# 58: Most rotator cuff tears don't hurt

[Music] Welcome to the Shoulder Physio Podcast, a podcast dedicated to exploring meaningful topics in Musculoskeletal healthcare. I'm your host, Dr. Jared Powell. Before we begin, the primary purpose of this podcast is to educate and inform. The views expressed in this podcast by myself and any guests are information-only. Do not constitute professional advice and our general in nature. If you act on the basis of any podcast episode, you should obtain specific advice from a qualified health professional for for-proceeding. G'day, it's Jared. I want to talk today about a study that I think every clinician who works with shoulder pain should read and that every clinician who orders or interprets imaging for broad spectrum, Musculoskeletal complaints, needs to seriously reckon with. It's called the Fmaj study, F-I-M-A-G-E, or the Finnish imaging of shoulder study, published in Jama, Internal Medicine, in February, 2026. Now, it's a population-based MRI study of the shoulders of 602 Finnish adults, age 41 to 76. They scanned both of their shoulders. They used 3T MRI, which is the high resolution stuff, and they then had three experienced Musculoskeletal radiologists reading the scans independently blinded to the symptoms of the participant. They then compared what was on the scan to whether the person had shoulder pain or not. Now, the headline finding has been making the rounds all over social media over the last couple of months. 98.7% of participants had at least one rotator cuff abnormality on MRI. Let's say nearly everyone in the study. But I want to make a case in this episode that the headline isn't the most interesting part of the paper. The more interesting parts are buried a few layers in as they tend to be, and I think they raise some very important questions about how we use imaging, how we talk about what imaging shows, and whether the language we're inherited from orthopedics remains fit for purpose. So let's rip in. So part one, what the Fmaj study actually found. Let me give you the numbers clearly and precisely to set up this episode. 602 participants, median age of 58, half male, half female, drawn from a nationally representative Finnish health cohort. So this is not a clinic sample. It's not people who showed up to clinic because their shoulder hurt. It's a random sample of the general adult population. So of these 602 individuals, only 71234567 had structurally normal rotator cuffs on MRI. So that equates to roughly 1.3% of the sample. So if I had to give you a different headline for this study, it would have been that 99% have abnormalities. It would be a clean rotator cuff scan after age 40 is statistically rare, rarer than rain in the Aussie Outback. In fact, it is the exception. It is not the rule. So of the 99% with abnormalities, only 25% had what was classified as tendonopathy. 62% a whopping 62% had partial thickness tears and 11% had full thickness tears. So interestingly in this sample, the partial thickness rotator cuff tear was the most normal or prevalent finding. Now the second important piece, they didn't only look at prevalence. They compared what was on the scan to whether the shoulder was symptomatic or painful or not. And remember, they imaged both shoulders left and right. So they could compare a person's painful shoulder to their non-painful shoulder, controlling for everything that's the same about that individual. What they found was that 96% of asymptomatic or non-painful shoulders had a rotator cuff abnormality and 98% of symptomatic or painful shoulders had a rotator cuff abnormality as well. So the difference between the group groups was let's say 1.8% with a confidence interval, which is important, that crosses zero. So in other words, if you walked into a room of 100 people and you knew which of their shoulders hurt, a rotator cuff MRI would not reliably tell you which one. Maybe the most striking individual statistic in this paper. Of the 96 full thickness tears, they identified 75 of them, which is roughly 80% in the asymptomatic shoulders of the 26 people who had bilateral full thickness rotator cuff tears. 17 of them reported no shoulder symptoms in either shoulder. None, that warrants repeating. 2/3 of people with bilateral full thickness rotator cuff tears were walking around feeling absolutely fine from a shoulder perspective. Quick one before we continue. On Saturday, June 27th, I'm running a face-to-face workshop in Brisbane, Queensland. It's my first live teaching event in four years and I'm super excited to be in a room with clinicians again. This is where the magic happens. So if you want to sharpen how you assess, reason through and treat shoulder pain, in person with me, with real discussion and real cases, this is it. Spots are limited, head to shoulderphysio.com for all the details. Right, back to it. Okay, part two, the age story. So the other thing that Farmaj makes vivid is how rotator cuff pathology evolves over time. Rotator cuff pathology seems to be less about an instance of trauma that causes a tear but rather a trajectory of age-related change. So in the study, in the youngest participants, those aged between 41 and 44, the dominant finding was 10 and up a thick. Full thickness tears were essentially absent. As age increased, that picture changed. Partial thickness tears became more common. Full thickness tears started appearing by age 70 and over. Roughly 28% of participants had a full thickness tear in at least one shoulder. So the rotator cuff seems to follow a predictable degenerative pattern with age. Tendentopathy in middle age, partial tears in the 50s and 60s, full thickness tears in the older decades. This is what tissues do as we get older. It's what skin does. It's what hair does. In my case, it fell out sadly. It's what are indivutable discs do. It's what menis guy do. It's what cartilage does. Tissues accumulate change. These are living things and they are not immune to the effects of time. They fray. They thin. They develop defects. And I want to make a point here that can get lost in the way we talk about this stuff. When we use the word tear, we tend to imagine an event or specific moment in time. We imagine a person reaching up and feeling something snap. We imagine a person lifting a weight that carries with it a trauma and a failure. Something happen. Something with a need, cause and effect. And this can happen at times. But this doesn't appear to be the normal order. So in most cases, this is not what's happening at all. We imagine a trauma, a failure. Something that happened. But in most cases, that's not what's happening at all. What for Mars is showing and what the broader literature has been showing for many years is that most rotating cuff tears are not events. They are accumulations. They are attrition. They are the slow, attritional outcome of decades of loading, of vascular changes, of cellular changes. They're more like the way a piece of fabric thins when it's been folded and unfolded thousands of times than they are like a piece of cloth being ripped. Now this is not simple semantics or pedantry. I'm not getting caught in the minor details here. The language we use shapes what the patient hears and what the patient hears shapes what they believe is happening to their body. What they believe needs to be done about it and how they understand themselves moving forward. Okay, part three. This is not new information and it isn't just in the shoulder. Now I want to zoom out because if you've been around the musculoskeletal literature for a while, you'll recognize that the Femar study that we're talking about today is not a one-off. It is the latest entry in a long and consistent line of studies showing the same thing in different anatomical regions and different musculoskeletal conditions. Let's take the England study from 2008 published in the New England Journal of Medicine. They did MRIs of the knees of nearly a thousand people aged 50 to 90 in the general population. They found that meniscal tears were present in 35% of all participants and crucially, the prevalence of meniscal tears were almost identical in people with knee pain and people without it. 61% of people who had a meniscal tear on MRI had no knee pain at all in the previous month. The meniscal tear on its own told you almost nothing about whether the knee hurt. Let's take the Jensen study also in the New England Journal of Medicine published way back in the 1990s. They did a number spine MRIs on people who had no back pain. What they found was that 52% had a disc bulge, 27% had a disc protrusion, only 36% had completely normal discs. And these are people with no symptoms at all. There are similar studies in cervical spine imaging in hip imaging where labral tears and care morphology which is the shape of your hip are extraordinarily common in asymptomatic populations. We also see it in ankle imaging as well. The pattern is holding consistently across the body. Structural findings on imaging are common. They accumulate with age and they correlate poorly with whether the person has pain or not. So I think for March isn't a shoulder story but it is the shoulder chapter of a much bigger story or bigger narrative or bigger book. And the biggest story or bigger narrative is the model that links pain to a specific structural finding. The model that says this hurts because that's torn has been failing on its own terms for decades. Study after study. Region after region. The body refuses to cooperate with the simple mechanical narrative and yet here we are talking about it today. This narrative is persisting. You see it in the clinic every week as do I. It persists in radiology reports. It persists in patient conversation. It persists in the wider conversation in society. It persists in surgical decision making. Even though the evidence undermining it has been around for 30 years or more. Why? It begs the question why? Why is this narrative persisting? I think there are a few reasons. One is that the simple mechanical story is intuitive. It's easy to explain. It's easy to draw on a whiteboard. It feels satisfying to both the clinician and the patient. The pain has a cause and the cause is visible and the visible cause can be addressed. Another reason is that imaging is impressive. Technology is impressive. A 3T MRI is an extraordinary bit of kit. Extraordinary bit of technology. It produces beautiful high resolution pictures and there's a deep human tendency to assume that something that looks impressive must be telling us something important. We over trust the image because the image looks like the truth. A third reason is that there's an entire economic and professional ecosystem built around imaging driven decision making. Radiology revenue, surgical referrals, injection procedures, the whole apparatus, the whole system assumes that the picture is important in a particular way. To question that assumption is to question a lot of practice that depends on it and we're not very good at questioning our practice where many lively hoods are built upon it. Part 4. But is prevalence the same as normal? So I want to deviate slightly for a second and still man an objection that I think a thoughtful listener will already be forming because it's a good one and I don't want to dodge it. I want to confront it and talk about it honestly. So the objection goes something like this. Just because something is prevalent doesn't make it normal. Lots of people are overweight. That doesn't make being overweight benign. Smoking was once near universal in many populations. That did it make lung damage normal. High attention is common. We still treat it. So the move from 99% of people over 40 have rotated up abnormalities. Two, therefore these are just normal age-related changes needs a stronger argument than prevalence alone. And I think that's right. Prevalence and normality are not the same thing. We need to be honest about that. So let me try to make the argument because it's more interesting one than the prevalence statistic on its own. The reason we tend to call obesity normal, even though it's common, isn't the prevalence. It's in fact the harm trajectory. Obesity is associated with cardiovascular disease, diabetes, increased mortality and joint problems. The prevalence isn't what makes it pathological. The downstream consequences are same with smoking, same with hypertension. The label abnormal earns its keep by predicting something we care about clinically. Now let's apply that test to rotaticaf abnormalities. The question is what do these structural findings predict? Specifically do they predict the thing that we've been using them to predict which is shoulder pain and dysfunction in the person sitting in front of us? And what the Femar study shows and indeed what the broader literature has been showing for decades is that answer is not really, not in any clinically useful way. When you control for the actual clinical picture, the structure finding loses almost all of its discriminating power. Most full-fitness tears are in people who feel fine, same with partial thickness tears, same with tendonopathy. The scan doesn't reliably predict the pain. So this isn't an argument that prevalence equals normality. It's a more nuanced and specific argument. It's an argument that the structural finding has very poor discriminating power for the specific clinical question we've been using it to answer. This is a narrower claim, but it's a defensible one and it's the one that the Femarge data supports. Now I want to be careful here because this is where the steelman has to keep working. It is possible that rotaticaf degeneration matters for things other than current pain. It might matter for long-term function. It might matter for biomechanical efficiency or strength or for future symptom development or for outcomes after a traumatic injury. Indeed we do have data that suggest the bigger rotaticaf tear gets the more likely it is to be associated with shoulder pain. Femarge doesn't answer any of these questions specifically. So I'm not making the absolute claim that these findings are benign in every sense. I'm making the narrower claim that they don't reliably explain why a particular shoulder hurts today on average or in most situations. The most defensible position I think is this. The label abnormal should earn its keep. If a finding doesn't predict the clinical presentation it's been blamed for, the label needs to be reexamined. That doesn't mean we throw out the label entirely. It means we hold it more lightly. We stop treating the scan as the answer and start treating it as one piece of information that has to make sense alongside everything else. If we discovered tomorrow that BMI or waste circumference or physical activity levels didn't predict cardiovascular events, didn't predict diabetes, didn't predict mortality. If all the downstream harms turned out to be confounded by other factors, then yes, the case for treating elevated BMI as a clinical abnormality would weaken substantially. We'd have to rethink it and I would be the first to do it. That's the position wherein we throw ataticaf imaging. The downstream prediction is much weaker than we assumed and the response to that should be intellectual honesty and certainly not doubling down on a model that's been failing for decades. Part five. So is imaging useless? Again, I want to tread carefully here because this is the place where the argument can swing too far in the other direction. Nothing in the Firmar study and nothing in this podcast episode is saying that imaging is useless. Imaging is extraordinarily valuable in the right context. If you've got a young person with an acute traumatic injury and a sudden loss of strength, an MRI can be of profound utility. If you're worried about a tumor or an infection or a fracture, imaging is the gold standard. If a patient has failed non-surgical management and you're seriously considering surgery, imaging informs that decision. These are clear, defensible use cases for shoulder MRI or ultrasound or X-ray or whichever you want to use. The problem is the routine reflexive use of imaging in a traumatic shoulder pain, which is the situation Firmar is addressing and the clinical presentation you're most likely see in clinic. When you image someone with vague, gradual onset shoulder pain in their 50s, the prior probability of finding something abnormal on the scan approaches 100%. And when the prior probability is that high, the scan loses almost all of its diagnostic discriminating power. It's not telling you anything specific about why this person hurts. It's telling you what their tendons look like, which statistically look the same as everyone else's tendons of that age. So the authors of Firmar put it in a way that I will quote almost directly. They argue that we should shift from asking whether abnormalities exist to whether they plausibly explain the clinical findings. This is a seemingly small shift in language but a big shift in reasoning. The first question doesn't abnormality exist, almost always returns. Ah yes, the second question, does this abnormality plausibly explain what's happening with this person is much more interesting and much harder to answer and much more likely to lead to good clinical decisions. This reframe puts the clinician back in the driver's seat. The scan is no longer the answer. The scan becomes one piece of information at data point that has to be integrated with everything else. The history, the examination, the patient's goals, values and preferences, the trajectory of their symptoms, the plausibility of the clinical hypothesis. The scan stops being averted and again, as I mentioned a moment ago, it becomes a simple data point. Part 6, the language problem. I want to come back to the language thing. The authors are for March explicitly recommend that clinicians and radiologists move away from the word tear. They suggest alternatives such as lesion, defect, fraying, disruption, structural alteration, degeneration. The reasoning is that the word tear carries connotations of trauma, of acute failure, of something being broken. And those connotations shape what the patient believes about their body and what they believe needs to be done about it. Now I've been thinking about this a lot lately, not just lately, basically for the last 10 years, because if you take the findings of the FMAJ study seriously, if you take the broader imaging literature seriously, then the picture that starts to take shape is that most rotating cuff tears are not really tears in the way that word implies. They're not events, they're processes, they're nutritional, they're cumulative, they're degenerative changes that have been progressing for years, and that have at some point along the trajectory, crossed a threshold of visibility on a scale. Calling that a tear is a category error. It imports or trojan horses the language of trauma into a phenomenon that is much more like aging than like injury. This language carries a whole cascade of consequences with it. The patient hears torn and pictures damaged, they protect the shoulder, they stop using it the way they used to, they become anxious about loading it, they start to believe that movement is dangerous, they look for someone or something to fix the broken thing. And in many cases, the very behaviors that the diagnosis encourages or facilitates are the behaviors that prolong the pain experience and undermine recovery. Now I'm not the first person to make this point and I won't be the last, but I think the time is over due for our profession and all professions that work with individuals with shoulder pain or any musculoskeletal pain for that matter. To reckon with the language we've inherited and ask whether it actually serves our patients. I'm increasingly convinced that we should be talking about rotating cuff changes, the way dermatologists talk about skin changes as something the body does over time, not something that breaks suddenly. Part 7, so what does this mean for practice? Let me try and land this in a way that is honest and helpful for you, the clinician. So if you take for March seriously and I strongly suggest you do a few things follow. First, routine imaging for atraumatic shoulder pain is not justified. It rarely changes management, it almost always finds something and what it finds is usually clinically ambiguous. The authors of for March including a senior orthopedic surgeon explicitly call into question the clinical value of routine imaging in this population and I think that should carry some weight. Second, when imaging is done for whatever reason the findings need to be interpreted with the population prevalence in mind. If 99% of people in this age group have an abnormality then finding an abnormality in your patient is not informative on its own. The question is not is there pathology? The question is does this pathology plausibly explain what I am seeing in this patient clinically? And often the answer is no or at least we don't know. Third, the way we talk to patients about imaging findings matters enormously. We can describe the same MRI in two completely different ways. For example, we can say you have a partial thickness tear of the super-spinatus tendon which is concerning and may need further investigation or we can say your scan shows changes that we'd expect to see in most people of your age including people who have no shoulder pain at all. These changes are very common they're not necessarily related to your symptoms and they don't tell us that anything is broken. Both of these statements are technically accurate. One of them I think is much more useful to the person sitting in front of you. And finally I think this kind of evidence should make us more humble about the limits of structural explanations about pain. Pain is not a simple readout of tissue damage. It never has been. The body as machine model the model where pain is the alarm bell of broken parts keeps failing and for Marge is one more piece of evidence that we need a different model. I don't think we need a model that ignores tissue but a model that holds tissue findings in their proper place as one variable among many with limited explanatory power on their own. So in conclusion the headline of the Marge is that 99% of adults over 40 have rotated up abnormalities on MRI. But the most important story in my view is what the prevalence does to the meaning of any individual scan. When everyone has the finding the finding stops being a finding it becomes a feature of being a body that has lived for a few decades or more. This is a freeing conclusion. It frees us from the burden of finding the broken part. It frees the patient from the belief that their shoulder is damaged and it pushes us back toward the kind of clinical reasoning that for Marge itself is calling for contextual, integrated, humble about the limits of imaging and focused on the person rather than the picture. Thanks for listening. I'm Dr. Jared Powell and this has been another episode of the Shoulder Physio podcast. I'll talk to you next time. The Shoulder Physio podcast would like to acknowledge that this episode was recorded from the lands of the terrible ang people. I also acknowledge the traditional custodians of the lands on which each of you are living, learning and working from every day. I pay my respects to elders past, present and emerging and celebrate the diversity of Aboriginal and Torres Strait Islander peoples and their ongoing cultures and connections to the lands and waters of Australia.