A clinical synthesis based on the landmark review by Dupont (Clinics in Sports Medicine 1997), the diagnostic meta-analysis of Stubbings and Smith (The Knee 2014), the therapeutic meta-analysis of Schindler (KSSTA 2014) and the only randomised trial comparing resection with physiotherapy (Sauer, 2022): 55 references checked one by one on PubMed.

What is a synovial plica, and why is its presence not a disease?

An embryonic remnant, not an abnormality

In the embryo of about eight weeks, the space between the distal femoral epiphysis and the proximal tibial epiphysis is occupied by mesenchymal tissue that temporarily divides the future joint cavity into compartments. This tissue resorbs during development, but rarely completely. What remains of it are folds of the synovial membrane: the plicae. Dupont, in the review that is still the reference, puts it unambiguously: incomplete resorption leaves synovial folds in most knees.¹

That sentence deserves to be read slowly, because it contains the whole clinical point of the subject. A structure present in most knees cannot, by construction, explain a rare symptom. It has the status of the mucoid ligament or the partial discoid meniscus: an anatomical variation, and not a sign. As Tindel already recalled in the Orthopedic Clinics of North America, plica syndrome is « an uncommon entity, diagnosed far too often in the presence of another condition ».¹⁵ Thirty-four years later, the sentence has not aged.

The four folds, and the only one that really counts

The nomenclature is organised in relation to the patella: suprapatellar, mediopatellar, infrapatellar, lateral. They are not at all equal clinically.

The mediopatellar plica is the only one that counts in everyday practice. Its position explains this: it runs along the medial border of the patella and can, if it thickens, come to interpose itself between the medial patellar facet and the anterior aspect of the medial femoral condyle during flexion-extension.⁹ The other three have, as a rule, no clinical significance. The suprapatellar plica only becomes an issue in its complete, non-perforated form, which isolates the quadriceps pouch from the rest of the joint: a genuinely rare anomaly.¹² The infrapatellar plica, or ligamentum mucosum, is so ordinary that it defines a normal image: it forms the roof of the horizontal cleft of Hoffa's fat pad, found on 90 % of knee MRIs.¹⁰ As for the lateral plica, it is described in 1 to 3 % of people and is still the subject of isolated cases published one by one.¹

Why the prevalence figure varies fourfold

This is where the literature becomes treacherous. One reads in turn that the medial plica is present in 22 %, 38 %, 48 %, 80 % or 95 % of knees: five figures all published, and all correct. These figures do not contradict each other: they do not describe the same knees, nor with the same eyes.

Three readings force themselves on you in front of this graph. First, the two highest figures come from operated knees. Gurbuz reviewed 318 arthroscopy videos and found a mediopatellar plica in 95 % of cases;³ Nakayama analysed 3,889 arthroscoped knees and reports 79.9 %.² These populations are, by definition, those of patients whose knee was bad enough to justify an operation. Quoting « 80 % » as a general population prevalence is a common abuse of language, and it mechanically inflates the impression that the plica is everywhere, and therefore that it explains everything.

Second, the only clean measurement in an asymptomatic population is Paczesny's. His control group comprised 32 volunteers with no complaint, that is 60 knees: 37 with no plica echo at all, 16 with an echo meeting one criterion, and 7 meeting two.⁵ Twenty-three knees out of sixty, then, carried a plica visible on ultrasound without their owner complaining of it. And seven of them would have met the majority of the criteria used to make the diagnosis.

Third, the labels are not interchangeable. Blok finds 21.8 % in 912 arthroscopies, but he is counting hypertrophic plicae, not plicae as such.⁶ Zmerly, in his review, gives a range of 18 to 60 % without the definition retained always being explicit.⁸ Comparing these figures without reading their denominator produces a false controversy where there is only a blurred definition.

The Sakakibara classification, and what it really predicts

The reference morphological description is Sakakibara's, taken up by almost all the arthroscopic series. It distinguishes four forms, from the simple ridge to a genuine curtain. Nakayama gives the most solid distribution, on 3,889 knees.²

This table contains a lesson that is almost never drawn: the most frequent type is the least likely to explain anything. A type A, a simple ridge, is found in more than one knee in three and has no mechanical reason to interpose itself. It is types C and D that concentrate the symptomatic descriptions, and together they account for a little more than one knee in five among knees that have already been operated on.

Hayashi provides the most useful nuance for reading an MRI. In 342 painful knees, 47.7 % had a visible mediopatellar plica: 22.2 % type A, 20.2 % type B, 5.3 % type C. The association with cartilage damage was limited to the medial patella (aOR 2.12; 95 % CI 1.23-3.64 for all types combined, aOR 4.20 for type B), and no association was found with subchondral bone lesions, in any patellofemoral subregion.⁴ A plica visible on the MRI of a painful knee is therefore, in the great majority of cases, a flatmate, not a culprit.

What histology shows when the plica really is diseased

If the normal plica is silent, what does a pathological plica look like? The most recent answer comes from a German histopathological study of 45 plicae taken from patients operated on for a documented plica syndrome. The result is clear: fibrosis in most of the specimens, chronic inflammatory reactions clearly dominating acute ones, hypertrophy of the synovial lining cells, lymphocytic aggregates present, bleeding and fatty change rare.⁷

This signature is consistent with the accepted pathophysiological model: a blunt injury, an abrupt return to activity or a transient synovitis inflames the fold, the inflammation progresses to fibrosis, the fibrosis makes the fold lose its elasticity, and the stiffened fold catches instead of gliding.⁹ It also gives the measure of the time involved: this is not an acute mechanism, it is a chronic story. Kan quantified it, in patients whose cartilage was already damaged, the delay between symptom onset and surgery was 29.0 months against 11.6 months for those whose cartilage was intact (p = 0.043).¹⁴

One reservation, and it is a large one: this histology describes plicae already selected by a surgeon who judged them pathological. Without a comparison with asymptomatic plicae taken under the same conditions, we do not know whether the fibrosis observed distinguishes the diseased from the healthy, or whether it is simply the ordinary state of an adult fold. The authors say so themselves in calling for a dedicated histopathological score to be built.⁷

Which clinical picture really raises the suspicion that the plica is responsible?

Four elements, and they must coexist

The reviews in this field converge on a description that has been remarkably stable for thirty years. The suggestive picture combines four elements, and it is their combination that counts, not their presence in isolation.

1. Anteromedial pain, dull, intermittent, sitting medial to the patella and above the joint line. This supra-joint-line site is discriminating: meniscal pain projects onto the joint line.¹ ⁴
2. Snapping, catching or clicking that is reproducible during flexion-extension, often felt by the patient themselves, sometimes audible. Pseudo-locking, a catching sensation without genuine loss of extension, is more characteristic than true locking.⁶
3. A painful palpable cord on the medial side of the knee, rolling under the finger against the anterior aspect of the medial femoral condyle. It is the only truly specific element of the examination.⁵ ⁶
4. Pain made worse by activities that repeat flexion-extension : stairs, squatting, running, cycling, and above all prolonged sitting followed by standing up.² ⁴

The palpable cord: the sign that carries the diagnosis

If a single element of the examination had to be kept, it would be this one. The principle is simple: the mediopatellar plica, when it is thickened and fibrous, becomes palpable as a cord stretched vertically along the medial border of the patella, between its superomedial edge and the fat pad. With the knee extended and the quadriceps relaxed, the thumb rolls it against the condyle. The sign has value only if it reproduces the patient's pain, and not merely if they feel a cord, since the cord also exists in the healthy person.

Griffith and LaPrade stress a detail that tends to be overlooked: these patients very often have stiff hamstrings, and this stiffness is part of the picture just as much as the pain.⁶ It also has a direct therapeutic consequence, which we will come back to.

The corollary is immediate, and it is the heart of the approach: the diagnosis is made by hand, not on a report. Al-Hadithy puts it in one sentence in his review: « Diagnosis is made by exclusion. »³

The triggering context, and why it counts

Plica syndrome almost never appears without a story. Two contexts dominate, and they are found throughout the literature.

Direct blunt trauma to the anteromedial aspect of the knee: a knock, a fall onto the flexed knee, a sporting contact. In Schindler's meta-analysis, which brings together 969 operated patients from 23 studies, trauma is considered the cause in 57 % of cases.¹ That is a high proportion, and it clearly distinguishes plica syndrome from a pure overload disorder.

An abrupt return to or increase in activity, the second classic mechanism: resuming running after a break, the start of a season, a change in training volume. Camanho, in a Brazilian series of 70 knees, attributes the hypertrophy of the fold directly to « unsuitable physical activity » and bases his conservative treatment on readjusting that activity.¹³

In both cases, the pathophysiological mechanism proposed is the same: a local, transient synovitis inflames the fold; the inflammation progresses to fibrosis; the fibrosed fold loses its elasticity and catches instead of gliding.¹ This sequence is consistent with the histology, which shows chronic inflammation dominating and fibrosis in the large majority.

This gap deserves to be named, without overstating it. The German surgeons describe a young woman in 77.7 % of cases; Schindler's meta-analysis, on 969 operated patients, reports a balanced sex ratio.¹ ⁸ The truth lies between the two : the 2025 systematic review, the largest and the most recent, counts 57 % women out of 527 patients.¹⁵ Hufeland's prospective series rises to 73.7 %, but on 38 knees in people under 30, which allows no generalisation.⁷ There is therefore a female predominance, more modest than the one the profile describes, and the profile circulates faster than the data that support it.

The forms that mislead: type D and the rare sites

Two configurations are worth knowing because they present in disguise.

The type D plica (fenestrated, duplicated or torn) gives, in slightly fewer than half of cases, symptoms that are not at all on the expected side. Matsusue reviewed 20 cases of symptomatic type D: 11 had symptoms in the medial compartment, but 9 showed patellofemoral or lateral compartment signs, mimicking a lateral meniscal tear or patellar subluxation, with no preoperative diagnosis possible.¹¹

The lateral plica is the other pitfall. Described in 1 to 3 % of people, it gives anterolateral pain that suggests anything but a plica. Kosaka gathered ten cases in athletes of 19 years' mean age, all having failed conservative treatment, with anterolateral pain, restricted range and catching.¹² The practical message fits in one line: in the athlete with persistent anterolateral pain, the lateral plica is at the bottom of the list, but it is on it.

Why must the more frequent causes of anterior pain be ruled out first?

The arithmetic, before the clinical work

Even before examining the knee, one thing is known: among patients who consult for anterior knee pain, the great majority have patellofemoral pain syndrome. Smith's meta-analysis, which pools 23 studies, gives an annual prevalence of 22.7 % in the general population and 28.9 % in adolescents; the point prevalence in adolescents is 7.2 % (95 % CI 6.3-8.3), and rises to 22.7 % (95 % CI 17.4-28.0) in adolescent female athletes.¹

Against that, there is no estimate at all of the frequency of plica syndrome in the population. Not one. The reviews call it an « uncommon entity »⁵ or a « rare » condition⁶ without ever pointing to a survey. The only order of magnitude available comes from an arthroscopic series: in Blok, 12.9 % of the 912 operated knees had no abnormality other than a plica and the associated chondromalacia: a proportion among knees already selected for surgery, which says nothing about the general population.¹⁵

This finding dictates the order of operations. Faced with anterior knee pain, the starting point is the patellofemoral approach, as codified by the clinical practice guideline of the Academy of Orthopaedic Physical Therapy published in the JOSPT in 2019.³ Plica syndrome is not an alternative to that approach: it is what is invoked when it has failed, and only if the picture from the previous chapter is complete. The dedicated article sets it out: patellofemoral pain syndrome, anterior knee pain.

The detail is worth developing, because it is almost never noted. In Boling's study, carried out on 1,525 students of the United States Naval Academy followed for two and a half years, patellofemoral pain syndrome is defined as retropatellar pain during at least two activities from a list, plus « negative findings on examination of the ligaments, menisci, bursae and synovial plica ».² In other words, the 22 per 1,000 person-years of incidence reported by that cohort describe, by construction, patients in whom the plica has been ruled out. The two diagnoses are not competing for the same ground: one is the first-rank diagnosis, the other the exception that must have been eliminated in order to make it.

The differential of anterior knee pain, entity by entity

Here are the entities to work through before settling on the plica, with, for each, what points to it and what separates it from the synovial fold.

The most difficult neighbour: the fat pad

One line of the table deserves developing, because it is the one where error is most likely. Hoffa's fat pad syndrome shares with the plica an anterior site, a mechanical character, a young and sporting population, and even the proposed mechanisms: inflammation then fibrosis of a richly innervated tissue.⁷

Two objective elements confirm this. First, the two entities share the same morphological sign: Kim showed that the patella-patellar tendon angle, which measures the sagittal tilt of the patella, is significantly smaller both in fat pad syndrome (137.3 ± 4.9 degrees) and in plica syndrome (138.1 ± 3.2) than in controls (141.4 ± 2.9), with no significant difference between the two syndromes.⁸ A later study of 405 MRIs confirms the absence of discrimination between plica, patellar chondromalacia and fat pad on this criterion.⁹

Second, the error is documented at arthroscopy. In the study validating the MPP test, of 13 false positives, seven corresponded to impingement of the synovial fringe of the fat pad, five to a localised synovitis and one to a focal cartilage lesion of the medial condyle.¹² The best-performing test in the field therefore errs, in the majority of cases, in exactly the direction the anatomy predicts: it confuses the fold with its fatty neighbour.

The practical consequence is modest but real: both entities are first managed in the same conservative way (relative unloading, motor control work, quadriceps management), which makes the distinction less critical in the first line than it would become if surgery were being considered.⁷

What getting the order wrong costs

Getting the order wrong is not neutral. Three consequences, in increasing order of seriousness.

You delay the treatment that works. Patellofemoral pain syndrome has codified management whose effectiveness is established: combined hip and knee exercise, education, load management.³ Attributing the pain to a plica means deferring that management in favour of a hypothesis whose frequency and optimal treatment are both unknown.

You steer towards a procedure whose benefit is not demonstrated in this indication. The general context is telling: in patients with a degenerative knee, the systematic review carried out for the BMJ Rapid Recommendations concludes, from 13 randomised trials, that arthroscopy brings a pain reduction of 5.4 points out of 100 at 3 months (95 % CI 2.0-8.8) and 3.1 points at 2 years (95 % CI -0.2 to 6.4), that is, in the authors' terms, no important long-term benefit.¹⁰ Plica syndrome is not osteoarthritis, and that conclusion does not transfer mechanically to it; it does, however, recall how easy it is to overestimate what an arthroscopic procedure settles.

You miss something else. That is the most concrete risk, and it is illustrated by a case published in the JOSPT : a 13-year-old basketball player referred for anterolateral pain, in whom the MRI reveals both an infrapatellar synovial plica and bilateral juvenile osteochondritis dissecans.¹³ The plica was indeed there. It was not the problem.

What are the clinical tests and imaging really worth?

The MPP test, described precisely

The reference test is the one described by Kim in 2007, referred to in the literature by the abbreviation MPP for medial patellar plica. Its procedure is simple, and the precision with which it is carried out matters:

1. Patient supine, knee extended, quadriceps relaxed.
2. With the thumb, apply pressure over the inferomedial part of the patellofemoral joint, so as to interpose the medial plica between the medial femoral condyle and the patella.
3. Maintain that pressure and flex the knee to 90 degrees.
4. The test is positive when the patient has pain with the knee extended, and that pain disappears or clearly decreases at 90 degrees of flexion.³

The logic is mechanical: in extension, the thumb traps the fold against the condyle; in flexion, the plica moves and slackens, the interposition ceases, the pain subsides. It is this abolition of the pain in flexion that gives the test its value, far more than the pain itself: a point often forgotten in practice, where people are content to check that medial palpation is painful.

Kim validated it on 172 knees, comparing 66 knees with anteromedial pain with 106 control knees with lateral joint line pain: sensitivity 89.5 %, specificity 88.7 %, positive predictive value 78.7 %, negative predictive value 94.4 %, accuracy 89.0 %.² The negative predictive value is the most useful figure in community practice: a negative MPP test makes the diagnosis very unlikely.

Why MRI disappoints, and what to ask of it

The meta-analysis of Stubbings and Smith gives MRI a sensitivity of 0.77 and a specificity of 0.58.¹ That last figure is the most important in the whole chapter, and it follows directly from chapter 1: MRI sees a plica very well, and a plica is very frequent in people who do not have the syndrome. A test cannot be specific for a disease if it detects a normal structure.

The exercise has nonetheless been attempted: since not every plica is pathological, can the pathological ones be told apart by a measurement? Kim set about it on 405 examinations, comparing 167 syndromes confirmed by MRI and arthroscopy with 226 unaffected knees. Thickness was the only significant factor on multivariate analysis: 2.3 ± 0.5 mm against 1.0 ± 0.8 mm, with an optimal threshold of 1.8 mm, but a sensitivity of 75.9 %, a specificity of 65.4 % and an area under the curve of 0.727.⁵ An AUC of 0.73 is poor discrimination: that does not make a diagnostic criterion, at most a supporting argument.

Blanke reaches a similar figure but from a different angle, and a far more useful one: in 117 patients treated conservatively for three months, the diameter of the plica was 0.8 ± 0.3 mm in those whose treatment succeeded against 1.6 ± 0.4 mm in those who had to be operated on (p < 0.05), with a significantly higher rate of plica-cartilage contact in the second group.⁸ Calibre is therefore not a diagnostic criterion; it is a prognostic factor for response to conservative treatment. That nuance changes everything one does with it.

A historical remark illuminates the division of roles. As early as 1996, Nakanishi compared MRI and arthroscopy in 40 knees: 27 of the 29 plicae seen at arthroscopy were correctly identified on MRI, and 9 of the 11 knees without a plica correctly classified negative.⁶ MRI therefore knows how to see the structure: the authors indeed concluded that it was useful as a « screening method before arthroscopy ». What no study has managed to make it do is distinguish the guilty fold from the innocent one.

There remains one good reason to request it: ruling out the other causes. That is in fact its only solid justification in this context: looking for a meniscal tear, an osteochondritis, a cartilage lesion, bone oedema. It is useful for what it eliminates, not for what it asserts.

Dynamic ultrasound, the genuinely good imaging for the subject

With equal performance on sensitivity (0.90) and clearly better specificity (0.83 against 0.58), dynamic ultrasound outclasses MRI for this precise indication.¹ The explanation lies in one word: dynamic. Paczesny defined three criteria observed during transverse mobilisation of the patella, with a 12 MHz linear probe: a continuous echo gliding over the medial femoral condyle during medial and lateral displacement of the patella; the entry of that echo under the patella during medial displacement; and reproduction of the pain or discomfort during the examination.⁴

The third criterion is the most important, and it explains the gain in specificity: dynamic ultrasound does not simply see the plica, it watches what it does and asks the patient whether it hurts. It is an instrumented clinical examination more than an imaging test. In 91 suspect knees, accuracy reached 88 %, sensitivity 90 %, specificity 83 %.⁴

The same work contains its own warning, already quoted: in the asymptomatic control group, 16 knees met one criterion and 7 met two.⁴ The rule of three criteria is therefore not decorative: it is what separates the signal from the noise.

The problem nobody solves: the reference is circular

A weakness that runs through the whole diagnostic literature of this field must be named. All these performances are measured against arthroscopy, that is to say against the observation that a plica is present and looks abnormal to the surgeon's eye. But that is exactly the question at issue: is the plica seen the cause of the pain? Validating a test against an examination that does not answer that question means measuring the agreement between two ways of looking at a fold, not the ability to identify a patient.

Stubbings and Smith say so explicitly, and it is the most useful sentence of their article: the reference standard for future research in this field should be the resolution of symptoms after resection, and not arthroscopic assessment.¹ Twelve years later, none of the studies gathered for this article has taken up that recommendation: all continue to validate themselves against arthroscopy. All the sensitivity and specificity values in this chapter must therefore be read as inter-method agreement, not as diagnostic performance in the strict sense.

Real practice, for its part, follows none of these recommendations: in the survey of 238 German knee surgeons, 58.0 % make the diagnosis on the combination of MRI plus clinical examination, and 54.2 % resect the plica as soon as they observe intraoperatively a significant effect on the patellofemoral joint.¹⁰ A decision taken during the operation, on a visual criterion, in a patient who is already asleep.

Which rehabilitation should be offered as a first line?

The real state of the evidence, and a contradiction to understand

Two statements circulate, and they seem incompatible. Schindler, in his 2014 meta-analysis, writes that symptomatic plicae can first be treated with physiotherapy measures and structured exercise programmes, « but that success rates are generally low ».⁸ Camanho, on the contrary, reports nearly 90 % good results after 60 days of rehabilitation in 62 knees treated without surgery.⁴

Both are true, and their contradiction is instructive. They do not describe the same population. Schindler pools 23 studies of operated patients: by construction, they are those in whom conservative care failed: the series contains nobody who recovered with rehabilitation. Camanho describes a cohort taken in the first line, all of them coming to consult, most cured before any surgical indication arose. Reading one in the light of the other means confusing the failure rate of the non-responders with the failure rate of everybody. This selection bias runs through the entire literature of the field, and it must be kept in mind at every figure in this chapter and the next.

The most recent and most useful data have just settled part of it. The systematic review with meta-analysis by Franco, published in 2025 in the Journal of Experimental Orthopaedics, brought together six clinical trials published between 2008 and 2023, that is 527 patients (57 % women). Over the 12 to 24-month follow-up period, the pooled functional improvement measured on the Lysholm score is 24.5 points (95 % CI 18.7-30.2; p < 0.001), and, this is the central result, with no statistically significant difference in favour of either approach (p > 0.05).¹ The authors conclude in favour of a conservative-first strategy.

Two reservations, which the authors set out themselves. The heterogeneity is substantial, I² = 85.3 % : the six studies do not measure quite the same thing in quite the same people. And the overall level of evidence is rated IV, four of the six studies being case series. This is not a demonstration of equivalence; it is the absence of a demonstration of superiority, which is not the same thing, but which is amply enough to justify starting with rehabilitation.

Modalities and level of evidence, modality by modality

There is no graded clinical practice guideline for plica syndrome. The table below therefore offers a GRADE reading built for this article from the study designs available, and not taken from an existing consensus: it is a reasoned appraisal, not an official rating.

The programme, and the reasoning behind it

For want of a trial comparing protocols with each other, the content of rehabilitation is deduced from the mechanism and from what the series describe. Four axes, in order of priority.

1. Reduce the loading that irritates the fold, without stopping. The mechanism is repeated irritation in loaded flexion-extension: it is therefore those movements that must be temporarily dosed, not activity in general. Camanho explicitly bases his protocol on « readjusting the sporting activity » and on « muscular rebalancing ».⁴ In practice: reduce the volume of stairs, deep squats and long out-of-the-saddle cycling sets; keep walking, cycling with a high saddle and low resistance, and swimming.

2. Strengthen the quadriceps, stretch the hamstrings. This is the combination most consistently described for twenty years. Griffith and LaPrade insist that these patients frequently have stiff hamstrings and that flexibility work must accompany the strengthening, not follow it.⁷ The logic is coherent: a stiff posterior chain increases the patellofemoral load in flexion, and therefore the pressure on the interposed fold.

3. Work on neuromuscular control of the lower limb. This is the best-argued axis from a causal point of view, and it comes from an unexpected source: a surgical team. In his ten-year follow-up of 52 operated knees, Paczesny observes that the results are significantly better in patients whose functional tests are normal: single-leg squat, modified Ober test, palpation of vastus medialis obliquus. His conclusion is explicit: « poor neuromuscular control may contribute to abnormal patellar tracking, leading both to irritation of the medial plica and to further cartilage degradation ».⁶ In other words, the fold may well be a consequence, and the control deficit the cause. Franco takes up the argument and draws the surgical consequence: patients whose patellar tracking is altered should not be treated by resection alone.¹

4. Reintroduce load progressively. The return to provoking activities is done in stages, guided by pain at 24 hours rather than by a calendar. That is the general principle of load management, which has not been tested specifically in this indication but which is the one in the patellofemoral recommendation of the JOSPT, applicable here by clinical proximity.⁹

A word on kinesiology taping, the only additional modality with a dedicated randomised trial. Genc and Duymaz randomised 80 patients between six weeks of exercise alone and six weeks of exercise plus taping: both groups improve, the taping group does better on pain, pain threshold, the stair climbing and descending test and quality of life, and it is the only one to improve the disability score significantly.² The trial is not blinded, which, for an intervention as visible as coloured tape, weighs heavily on the interpretation. The most solid result of that trial is in fact not the effect of the tape: it is that the control arm, six weeks of exercise, improved significantly on almost every outcome.

How long before concluding failure

The published intervals vary, and that should be said rather than settled arbitrarily. Camanho obtains his results at 60 days.⁴ Blanke sets the duration of conservative treatment at three months before considering surgery.⁵ Al-Hadithy mentions conservative treatment that can run to six months.¹⁰ Morelli and Braxton mention a widespread practice of six to eight weeks before considering an injection.¹¹ None of these thresholds has been compared with the others.

A compromise marker: three months of well-conducted, documented rehabilitation. That is the interval of the only study that has tried to characterise the non-responders, which gives it an empirical anchor the others do not have.⁵ Three months allow time for a complete strengthening cycle, and stay well short of the 29 months of mean delay associated, in Kan's series, with established cartilage lesions.

Blanke's data deserve a final comment, because they are reassuring and are rarely read that way. Of his 117 patients, 41 recovered with rehabilitation and 76 were operated on. At final follow-up, the two groups are level: Lysholm 96.25 against 95.93, return to sport 96.2 % against 97 %, Tegner score 6.0 against 6.01, with no statistical difference.⁵ Failing conservative treatment therefore does not condemn anyone to a poor result, but starting with it loses nothing either.

What is arthroscopic resection worth, and on which series?

What the two meta-analyses say

Two quantitative syntheses exist, and they converge. Schindler, in 2014, identified 23 studies reporting the clinical outcome of 969 patients after open or arthroscopic excision. Mean age 25, a balanced sex ratio, trauma implicated in 57 % of cases. At a mean follow-up of 27.5 months: 64 % of patients free of symptoms, 26 % improved, 10 % considered failures.¹

Gerrard and Charalambous, in 2018, retained 12 articles covering 643 knees, of which 7 articles and 235 knees were usable for the meta-analysis proper. The rate of good or excellent results is estimated at 84.2 % (95 % CI 72.8-91.4). In the patients who had received non-surgical treatment before the operation alone, it falls to 76.1 % (95 % CI 60.1-87).²

That gap between 84.2 % and 76.1 % is instructive: the best results are obtained in those who had not tried anything else. The authors draw the opposite conclusion to the expected one: they suggest that resection could be considered « either as a first line or after failure of non-surgical treatment ». That is one possible reading. Another, at least as plausible, is that the group not previously treated contains more patients who would have recovered on their own.

The only randomised trial, and what it does not prove

A single randomised controlled trial has compared arthroscopic resection with structured physiotherapy in isolated medial plica syndrome. Sauer included 48 patients, randomised between the two arms, with the change in the Lysholm score at two years as the primary outcome. The baseline scores were comparable: 65.8 and 66.3. At two years, the surgery group reached 89.7 against 74.6 for the physiotherapy group, a statistically significant difference (p = 0.007).³

It is the result most favourable to surgery in the whole literature. It calls for four reservations, none of which cancels it but whose accumulation counts.

• Thirty-three patients out of 48 were reassessed at two years, that is 31 % lost to follow-up. In a trial of that size, the withdrawal of fifteen participants can overturn a result depending on the direction in which they went.
• Blinding is impossible for this kind of comparison, and there is no sham surgery arm. Part of the gap may reflect the expected effect of an invasive procedure on a self-reported outcome.
• The content of the physiotherapy arm is not standardised across trials in the field, and it is not known whether the « structured physiotherapy » of that trial contained the neuromuscular control work that other authors consider central.
• The trial has not been replicated : four years on, the 2025 systematic review still finds only one randomised trial comparing the two approaches.

That is why the 2025 meta-analysis, which includes it alongside five other works, does not find a significant difference between the two approaches across the 527 patients as a whole.⁴ A single trial with 33 evaluable patients does not tip a pooled result.

The inclusion criteria, in figures

This is the most important point of the chapter, and it can be demonstrated. If the surgical series described the same plicae as the general arthroscopic population, the distribution of Sakakibara types would be comparable in them. It is not at all.

This graph says two opposite things, and both must be held. On the one hand, it is reassuring about the practice described : the surgeons in these series clearly do not operate on just any plica. They massively select the broad shelf forms covering the condyle, the ones whose interposition mechanism is plausible. On the other, it forbids generalising their results : when a meta-analysis announces 84 % good results, it is talking about a population in which type C is four to six times more represented than in the ordinary arthroscopic population. Those 84 % do not apply to the patient whose MRI mentions a « mediopatellar plica » with no further detail.

Franco sets the limit out in full: few studies manage to exclude the other knee conditions coexisting with the plica, which limits the ability to attribute the results observed specifically to plica syndrome.⁴ The authors go further, and the question they raise is worth quoting: since ICRS grade above 2 cartilage lesions weigh heavily on functional scores, might it not be the cartilage lesion itself that limits the result, and a procedure on the cartilage that would be more relevant than resecting the fold?

What really predicts a good result

Three factors emerge consistently, and they are more useful to the clinician than the overall rates.

The integrity of the cartilage. This is the best-established factor. Kan compared 57 knees according to the severity of the cartilage damage: postoperative results were significantly better in the group with little cartilage damage (p = 0.0017).⁹ Paczesny confirms it at ten years: lesions above ICRS 1 significantly lower the final Lysholm.⁵ Prejbeanu observes that patients with little cartilage damage recover fastest.⁶

The brevity of the history. Again in Kan, the delay between symptom onset and the operation was 11.6 months in the group with preserved cartilage against 29.0 months in the group with damaged cartilage (p = 0.043).⁹ Schindler cites the same good-prognosis factors: a young patient, localised and recent symptoms, no induced chondromalacia.¹

The absence of other pathology. This is the most often repeated and the least often verified condition. Hufeland deliberately excluded any patient with a cartilage lesion beyond ICRS grade I or requiring an associated procedure, and obtains excellent results: KOOS from 50.2 ± 19.1 to 80.7 ± 15.3, Kujala from 52.6 ± 16.6 to 80.7 ± 16.5, pain from 7.9 ± 2.0 to 3.1 ± 2.6 on a numerical scale.⁷ Notably, in that series neither trauma, nor sporting level, nor the morphological type of plica significantly influenced the result.

The ten-year follow-up, and what it changes

Almost all the series stop between one and three years. Only one has gone to ten, and its result deserves to be known: in 52 knees of 50 consecutive patients followed prospectively, the Lysholm score rises from 52 preoperatively to 80 at final follow-up.⁵ Compare that with Prejbeanu's 94.5 at three years⁶ and Sauer's 89.7 at two years³: the benefit erodes.

That erosion is not surprising, and the explanation proposed by the authors is the one that closes the loop of the whole article. They observe a negative correlation between the final score and age, a clear adverse effect of cartilage lesions, and better results in patients whose functional tests are normal, hence their conclusion: poor neuromuscular control contributes to abnormal patellar tracking, which in turn sustains both irritation of the plica and cartilage degradation.⁵ Resecting the fold without correcting what irritates it amounts to treating the symptom of a problem that has been left in place.

One last point, on safety. The 2025 meta-analysis reports no complication in either arm of the six included studies.⁴ The German survey gives a more realistic picture of practice: haemarthrosis and persistent pain are the most reported complications, and 83.2 % of surgeons estimate their complication rate at below 11 %.¹² Franco moreover notes that short-term complications are a « notable gap » in the literature.⁴ A procedure with low morbidity is not a procedure without consequence, especially when its comparative benefit has not been demonstrated.

What difference does the patient's age make, from the adolescent to the older adult?

The starting fact: the plica itself does not change with age

It is a result easy to miss in Nakayama's series, and it governs this whole chapter. In 3,889 knees, the distribution of plica types « was found in all age groups and in both sexes », with the single exception of women aged 20 to 39.¹¹ The medial plica is as frequent at 60 as at 20.

What varies with age is therefore not the structure: it is the probability that it explains the pain. At 15, the competition is called osteochondritis dissecans, apophysitis, patellar instability. At 55, it is called patellofemoral osteoarthritis and degenerative meniscal tear. In both cases, the same MRI image does not have at all the same value.

The sporting adolescent: the population in which the diagnosis is most often made

This is the classic ground. Schindler's meta-analysis gives a mean age of 25 for 969 operated patients;³ Hufeland's prospective series, which includes only people under 30, has a mean of 16.2 ± 4.7 years, with patients from 9 to 26.² The German survey confirms that surgeons place their typical patient between 21 and 30 in 57.6 % of cases.⁹

Irha and Vrdoljak devoted an article to this population whose title says the essential: « a diagnostic pitfall in adolescent athletes ».¹ The pitfall is twofold, and it works both ways.

In one direction, it is missed. The picture is modest, standard imaging is normal, and anteromedial pain in a sporting adolescent is easily filed under « growing pains » or « patellar syndrome ». The diagnosis is frequently delayed. Hufeland indeed titles his article « an underestimated pathology in young patients » and reports, in 35 operated patients, preoperative pain of 7.9 out of 10, that is not a minor nuisance.²

In the other direction, it is invoked in place of what matters. That is the major risk, and it is documented by a case published in the JOSPT : a 13-year-old basketball player, pain in the anterolateral aspect of the left knee for four weeks, worse with basketball, referred by direct access to a physiotherapist. The MRI requested after a medical opinion finds an infrapatellar synovial plica, and bilateral juvenile osteochondritis dissecans.⁴ In an adolescent, osteochondritis is looked for before everything else.

Two recent cases illustrate the coexistence, which is the rule rather than the exception. Vasiliadis reports a 14-year-old handball player, her knee locked at 20 degrees of flexion after an injury at training, in whom the MRI finds a hypertrophic medial plica and a horizontal tear of the medial meniscus; arthroscopy treated both.⁵ Sherafatvaziri describes a 12-year-old boy with a symptomatic hypertrophic lateral plica two years after a direct injury: a case that is rare in its age, in its site, and in the fact that he was not an elite athlete.⁶

The practical lesson is simple: in the adolescent, finding a plica does not excuse you from looking for the rest. It is almost always there.

The older adult: when the hypothesis becomes unlikely

At the other end of the spectrum, the reasoning reverses. The older the patient, the less defensible the plica hypothesis, for three cumulative reasons.

The series contain practically none of them. The mean age of 969 operated patients is 25.³ Hufeland excludes people over 30. None of the series gathered for this article reports results specific to patients over 50.

The result of surgery deteriorates with age. Paczesny showed, in his ten-year follow-up, a significant negative correlation between the final Lysholm score and the patient's age.⁷ The older the patient at the time of the procedure, the less the procedure returns.

The diagnostic competition becomes overwhelming. After 50, anterior knee pain is above all a matter of patellofemoral osteoarthritis or of a degenerative meniscal tear. And it is precisely in that context that the benefit of arthroscopy is best refuted: the systematic review of the BMJ Rapid Recommendations, on 13 randomised trials, concludes that there is no important long-term benefit in pain or function.¹³

A published case illuminates that limit better than any reasoning. Kambhampati describes a 70-year-old woman with a previously undescribed plica variant, resected at the same time as a loose body: the relief lasted eleven months, after which she underwent total knee replacement for the underlying osteoarthritis.¹² The plica was indeed there, and indeed atypical. It was not the problem.

A particular population: the already operated knee

One situation deserves to be known by the physiotherapist who follows postoperative patients. Yin reviewed 1,085 second-look arthroscopies carried out after anterior cruciate ligament reconstruction: a medial plica was present in 722 patients, that is 66.5 %, an incidence the authors describe as significantly higher than the one usually reported. All these plicae had avascular, fibrosed and thickened edges.¹⁴ The incidence varied significantly with the delay between injury and reconstruction and with the slowness of progress in postoperative rehabilitation, but not with the stability of the knee.

That result does not establish causality: these are knees reselected for a second arthroscopy, and therefore a particular sample. It does nonetheless justify thinking of it in persistent anteromedial pain with catching after ACL reconstruction, particularly when regaining flexion has been laborious. Postoperative ACL rehabilitation has its own pitfalls, and this one rarely appears on the list.

What do published clinical cases teach us?

The one that was hiding something else: Benedetti, 2019

A basketball player of 13 years presents by direct access to a physiotherapist for pain in the anterolateral aspect of the left knee that appeared four weeks earlier, worse after matches. After examination, the physiotherapist seeks the opinion of the GP, who requests an MRI. It finds an infrapatellar synovial plica, and bilateral juvenile osteochondritis dissecans.¹

What this case teaches. The plica found on imaging was real and could, on its own, have explained the picture plausibly. The lesion that carried the prognosis was the other one. It is the case to keep in mind whenever a report mentions a plica in an adolescent: the right question is not « can the plica explain this? », it is « what else is there? ».

The one that had two: Vasiliadis, 2024

A handball player of 14 years locks her left knee at 20 degrees of flexion during training. Tenderness sits on the medial joint line, the radiographs are normal. The MRI shows a hypertrophic medial parapatellar plica and a horizontal tear of the medial meniscus. Arthroscopy reveals an extremely hypertrophied plica, covering a large part of the medial femoral condyle and reaching up to the trochlea, with an atypical distal insertion into the intermeniscal ligament: an anatomical variant not described until then. The plica is excised, the meniscus repaired. At one month, the patient is asymptomatic; at three months, she has resumed her weekly training.²

What this case teaches. Two things. First, that coexistence is the rule: here, two lesions each capable of locking a knee. Second, that the anatomy of the plica is not fixed: an insertion into the intermeniscal ligament mechanically explains a locking that the usual arrangement would not.

The one whose image alarmed for nothing: Kim, 2021

A boy of 17 years, with no history of trauma or surgery, consults for recurrent pain in the right knee with snapping in active extension, preventing him from extending the knee fully, while a third party can extend it passively without difficulty. The MPP test is positive. The preoperative MRI shows a hypertrophic medial plica and an abnormal notch in the articular surface of the medial femoral condyle, raising the fear of an established cartilage lesion. At arthroscopy, a surprise: the cartilage of that notch, sitting under the plica, is perfectly intact. The authors conclude that the notch was hollowed out gradually under the pressure of the fold, with no associated cartilage lesion.³

What this case teaches. A worrying image is not a lesion. This case is also the only one in the set to describe the MPP test precisely, the procedure set out in chapter 4.

The one where conservative care really did fail: Marín Fermín, 2022

A man of 29 years consults for anteromedial pain in the right knee of gradual onset, worse going downstairs and on prolonged sitting. Examination finds medial parapatellar tenderness, a palpable click in that area during extension, and stiff hamstrings. The MRI shows a duplicated medial plica, with hyperintensity of the medial portion of the fat pad. After conservative treatment failed, arthroscopy confirms two superimposed plicae as well as hypertrophy of the fat pad; both folds and the fatty tissue are resected until all contact with the trochlea disappears through the whole range. At four weeks, the symptoms have completely gone; no recurrence at one year.⁴

What this case teaches. Successes must be told too. This patient had the complete picture from chapter 2, including the stiff hamstrings described by Griffith, plus a rare anatomical variant, and he had a well-indicated procedure after documented failure. It is the sequence this article recommends, applied.

The one that surgery did not settle: deBettencourt, 2024

A committed runner is diagnosed with synovial plica syndrome. His pain resists, in turn, arthroscopic resection of the plica, physiotherapy, non-steroidal anti-inflammatory drugs and intra-articular corticosteroid injections. He finally receives three series of lumbar sympathetic blocks, with significant and lasting relief. The authors conclude that a sympathetic component might exist in this condition.⁵

What this case teaches. It is the necessary counterweight to the previous one, and it represents the 10 % of failures in Schindler's meta-analysis: one patient in ten, a proportion rarely quoted alongside the 64 % who became symptom-free. It also recalls that chronic pain that resists everything does not call for one more procedure, but for a re-examination of the model that explains it. The sympathetic hypothesis remains, at this stage, a hypothesis drawn from a single case.

The one that came from the most ordinary fold: Radu, 2015

A middle-aged woman, with no notable history of trauma, presents with anterior knee pain and limited extension, resistant to conservative measures. The preoperative MRI shows a significantly thickened infrapatellar plica, tethering Hoffa's fat pad in the anterior interval of the knee. Arthroscopic resection of the infrapatellar plica leads to complete resolution of the symptoms within six months.⁶

What this case teaches. It concerns the infrapatellar plica, the one chapter 1 says is almost always silent and defines a normal image on 90 % of MRIs. It can therefore, exceptionally, be responsible, through a mechanism different from the others, not by interposing itself but by tethering the fat pad. That is the limit to place on any rule: « almost always silent » is not « always silent ».

How do you apply this concretely in the clinic?

The sequence, in order

1. Rule out the red flags first of all. True irreducible locking, a major effusion of abrupt onset, fever or constant non-mechanical pain, an adolescent with recurrent effusions. These situations immediately fall outside the scope of this article.

2. Carry out the full patellofemoral approach, and see it through. This is the step most often skipped when a report mentions a plica. Combined hip and knee exercise, education, load management, over six to twelve weeks, according to the recommendation of the JOSPT.³ This step is not an administrative preliminary: it is the treatment of the most likely diagnosis.

3. If it fails, check that the plica picture really is complete. Four elements, and all of them are needed: anteromedial pain above the joint line, reproducible snapping or pseudo-locking, a palpable cord whose palpation reproduces the patient's pain, a triggering context of trauma or a return to activity. Do the MPP test respecting its procedure: the key is the disappearance of the pain at 90 degrees of flexion, not the pain in extension.⁶ A negative test makes the diagnosis very unlikely: negative predictive value 94.4 %.

4. Do not request an MRI to confirm a plica. Its specificity is 0.58.⁷ Requesting it to rule out a meniscal tear, an osteochondritis or a cartilage lesion is legitimate; requesting it to affirm the syndrome is not. If doubt persists and a trained operator is available, dynamic ultrasound is the appropriate examination, provided all three criteria are required and not just one.

5. Treat, for three months, documenting it. Load adaptation, quadriceps strengthening, hamstring flexibility, work on neuromuscular control and dynamic valgus. Document the content and the scores: that is what will make it possible, if it fails, to speak of failure rather than of insufficient treatment.

6. Refer if and only if the failure is documented. Two elements strengthen the indication: a thick plica: beyond 1.5 to 1.8 mm depending on the study, and plica-cartilage contact on imaging, both associated with non-response to conservative treatment.²

What is said to the patient, word for word

Faced with a report that mentions a plica, with no clinical picture: « What is written on your MRI is accurate, but it is not an abnormality. A plica is a small fold inside the knee that most people have had since birth. When it is looked for on ultrasound in people with no knee pain at all, it is found in more than one knee in three. Its presence therefore says nothing about the origin of your pain. What we are going to do is look for what really explains what you feel. »

Faced with a genuinely suggestive picture: « It is possible that this fold is irritated and comes up against the bone when you bend your knee. The good news is that in most cases this settles with well-conducted work over three months, and that the patients who do end up needing an operation finish at the same level as those who recovered without surgery. So there is nothing to lose by starting there. »

Faced with a request for a quick operation: « Surgery gives good results, between 64 and 84 % depending on the study. But those figures come from very selected patients, often after several months of failure, and the only study that followed patients for ten years shows that the benefit fades with time. The most recent synthesis, on 527 patients, does not show that surgery does better than rehabilitation. »

What is measured, and when to redirect

The scales to use are those of the studies cited, which makes it possible to place oneself: the Lysholm score (used by almost all the series, with a pooled gain of 24.5 points under treatment, surgical or not), a numerical pain scale, and for the athlete the Tegner score for activity level.¹ ⁸

Frequently asked questions

I have a plica on my MRI. Does it need operating on?

No, not on that basis. The plica is a normal structure, present in a large part of the population with no symptoms at all: in the only published asymptomatic control group, 23 knees out of 60 carried one visible on ultrasound. A plica seen on imaging is an argument only if the clinical picture is complete, and even then, first-line treatment is rehabilitation.

Can plica syndrome recover without surgery?

Yes, and that is the most frequent outcome when it is taken in the first line. A series of 62 knees treated without surgery reports nearly 90 % good results after 60 days of rehabilitation.⁴ The most recent meta-analysis, on 527 patients, shows no superiority of surgery over conservative treatment at 12-24 months.¹

How long does it take to get better?

The published intervals run from 60 days to six months depending on the protocol. A reasonable marker is three months of well-conducted rehabilitation before concluding failure, that is the interval of the only study that has tried to characterise the non-responders.² None of these thresholds has been compared with the others in a trial.

Can I keep doing sport?

Generally yes, with adaptation. It is loaded, repeated flexion-extension movements that irritate the fold: stairs, deep squats, long out-of-the-saddle cycling sets. Walking, cycling with a high saddle and low resistance and swimming are usually well tolerated. Stopping altogether is neither necessary nor desirable.

My knee clicks. Is that serious?

An isolated, painless click is not a disease and is not treated. It is the combination of snapping and anteromedial pain and a painful cord on palpation that raises the suspicion of plica syndrome. A click alone justifies neither imaging nor treatment.

Why does my doctor talk about patellar syndrome and my MRI about a plica?

Because they are not answering the same question. The MRI describes what it sees, and a plica is visible in many people. The doctor is looking for what explains the pain, and anterior knee pain is, in the great majority of cases, patellofemoral pain syndrome, whose annual prevalence reaches 22.7 % in the general population.¹³ Both reports can be accurate at the same time.

Is ultrasound really better than MRI?

For this precise question, yes: equivalent sensitivity (0.90) and clearly higher specificity (0.83 against 0.58).⁷ The reason is that it is dynamic: it watches the plica move and asks the patient whether that reproduces their pain. MRI, on the other hand, remains superior for ruling out the other causes, which is often the real reason for the examination.

After a resection, can the pain come back?

Yes. Schindler's meta-analysis counts 10 % failures, and the only published ten-year follow-up shows a Lysholm score of 80 out of 100 at final follow-up, against 89 to 95 in the series at two or three years.⁸ ⁹ The factors associated with a poorer result are older age, a pre-existing cartilage lesion and deficient neuromuscular control, and that last point is precisely what rehabilitation can correct.

Going further on the anterior knee

Plica syndrome only makes sense within a broader approach to anterior knee pain. The compulsory starting point is patellofemoral pain syndrome, which accounts for the great majority of cases and whose management is far better established. For the differential, degenerative meniscal tears, patellar tendinopathy, pes anserine bursitis and prepatellar bursitis cover the essentials of what has to be ruled out.