Can We Trust Pain-Free Shoulder Force Testing?
The Shoulder Strength Testing Out of Pain (SSTOP) study
Tim Cook - Physiotherapist, lecturer and researcher
The Shoulder Strength Testing Out of Pain (SSTOP) study
Tim Cook - Physiotherapist, lecturer and researcher
A dynamometer gives us a number. But how confidently can we interpret it?
When assessing someone with rotator cuff-related shoulder pain, measuring force can seem an attractive way to quantify their difficulty. We might compare shoulders, repeat a measurement or use the result to inform rehabilitation.
However, those decisions depend on understanding how much readings vary.
In the Shoulder Strength Testing Out of Pain (SSTOP) study, my co-authors and I investigated whether pain-free shoulder force testing produced sufficiently consistent measurements—and how those measurements related to the pain and disability people reported.[1]
Our findings highlight an important distinction: a reassuring reliability score does not necessarily mean that individual readings agree closely enough to interpret small differences confidently.
Dynamometers offer a practical way of measuring force. Clinical guidance recommends handheld dynamometry when assessing isometric shoulder strength in people with rotator cuff disorders.[3]
However, the way we test matters.
Asking someone to push as hard as possible despite pain is different from asking them to stop when pain begins. The resulting numbers represent different aspects of performance.
We wanted to investigate a pain-onset approach: participants gradually increased force against fixed resistance and stopped when they first experienced pain.
The intention was to measure the force they could produce before pain began, rather than asking them to continue through it.
I think of this as pain-free force tolerance under particular testing conditions. It should not be treated as a pure measure of maximal muscle strength.
We recruited 31 adults with rotator cuff-related shoulder pain affecting one shoulder for at least three months. Participants did not have shoulder pain at rest.
Two musculoskeletal physiotherapists independently assessed both shoulders on the same day using an externally fixed EasyForce portable dynamometer. Fixing the device meant that the examiner did not have to hold it against the participant’s effort.
We tested two movements: external rotation, pushing the forearm outwards with the elbow bent beside the body, and abduction, pushing the upper arm outwards.
Participants sat with back support. For external rotation, the cuff was positioned around the wrist. For abduction, it was positioned around the lower upper arm, with the shoulder slightly away from the body.
We standardised positioning, stabilisation, resistance placement and instructions, consistent with published considerations for muscle-strength measurement.[2]
The asymptomatic shoulder was tested first. Participants gradually built force over five seconds, stopping if pain began, and completed three repetitions per condition. We analysed both the average and highest readings.
We randomised which examiner tested each participant first. Participants could not see their scores, and neither examiner could see the other’s measurements.
This allowed us to examine two separate questions: how repeated efforts varied with the same examiner within a session, and how measurements agreed between examiners.
Repeated efforts with the same examiner showed good to excellent relative reliability. Intraclass correlation coefficients, or ICCs, ranged from 0.89 to 0.97.[1]
However, there was still substantial variation in individual readings.
Our within-session repeatability coefficients reached 34.5 newtons, depending on the movement and examiner. These describe the scale of differences expected between repeated readings because of measurement variability.
That distinction matters: participants could maintain their relative position within the sample while their own repeated measurements still varied appreciably.
These were repeated efforts during one appointment. They do not establish how reliably the test performs across separate appointments.
Between-examiner ICCs were also good, ranging from 0.83 to 0.89. But the limits of agreement were wide.[1]
Limits of agreement describe the range within which approximately 95% of differences between two examiners’ measurements would be expected to fall under similar conditions.
For average external rotation force on the painful side, these limits were approximately −26 to +25 newtons.
In other words, one examiner could record a substantially different value from the other for the same person on the same day.
Before analysing the results, we specified an acceptability criterion of ±15% of the overall mean force for each measure. Between-examiner agreement fell outside that criterion in every tested condition.
This was our study’s prespecified measurement-error criterion. It was not a threshold for a change that patients would consider important.
An ICC reflects how well measurements distinguish people within a sample, taking account of the variation between them.
Imagine a group containing people who produce quite different amounts of force. A test may distinguish those people reasonably consistently even when repeated measurements for each person fluctuate.
For clinical decisions, we also need to know how large those fluctuations are.
That is why I would encourage clinicians to look beyond a description such as “good reliability” and examine the absolute measurement error.
A test can distinguish people reasonably well while still being imprecise for interpreting small differences in one person.
We also calculated a relative pain-free force index, comparing the painful shoulder with the other shoulder, and examined its association with the Shoulder Pain and Disability Index, or SPADI.
These associations were weak to moderate and differed between examiners.[1]
Questionnaires assessing pain self-efficacy, pain catastrophising and fear of movement showed numerically stronger associations with SPADI.
Pain self-efficacy—confidence in carrying out activities despite pain—had the strongest association, at r = −0.71. Greater confidence was associated with less reported pain and disability.
These were exploratory findings from a small sample. They do not establish causation, show that confidence explains the force readings, or demonstrate that changing confidence will necessarily improve disability.
For me, they reinforce the value of understanding the person’s experience alongside their physical performance.
My interpretation is that pain-free force readings need context.
A small difference should not automatically be labelled improvement or deterioration. First, consider whether it could plausibly reflect the variability associated with that particular protocol.
A difference large enough to exceed measurement error is also not automatically important to the patient. Detectable change and meaningful change are separate questions.
Force testing may still contribute useful information, but I would interpret it alongside symptoms, confidence in using the shoulder and performance of activities that matter to the person.
The findings do not establish that dynamometry is useless. They highlight limitations of using this pain-free protocol as a standalone measure of recovery.
Our study involved a small sample from one centre, with testing completed on the same day.
We did not establish between-appointment reliability, responsiveness to rehabilitation or how much force change patients would regard as important.
Repeated efforts may also be influenced by familiarisation, fatigue or changes in the experience of pain. We cannot determine how much each contributed to the variation we observed.
The results should therefore remain specific to the population, equipment and pain-onset protocol studied. They should not be generalised to every form of shoulder force testing.
SSTOP showed that good relative reliability can coexist with substantial measurement error.
I carried out this study with Iker Bautista, Luke Emmett and Jeremy Lewis. Our findings encourage caution when interpreting small differences in pain-free shoulder force and support assessment beyond the dynamometer reading alone.
The number may contribute to the clinical picture. Understanding its uncertainty helps us decide how much weight to give it.
Read the open-access SSTOP paper.
You can also find the study in the Publications section of the Physio Principles website and a visual summary in the Infographics section.
This work was supported by the National Institute for Health and Care Research (NIHR) Applied Research Collaboration (ARC) Kent, Surrey and Sussex (KSS) Individual Development Award (IDA).
Cook T, Bautista IJ, Emmett L, Lewis J. The Shoulder Strength Testing Out of Pain (SSTOP) Study: reliability of fixed portable dynamometry for measuring pain-free shoulder force in people with rotator cuff-related shoulder pain (RCRSP). JOSPT Methods. 2026. https://doi.org/10.2519/josptmethods.2026.0037
Bohannon RW. Considerations and practical options for measuring muscle strength: a narrative review. BioMed Research International. 2019;2019:8194537. https://doi.org/10.1155/2019/8194537
Lafrance S, Charron M, Roy JS, et al. Diagnosing, managing, and supporting return to work of adults with rotator cuff disorders: a clinical practice guideline. Journal of Orthopaedic & Sports Physical Therapy. 2022;52(10):647–664. https://doi.org/10.2519/jospt.2022.11306
This article provides general education based on evidence and professional interpretation. It does not replace individual assessment or treatment. For personal advice, consult an HCPC-registered physiotherapist or another qualified healthcare professional. Clinicians should apply their own clinical judgement.
Tim Cook © 2026 Physio Principles