Hamstring muscle strain (pulled hamstring)

Sudden pain at the back of the thigh in mid-acceleration, and the athlete stops dead. Hamstring muscle strain has become the leading injury in men's professional football, and one of the most recurrent in all of sport: nearly seven reinjuries out of ten occur in the two months following return to play.

A clinical synthesis article. Every figure carries its source where it is written. The 46 references were verified against two independent databases (Europe PMC and CrossRef) at the time of writing: identifier, journal, year, pagination and full author list. Where a piece of data comes from a population that is not the reader's, men's professional football most often, the limitation is flagged at the figure itself.

In brief: the clinical synthesis

What the available literature allows us to state, in ten points. Each is taken up, sourced and qualified in the chapters that follow.

• It is the commonest injury in men's professional football. In the UEFA study run over 21 seasons and 3,909 players, hamstring injuries went from 12 % of all injuries in 2001/02 to 24 % in 2021/22, and from 10 % to 20 % of all lay-off days¹.
• About one injury in five is a reinjury : 18 % of the 2,636 injuries in that cohort, and 69 % of those reinjuries occur within two months of return to play¹.
• Reinjury strikes in matches, not in training : it is 9.25 times more likely there (95 % CI 7.67 to 11.15)¹. Returning to team training and being available for competition are not the same decision.
• The leading risk factor is a previous injury. A meta-analysis of 78 studies, 8,319 injuries in 71,324 athletes, gives a relative risk of 2.7 for any previous injury and of 4.8 for a recent injury²³.
• Two mechanisms, two diseases. The fast-running injury affects the long head of biceps femoris and heals in a few weeks; the slow-stretch injury affects the semimembranosus and its proximal free tendon, and may take a year : in fifteen professional dancers, return to the previous level took a median of 50 weeks (30 to 76)¹².
• The site weighs more heavily than the grade. Across 76 injuries in professional footballers, involvement of the free tendon of the long head of biceps took 56 days as a median against 24 days for the central tendon of the same muscle³³.
• But intramuscular tendon involvement is not a sentence. Across 165 athletes followed for a year, it was associated with no excess of reinjury (HR 1.05; 95 % CI 0.52 to 2.12)⁸. The contradiction with the track and field series is real and is explained by the population, not by a measurement error.
• No imaging classification predicts the return date on its own. Across 176 athletes, the three systems tested explained only 7.6 % to 11.9 % of the variance in the return timeline³¹.
• No set of return criteria has been validated. A systematic review of 25 articles lists a great variety of them, none of which has been validated²⁵. Conversely, an earlier return is associated with reinjury at twelve months across 308 injuries²⁶.
• Eccentric exercise is the only documented prevention, and it is little applied. The founding trial finds a relative risk of 0.29 for injuries overall and of 0.14 for reinjuries, with a number needed to treat of 3 to prevent one reinjury¹⁷ ; ten years later, the complete programme was applied in only 10.7 % of elite club-seasons²⁰.

Strain or tendinopathy: what exactly are we talking about?

Three different entities share the name of the same muscle group. Confusing them means applying one's timetable to the other, and that is one of the surest ways of manufacturing a reinjury.

The term “hamstrings” denotes a group of three biarticular muscles: the long head of biceps femoris and the short head laterally, the semitendinosus and the semimembranosus medially. All but the short head arise from the ischial tuberosity and cross the hip and the knee. That dual action explains both their vulnerability in running, where they brake knee extension while the hip flexes, and the variety of clinical pictures they produce.

The site already publishes two articles on proximal tendinopathy. It is not the same disease as a strain, and the table below exists so that the distinction can be made in thirty seconds, at the first history.

What is the real frequency of hamstring strain?

The solid figures come almost entirely from a single source: UEFA's prospective surveillance of elite European clubs. That is a strength, because the data collection has been standardised over two decades. It is also a limitation, to be taken into account before transposing to the clinic patient.

The first reference measurement dates from 2011. Across 51 teams and 2,299 players followed from 2001 to 2009, muscle injuries accounted for 31 % of all injuries and 27 % of total lay-off time. Among them, the hamstrings came well ahead with 37 %, ahead of the adductors (23 %), the quadriceps (19 %) and the calf (13 %). Sixteen per cent of muscle injuries were already reinjuries, and those cost significantly more lay-off days than index episodes².

Twelve years later, the same team published the long series. Across 21 consecutive seasons, 54 teams from 20 countries and 2,636 hamstring injuries, the share of these injuries in all injuries went from 12 % to 24 %, and the share of lay-off days they cause from 10 % to 20 %. Over the last eight seasons analysed, the incidence rose by 6.7 % a year in training (95 % CI 1.7 % to 12.5 %) and by 3.9 % a year in matches (95 % CI 0.1 % to 7.9 %)¹.

Translated into organisational terms, that gives usable landmarks: in a squad of 25 players, expect eight hamstring injuries per season, and 20 % of players will miss at least one training session or match for that reason during the season¹.

The median lay-off is 13 days (interquartile range 7 to 22). That median masks two very different populations: so-called structural injuries, with a visible fibre lesion, cost 17 days (IQR 11 to 25), functional injuries 6 days (IQR 4 to 10)¹. That is the first reason why announcing “three weeks” without having examined the patient makes no sense.

Why does this injury recur so much?

That is the question that governs everything else. An injury that heals in thirteen days and recurs one time in five is not a healing problem: it is a decision problem.

Of the 2,636 injuries in the UEFA cohort, 475 were reinjuries, that is 18 %. The striking fact is not that rate, but its distribution in time: 325 of those reinjuries, that is 69 %, occurred within two months of return to full participation¹. An earlier systematic review reached the same conclusion in different words: more than half of reinjuries occur in the first month after return²⁵.

A second fact, just as clear: reinjury strikes in competition. Its incidence is 0.88 per 1,000 hours in matches against 0.10 in training, that is a relative risk of 9.25 (95 % CI 7.67 to 11.15). The ratio is identical for early reinjuries¹. In other words, a player can get through two weeks of team training without incident and be injured at the first real competitive exposure: training does not test what a match demands.

What really increases the risk

The best available synthesis is the 2020 update: 78 prospective studies, 8,319 injuries including 967 reinjuries, in 71,324 athletes. Four factors emerge with a statistically significant association²³ :

Two less expected lessons come out of that same meta-analysis. First, a previous anterior cruciate ligament rupture (RR 1.7) and a previous calf injury (RR 1.5) are risk factors in their own right: the history must not be confined to the thigh. Second, and this is the most useful sentence in the abstract, reinjury risk is better assessed from clinical data than from the MRI features of the index injury²³.

That last statement was refined in 2024 by a multicentre cohort of 330 acute injuries confirmed on MRI, from the pooling of four prospective studies. At twelve months, five elements were independently associated with reinjury: discomfort on the active knee extension test (adjusted relative risk 2.52; 95 % CI 1.10 to 5.78), a lower straight leg raise angle on the injured side, involvement of the myotendinous junction on MRI (aRR 3.10; 95 % CI 1.39 to 6.93), less extensive anteroposterior oedema, and a shorter time to return to play²⁶.

How does the injury happen, and why does the mechanism change everything?

Two very different movements produce two injuries that carry the same name, sit in different muscles and do not have the same prognosis. Telling the two apart in the history is worth more than many investigations.

The fast-running injury

By far the commonest. In the UEFA cohort, running or sprinting is the mechanism reported in 61 % of the 2,031 injuries for which the mechanism was recorded, and up to 62 % of structural injuries¹. It preferentially affects the long head of biceps femoris, far more often involved than the semimembranosus and semitendinosus¹.

A video analysis published in 2024 refined the real circumstances, reviewing broadcast footage of 63 match injuries in the Qatari league between 2013 and 2020. Running was involved in 86 % of cases, but the commonest situation was not the flying sprint: it was acceleration over the first ten metres, found in 24 % of all injuries. Pressing was involved in 46 % of cases, and in short accelerations, indirect contact between players and loss of balance were present in 53 % and 67 % of situations. The authors conclude explicitly that prevention research should look beyond high-speed running alone²⁸.

One last field finding, often quoted and for once solidly measured: nearly half the injuries occurring in matches happen in the last fifteen minutes of each half, which departs significantly from the expected distribution¹.

The stretch-type injury

Much rarer, and much longer. It accounts for only 5 % of the mechanisms reported in the UEFA cohort, but 10 % of semimembranosus and semitendinosus injuries¹. The movement involved is slow hip flexion with the knee extended: the splits, a high kick, a deep lunge, a slide to reach a ball.

The reference description covers fifteen professional dancers followed prospectively for two years after a first injury. All had been injured during a slow hip flexion movement with the knee extended, and all had initial symptoms that were relatively mild. Every injury sat in the proximal part of the posterior thigh, near the ischial tuberosity. The semimembranosus was involved in 87 % of cases, and every semimembranosus injury involved its proximal free tendon. The median time to return to the previous level was 50 weeks, with a range of 30 to 76 weeks, and no initial clinical or MRI parameter predicted it¹².

How do you examine a strain in the acute phase?

The positive diagnosis is rarely difficult. The stake of the first consultation lies elsewhere: ruling out what belongs to the surgeon, and setting the baseline measurements that will serve to decide on return several weeks later.

The measurements that really serve, and why to record them from day one

A study of 131 MRI-confirmed injuries followed daily throughout rehabilitation tested which clinical measures actually tracked progress. Three stand out: strength in outer range, range in maximal hip flexion with active knee extension, and the length of the painful area on palpation. Intra-observer reliability of the strength and flexibility measures was excellent (95 % CI 0.85 or above). The authors also note that progress in these measures is not linear, and that at the same perceived running effort, actual speed varies greatly from one athlete to another²⁹.

That last point has a direct consequence: asking “what percentage are you at?” is a poor measuring instrument. You have to time or measure, not ask.

Is an MRI needed, and when?

MRI specifies the site, documents the record and reassures about the absence of a surgical lesion. What it does not do is set the return date: across 176 athletes with MRI at five days, the classification systems tested explained only 7.6 % to 11.9 % of the variance in time to return to sport, with wide overlap between categories³¹. The next chapter details that limitation.

In practice, MRI is justified when the answer changes management: a competitive athlete with a deadline to arbitrate, suspected tendon involvement from the mechanism or from palpation, diagnostic doubt, or a picture that is not evolving as expected. It is read in the first five days, and the classification system used must be named in the report, failing which the grade communicated cannot be interpreted³¹.

Which classification should you use, and what does it really bring?

Several systems coexist, and they do not say the same thing. Choosing is not a matter of school: a grade communicated without the name of the system that produced it is empty information.

The historical three-grade classifications, “mild, moderate, complete”, were judged inadequate from the early 2010s, for lack of diagnostic precision and prognostic value³. Three systems have established themselves since, each with its own logic.

Which one is actually used

The question was settled by the London international Delphi procedure, published in 2023. A first questionnaire was sent to 46 international experts (sports physicians, physiotherapists, surgeons, strength coaches and scientists), followed by a consensus meeting of fifteen then a second round with 112 experts, with a response rate of 88.4 % and an agreement threshold set at 70 %. Result: 58 % of experts use BAMIC, 12 % Munich and 6 % the Barcelona classification⁶.

What these systems are worth, measured rather than asserted

Three pieces of work answer this without appealing to any school of thought.

On reproducibility, the answer is rather reassuring. Forty athletes, two independent radiologists, three systems compared: intra- and inter-observer agreement runs from “substantial” to “almost perfect” for severity grades and overall classifications (kappa 0.65 to 1.00 intra, 0.77 to 1.00 inter). By contrast, the reliability of the anatomical subcategories, the very ones that carry the prognostic information, remains uncertain, with kappas ranging from -0.05 to 1.00³⁰.

On prognostic value, the answer is frankly negative. The same team followed 176 athletes with MRI at five days until return to sport. Timelines varied enormously within each category and between categories; the total variance explained by BAMIC severity grades and anatomical sites ran from 7.6 % to 11.9 %. The authors' conclusion, verbatim: none of the classification systems could be used to predict return to sport in their sample³¹.

On the substance, finally, a 2015 narrative review recalls that despite the abundance of literature on the subject, the classification of muscle injuries rests predominantly on expert opinion, and that little links the clinical or radiological features retained to an established pathology or a clinical outcome³⁴. The London consensus says nothing different when it recommends developing these systems further and calls for work on their validity and reliability⁶.

Does the site of the injury change the prognosis?

This is the point where the literature openly contradicts itself. Rather than picking a side, you have to understand why both sets of results are true in their own population.

The idea is old and intuitive: an injury that extends into the intramuscular tendon, that tendinous extension running inside the muscle belly, ought to heal less well than one confined to the fibres. A 2023 narrative review recalls the scale of the phenomenon: intramuscular tendon involvement is present in up to 41 % of hamstring injuries, and reinjury rates of up to 60 % have been reported in elite track and field athletes¹¹.

The series that confirm it

The seminal study covers 230 elite British track and field athletes, 44 of whom sustained 65 hamstring injuries with MRI within seven days. Times to return to full training differed significantly by BAMIC category: grade 3 injuries and “c”, intratendinousinjuries took significantly longer. Above all, the reinjury rate was significantly higher for “c” injuries, whereas it differed neither by numerical grade, nor by muscle involved, nor by proximal, central or distal location, nor by age or sex⁴.

Two independent cohorts point the same way on timelines. In professional Australian football, 41 injuries classified by the severity of intramuscular tendon involvement and its height give median returns of 21 days for low-grade involvement, 35 days for distal or isolated high-grade involvement and 88 days for high-grade proximal involvement of the long head of biceps associated with semitendinosus involvement¹⁰. In Spanish professional football, across 76 injuries, involvement of the free tendon of the long head of biceps took a median 56 days against 24 days for involvement of the central tendon of the same muscle (p = 0.038)³³.

The series that contradict it

Two prospective studies from the same group tested the same hypothesis in mainly football populations, with a standardised, criteria-based rehabilitation protocol.

The first included 165 athletes, with MRI within five days and prospective collection of reinjuries over twelve months. Intramuscular tendon disruption was present in 64 of them (39 %). The overall reinjury rate was 19 %, and it was identical with and without tendon involvement: 20 % against 20 %, hazard ratio 1.05 (95 % CI 0.52 to 2.12; p = 0.898). The same absence of difference for a waviness of the tendon⁸.

The second, in 70 participants, measured the return timeline. Injuries with no disruption returned in a mean 22.2 days (standard deviation 7.4); those with full-thickness disruption in 31.6 days (standard deviation 10.9), a significant difference. But the authors themselves stress that the gap represents a little over a week and that the overlap between the groups is such that the relevance for a given athlete remains limited. They recall in passing that this injury's reputation, “more than 50 days lay-off and up to 63 % reinjury”, rested on retrospective series at high risk of bias⁹.

A recent element strengthens the interest in the site, but shifts the target. In the multicentre cohort of 330 injuries already cited, it was not intramuscular tendon involvement that emerged as an independent factor for reinjury at twelve months, but involvement of the myotendinous junction, with an adjusted relative risk of 3.10 (95 % CI 1.39 to 6.93)²⁶.

What rehabilitation, and at what level of evidence?

Three randomised trials structure this whole chapter. They are old, small, and none is blinded: that is the real state of the evidence on a condition that nevertheless costs a fifth of the lay-off days in professional football.

What is no longer done: isolated stretching

The 2004 trial remains, twenty years on, the most striking in the field. Twenty-four athletes with an acute injury were randomised between a programme of static stretching and isolated hamstring strengthening, and a programme of progressive agility and trunk stabilisation. Time to return to sport did not differ significantly (37.4 days against 22.2; p = 0.25). The reinjury rate, on the other hand, differed massively: in the first two weeks, 6 athletes out of 11 (54.5 %) against 0 out of 13 ; at one year, 7 out of 10 (70 %) against 1 out of 13 (7.7 %)¹⁶.

That result must be read for what it is: an enormous effect in a tiny sample, in a single trial never replicated identically. It does not prove that trunk stabilisation heals; it shows that a programme centred on passive stretching and analytical strengthening of the injured muscle is a bad idea, and that negative conclusion has held up well.

What is done: loading in lengthened positions

Two Swedish randomised trials compared a protocol emphasising lengthening exercises (protocol L) with a protocol of conventional exercises (protocol C), in athletes with MRI-confirmed injury.

• In elite football, 75 players: return in a mean 28 days (standard deviation 15, range 8 to 58) with protocol L against 51 days (standard deviation 21, range 12 to 94) with protocol C. A single reinjury over twelve months, in group C¹³.
• In track and field, 56 sprinters and jumpers: 49 days (standard deviation 26) against 86 days (standard deviation 34). Two reinjuries, both in group C. And, whatever the protocol, injuries involving the proximal free tendon took far longer: 73 days against 31 in group L, 116 against 63 in group C¹⁴.

These two trials are not blinded, come from the same team and a single country, and the differences observed are considerable for so simple an intervention. They nevertheless remain the best direct evidence available on the content of rehabilitation, and their message is consistent with the rest of the field: loading the muscle in a lengthened position is the lever.

What has been tested and adds nothing: bringing eccentric work forward

A natural question is whether lengthening exercises should be introduced as early as possible. A randomised superiority trial answered it: 90 male athletes with MRI-confirmed injury, randomised between introduction from the first day of rehabilitation and introduction after being able to run at 70 % of maximal speed, all otherwise receiving the same programme. Return to sport: 23 days (IQR 16 to 35) against 33 days (IQR 23 to 40), but an adjusted hazard ratio of 0.95 (95 % CI 0.56 to 1.60; p = 0.84), therefore with no significant difference. No difference either in reinjuries at 2, 6 and 12 months²⁷.

That trial is instructive for another reason: the raw medians appear to separate the two groups clearly, and the adjusted analysis shows that they do not. It is a good reminder of what differences in means are worth when read without a confidence interval.

What remains uncertain: platelet-rich plasma injections

The reference trial is Dutch: multicentre, randomised, double-blind against placebo, in 80 competitive and recreational athletes with an acute hamstring injury. The primary outcome, published in the New England Journal of Medicine³⁹, showed no speeding up of return to play. The secondary results published the following year found no difference in reinjury rate at one year (HR 0.89; 95 % CI 0.38 to 2.13; p = 0.80), nor in any other subjective, clinical or MRI outcome³⁸.

Later syntheses do not overturn that result. A 2021 meta-analysis covering 10 studies, 207 injections against 149 controls, finds a reduction in return time of 5.67 days and a reinjury hazard ratio of 0.88, both non-significant³⁷. A 2026 meta-analysis restricted to the six randomised trials (277 patients) finds a gain of 8.6 days (21.4 against 30.0 days), but with heterogeneity of 94,1 %, an unchanged reinjury rate (15 % against 16 %; p = 0.72), and certainty of evidence rated by its authors on GRADE as low to moderate because of risk of bias and imprecision³⁶.

When can sport be resumed, and on what criteria?

This is the decision that determines the one-year prognosis. It rests on criteria none of which is validated, and it is taken under pressure that has nothing medical about it.

Why a fixed timeline is a bad criterion

Three reasons, each measured.

One, the spread of actual timelines is enormous. The 13-day median of the UEFA cohort has an interquartile range of 7 to 22 days¹, and the cohorts that break it down by site run from 21 to 88 days¹⁰. A mean timeline applied to an individual is wrong in both directions, half the time too long, half the time dangerous.

Two, imaging does not make up for that spread. MRI classifications explain at best a tenth of the variance in the timeline³¹.

Three, and this is the decisive point, an earlier return is independently associated with reinjury. Across 308 injuries followed for twelve months, time to return to play emerges among the five factors independently associated with reinjury risk, alongside discomfort on the active knee extension test and involvement of the myotendinous junction²⁶. The very title of that publication mentions the value of delaying return to play.

What is validated, and what is custom

This must be clear, because many protocols circulate with a confidence the literature does not support. A systematic review examined 25 articles providing a definition of return to play or criteria for deciding it: 13 gave a definition, 23 described criteria. The recurring themes were absence of pain, strength and flexibility comparable with the sound side, agreement of the medical team and functional performance. The authors' conclusion: a great variety of criteria is used, none of which has been validated²⁵.

The 2023 London consensus adds the most useful complement by separating what the experts agree on from what they do not. The statements reaching the 70 % threshold concerned the choice and dose of exercises (78.8 % to 96.3 % agreement), the importance of the kinetic chain (95 %), criteria for progressing exercises (73 % to 92.7 %), running and sprinting (83 % to 100 %) and return to sport criteria (78.3 % to 98.3 %). By contrast, the numerical thresholds for flexibility (40 % agreement) and strength (66.1 %) did not reach consensus, any more than adjuncts did (68.9 %)⁷.

Does eccentric exercise really prevent reinjury?

It is the only widely documented preventive measure in the field. Its history is also a lesson in critical reading: depending on the method of analysis chosen, the same literature gives “half as many injuries” or “no conclusion possible”.

What the trials show

The founding trial is Danish: 50 clubs, 942 professional and amateur players, cluster randomisation, ten weeks of a progressive eccentric programme then weekly maintenance. Over a full season, 52 injuries in the control group against 15 in the intervention group. The rates per 100 player-seasons give¹⁷ :

• All injuries : 3.8 against 13.1, that is an adjusted rate ratio of 0,293 (95 % CI 0.150 to 0.572; p < 0.001). Number needed to treat: 13.
• New injuries : 3.1 against 8.1, ratio 0.410 (95 % CI 0.180 to 0.933; p = 0.034). Number needed to treat: 25.
• Reinjuries : 7.1 against 45.8, ratio 0,137 (95 % CI 0.037 to 0.509; p = 0.003). Number needed to treat: 3.

That last figure is the most important in the whole article for a private-practice physiotherapist: it is in the patient who has just been injured that the intervention is by far the most profitable. Treating three players with a previous injury prevents one reinjury.

A second randomised trial, run in 579 amateur footballers from 40 teams, confirms the effect outside the professional setting: 25 sessions over 13 weeks, incidence of 0.25 against 0.80 per 1,000 hours, odds ratio of 0,282 (95 % CI 0.110 to 0.721; p = 0.005). Two useful qualifications: the severity of injuries was not reduced, and adherence reached 91 %, which is characteristic of a protocol supervised by a study²².

What becomes of the effect when the method is tightened

A 2019 meta-analysis covering 15 studies and 8,459 athletes concludes to a risk ratio of 0,49 (95 % CI 0.32 to 0.74; p = 0.0008), that is “half the injuries”, a formula taken up everywhere since¹⁸.

Two years later, a methodological reanalysis takes up the same base. Its starting point is blunt: of the 15 studies included, only five genuinely randomised participants to the exercise. Restricting to those, then adding a trial identified by the update, the risk ratio becomes 0,59 with a 95 % confidence interval of 0.27 - 1.29, and a wider prediction interval still. The authors' conclusion: the evidence of a protective effect remains inconclusive, coming essentially from trials at high risk of bias, and at best a conditional recommendation can be made, for football¹⁹.

The real limitation is not the evidence, it is adherence

A 2015 survey measured the actual application of the programme in 50 professional clubs, 32 from the Champions League and 18 from the Norwegian league, over three seasons. The response rate was 100 %. Of the 150 club-seasons covered, the complete programme was carried out in 16 cases (10.7 %) and partly in 9 others (6 %); 125 club-seasons (83.3 %) were classed as non-compliant. The authors conclude that adoption is “too low to expect any effect at all on injury rates”²⁰.

Seven years later, the situation had not changed: of 17 teams in the UEFA study in 2020/2021, one used the complete original programme and four applied it to all or almost all the squad, that is an adoption of 13 %. Eleven teams reserved it for players with a previous injury. Yet the teams that applied it collectively had fewer injuries (5 against 11 per team ; p = 0.008) and a lower injury burden (12 against 35 lay-off days per 1,000 hours; p = 0.003)²¹.

Does strain affect women and older athletes differently?

Almost all the literature in this file describes male professional footballers. This chapter exists to name what that leaves in shadow, rather than pretend the question does not arise.

Women

The most direct data come from the UEFA study of elite women's clubs. Eleven medical leads from European professional clubs were invited to propose and then rate the modifiable risk factors for hamstring injury. Twenty-one factors were proposed, and the conclusion is clear: most are extrinsic, linked to the club, the team and the staff, and not to the players themselves. The highest rated were lack of communication between medical and coaching staff and the load imposed on players (3.9 out of 5 each), then the lack of regular exposure to high-speed actions in training (3.8) and playing two to three matches a week (3.7)⁴¹.

This study must be read for what it is: a collection of expert opinions, level of evidence III, across eleven clubs. It does not measure risks, it describes what clinicians in the field judge decisive. Its value lies elsewhere: it shifts the question from individual “fragility” towards organisation, which the male data independently support since match exposure is the setting of nine reinjuries out of ten¹.

One gap must be stated plainly: the large eccentric exercise prevention trials were conducted in men^17,22, and the meta-analysis concluding to a halving includes studies in women as well as in men without the effect having been established separately for each sex¹⁸. Transposing the male figures to women athletes is reasoning by default, not data.

The older athlete

Age is one of the two most consistent risk factors in the 2020 meta-analysis, along with previous injury²³. In the 2011 cohort, the incidence of muscle injuries rose overall with age, but, an important detail, that rise concerned significantly only the calf, and not the hamstrings, the quadriceps or the adductors². In other words, the effect of age on this particular injury is less clear-cut than is usually said.

The most typical clinic patient is in any case not a professional. A case published in 2025 describes a 52-year-old active woman, injured running between the bases in a recreational ball game, in front of three orthopaedic surgeons who witnessed the scene. Diagnosis made on the mechanism and clinical examination, non-operative treatment, gradual improvement and return to full activity. Its author stresses that these injuries are common in middle-aged recreational athletes, in whom activities that had been risk-free until then become a source of unexpected vulnerability⁴².

What do concrete clinical cases teach us?

Four published cases, chosen because each illustrates a point the cohorts do not show: what happens when return is decided on insufficient criteria, and what the rare forms look like.

How do you apply all this concretely in practice?

What follows condenses the previous chapters into an approach. Recommendations resting only on expert consensus or on custom are flagged as such.

At the first consultation

1. Date and qualify the mechanism. Fast running or slow stretch? The answer changes the order of magnitude of the timeline announced, from 23 to 43 days in the same trial depending on the type¹³, and up to a year on a pure stretch mechanism¹².
2. Rule out the red flags listed above, in particular proximal avulsion and sciatic nerve involvement.
3. Set the baseline measurements that will serve to decide on return: length of the painful area on palpation, strength in outer range, range in hip flexion with active knee extension²⁹. With no initial measurement, there will be no exit criterion.
4. Announce a range, not a date. And announce the reinjury risk at that moment, because it is the information that will determine the patient's adherence at the end of the programme.
5. Only request an MRI if it changes management. If it is requested, within five days, with explicit mention of the classification system used³¹.

During rehabilitation

1. Load in a lengthened position, progressively, rather than stretching passively and strengthening analytically^13,14,16.
2. Bring in agility and trunk control early: that is the arm of the 2004 trial whose one-year reinjury rate was 7.7 % against 70 %¹⁶, and the London consensus gives 95 % agreement to the importance of the kinetic chain⁷.
3. Do not try to bring eccentric work forward to save time: the trial that tested that strategy finds no gain²⁷.
4. Progress on capacity and symptoms, bearing in mind that progress in the clinical measures is not linear²⁹.
5. Reintroduce running then sprinting as a stage in its own right: it is the only area where complete absence of pain is a consensus (85.5 %)⁷.

When deciding on return

1. Check the active knee extension test : persistent discomfort is associated with a 2.5-fold reinjury risk at twelve months²⁶.
2. Add a high-speed test when the classic examination has become normal: Askling's active H-test¹⁵, and rate the patient's apprehension, which is data and not noise.
3. Do not confuse return to training with return to competition. The reinjury risk is 9.25 times higher in matches¹, and the target of rehabilitation must be match load (89.8 % consensus)⁷.
4. Explicitly plan for the two-month window following return, in which 69 % of reinjuries occur¹ : it is a load management period, not a normal one.

After the return

Keep up eccentric work. It is in the patient who has just been injured that the benefit is greatest: three patients treated to prevent one reinjury, against twenty-five to prevent a first injury¹⁷. The documented obstacle is not effectiveness, it is stopping the programme: the complete protocol was applied in only 10.7 % of professional club-seasons²⁰, and adoption was still 13 % seven years later²¹.

The timelines you can announce, and where they come from

None of these figures is an individual prognosis. They serve to place a patient within an order of magnitude, while naming the population they come from.

Frequently asked questions

A hamstring strain, how long is the lay-off?

The median observed in professional football is 13 days, with an interquartile range of 7 to 22 days¹. But that median covers situations running from 6 days for a functional injury to 88 days for high-grade proximal involvement^1,10, and up to nearly a year for a pure stretch mechanism¹². The mechanism and the site matter more than the grade.

Is an MRI needed for a strain?

Not routinely. The diagnosis is clinical. MRI is justified when it changes management: diagnostic doubt, suspected tendon involvement or avulsion, a sporting deadline to arbitrate. It specifies the site but explains only between 7.6 % and 11.9 % of the variance in the return timeline³¹ : it does not set a date.

Why do hamstrings recur so much?

Because return happens within a window of vulnerability that is not clinically visible. Eighteen per cent of injuries are reinjuries, of which 69 % within two months of return to play, and the risk is 9.25 times higher in matches than in training¹. An earlier return is independently associated with reinjury²⁶.

Is the Nordic hamstring really effective?

It depends on the question asked. Programmes that include it are associated with a reduction of about half in risk¹⁸, but a reanalysis restricted to trials genuinely randomising to the exercise gives a risk ratio of 0.59 with an interval of 0.27 to 1.29, therefore inconclusive¹⁹. The best-established effect concerns reinjuries, with a number needed to treat of 3¹⁷. In practice, it remains the best tool available.

Should a muscle that has just been strained be stretched?

Not in the form of isolated static stretching. The only trial that compared it with an agility and trunk stabilisation programme found 70 % reinjury at one year against 7.7 %¹⁶. Loading the muscle in a lengthened position, progressively and actively, is on the other hand what Askling's two randomised trials support^13,14. Stretching and loading in a lengthened position are not the same thing.

Does a platelet-rich plasma injection speed up healing?

The double-blind placebo-controlled trial, in 80 athletes, finds neither faster return to play nor fewer reinjuries at one year^39,38. Later meta-analyses swing between a non-significant gain of 5.67 days³⁷ and a gain of 8.6 days with heterogeneity of 94 % and certainty rated low to moderate³⁶. The reinjury rate is unchanged in every analysis.

Is it the same as hamstring tendinitis?

No, and it is a common confusion. A strain is an acute injury of the muscle belly, on a datable mechanism. Proximal tendinopathy is an overload condition of gradual onset, with deep pain under the buttock made worse by prolonged sitting^40,46. The two have different treatment and prognosis: see the article devoted to proximal tendinopathy and the one on how long hamstring tendinitis takes to heal.

Can a first injury be prevented in a patient who has never had one?

Yes, but the benefit is far smaller than in someone with a previous injury: the number needed to treat is 25 to prevent a first injury against 3 to prevent a reinjury¹⁷. In amateur athletes, a programme of 25 sessions over 13 weeks reduced the incidence with 91 % adherence²².