derived MSC preparations under Thai FDA oversight, alongside supporting infusion services. Cell counts, session structure, pricing and the inclusion of adjunct services are commercial and clinical arrangements determined at the clinic. They are not findings from the literature and are not presented here as evidence of effect.

Adjunct offerings — including NAD+ infusion and NK/NKT cell preparations — have no published clinical evidence supporting their use for joint disease, and form no part of the evidence base described on this page. In the joint literature, natural killer cells are studied principally as participants in osteoarthritis synovial inflammation rather than as a treatment for it.

Any specific treatment plan is a matter for a qualified physician who has examined you and reviewed your imaging.

Questions Worth Asking

Of your surgeon

1. What is my Kellgren-Lawrence grade, in each compartment?
2. What does the MRI show that the X-ray does not?
3. Given my age, what is my estimated lifetime revision risk?
4. What specifically happens if I wait one year? Two?
5. Which non-surgical options have I genuinely not yet tried, at full dose?

Of anyone offering cell therapy

1. What is the cell source, and to what manufacturing standard is it produced?
2. What dose, and what is the evidence for that dose?
3. What outcomes do you measure, over what period, and against what comparison?
4. What proportion of your patients go on to have joint replacement anyway?
5. What would make you tell me not to have this?

Of yourself

1. How much is my function actually limited today?
2. Am I hoping to feel better, or hoping to avoid an operation? Those are different goals with different evidence behind them
3. Am I willing to have spent the money and still need surgery?

Frequently Asked Questions

If I try stem cells and they don't work, can I still have surgery?

Yes. Intra-articular injection does not alter joint anatomy and does not remove the option of later replacement. Whether it improves later surgical outcomes is a separate claim, and there is no controlled evidence for it.

Will stem cell therapy let me avoid a knee replacement?

No study has established that. The one quantitative estimate of reduced conversion to surgery comes from non-randomised comparisons within an umbrella analysis whose principal finding was that no direct comparison exists ^[3]. The intervention with randomised evidence for deferring surgery is a structured programme of exercise, education, weight management and pain control ^[2].

Will it regrow my cartilage?

The Phase III trial found no significant MRI difference in cartilage defects against placebo at six months ^[4], and current systematic review evidence describes the effect as symptom-modifying rather than structure-modifying ^[7]. Improvements in pain and function are real; structural regeneration is not established.

How long does the benefit last?

Most randomised evidence runs to 6 or 12 months ^[4][5][6]. Beyond that the data are inconsistent — one 60-month follow-up found scores approaching baseline by year five ^[21], another five-year series reported maintained benefit in 11 uncontrolled patients ^[22].

Is more cells better?

The dose-escalation evidence says no. The high-dose group in the umbilical-cord dose-escalation trial had universal injection-related swelling and smaller improvements than the low and medium doses ^[29], and a dose-ranging bone-marrow trial found its lowest tested dose most effective, with adverse events concentrated at higher doses ^[30].

My surgeon says I need surgery now. Should I get a second opinion?

Unless there is an emergency — fracture, infection, tumour — most osteoarthritis decisions allow time to seek one, and a second opinion is reasonable before any elective operation. That is a general statement about elective surgery, not advice about your case.

Is stem cell therapy FDA-approved?

Not for osteoarthritis in the United States. It remains investigational for this indication in most jurisdictions. Sterling-certified partner clinics in Thailand operate under Thai FDA oversight; that is a different regulatory framework, not an equivalent approval.

Why does this page cite so much evidence against the treatment it describes?

Because that is what the literature contains, and a page that hid it would be less useful to you. The positive findings here are real and are reported at full strength. So are the limits.

Scope of This Review

This is a narrative synthesis, not a systematic review. It prioritises randomised trials, meta-analyses of randomised trials, national registry data and society guidance, and states where evidence is weak, contested or absent. Where a treatment's strongest evidence is older, the older work is cited; where newer evidence has changed the picture, that is noted. Effect sizes and confidence intervals are given as the original authors reported them.