Is My Dog's Limp Getting Better, or Am I Just Hoping?

If you are asking this, there is a fair chance somebody else in the house has already answered it differently. That disagreement is not a failure of attention. In the placebo arm of a blinded osteoarthritis trial, owners judged their dogs improved 39.7% of the time while the mean force plate readings underneath did not move, and the veterinarians did it slightly more often than the owners did. The fix is not to watch harder. It is to write three things down on a schedule: a questionnaire with a threshold you set in advance, a weekly video filmed the same way every time, and nine items that have a definition instead of an impression. All three are free and you can start tonight.

Why can't I just tell whether he's better?#

Because nobody can tell by looking, including the people who do it for a living.

Fifty-eight dogs with lameness from osteoarthritis were enrolled in the placebo arm of an FDA-approved trial, with force platform gait analysis running throughout. Owners reported improvement 39.7% of the time. Veterinarians examining the same dogs reported it 44.8% of the time at a walk, 44.8% at a trot, and 43.1% when checking for pain on palpating the joint. Meanwhile the mean ground reaction forces did not move. Taken dog by dog over 42 days, five got measurably worse by 5% or more, seven got measurably better by 5% or more, and forty-six did not change at all. The effect was significantly stronger the longer the trial ran.^[1]

A longer look was not a more reliable look. It was a more contaminated one.

Nobody in that trial was being careless. They were being asked to measure something with memory, and memory does not hold last month's dog still while you compare him to this one.

The second study is harder still on the eye. Three small-animal veterinarians with orthopedic training scored twenty-one dogs on both a numerical rating scale and a visual analogue scale, against force plate readings taken at the same moments. Agreement between the three of them was generally low, reaching an acceptable level only at the four-week time point. Only one force plate variable had a significant relationship with any observer's scores. And when the very lame dogs were taken out of the analysis, no significant relationship between any observer's scores and the force plate data survived at all. The authors' conclusion was that each observer uses an individually unique scale.^[2]

One thing to hold against that study before you carry it too far: those were research dogs with a surgically created lameness — a tibial osteotomy held with an external fixator — scored before surgery and at four and eight weeks. That is not your dog with a chronic, fluctuating limp. It tells you about the reliability of the instrument rather than about the disease.

So the eye works when the dog is severely lame, which is the situation in which you did not need help. On the mild-to-moderate limp you are trying to track from one Sunday to the next, it is close to useless.

"Better" is three questions wearing one coat#

Here is the distinction that reorganises the whole problem, and it is the reason two reasonable people in one house can watch the same dog and come to opposite conclusions. They are answering different questions.

• Is he putting weight through the leg? That is limb loading, and a force plate or pressure walkway is what measures it.
• Does it hurt less? That is pain, and owner-completed questionnaires measure it.
• Is he doing more of his life? That is function, and questionnaires and activity monitors measure that.

These do not move together. In 222 dogs with osteoarthritis, three validated owner questionnaires were tested against force platform analysis. They agreed with each other — the authors' word is "moderate" — the Liverpool Osteoarthritis in Dogs score correlating with the Helsinki Chronic Pain Index at 0.766. Against the force plate they were weak or absent: 0.232 for the LOAD total score, 0.297 for its lameness sub-score, 0.281 and 0.276 for the two halves of the Canine Brief Pain Inventory, and 0.128 for the Helsinki index, which did not reach significance at all. The authors say outright that they are not certain why the Helsinki index failed to correlate. And in the 79 dogs followed to six weeks of treatment, no significant correlation was found between change in any questionnaire score and change in peak vertical force or in any activity measure.^[3]

The temptation is to read that table as three instruments and pick the accurate one. That is the mistake. The questionnaires are not bad force plates. They are good instruments for a different quantity, which is why "he seems happier in himself but he's still limping" is a coherent observation rather than a contradiction, and why one household member can be right about the pain while the other is right about the leg.

“Because your dog can't tell you how much pain he or she is in.”

The three things to write down#

None of this is new science and none of it is behind a paywall. The first of the three is a validated questionnaire used in drug registration trials and free to download. The second is a filming routine. The third is a scoring checklist from a 2022 study of 27 dogs — promising rather than established, and labelled as such below. What none of them is is routine: nothing in a fifteen-minute consultation obliges anyone to put one in your hands.

1. A questionnaire, monthly, with the threshold set in advance#

The Canine Brief Pain Inventory is four questions about how severe your dog's pain is and six about how much it interferes with ordinary activities, each answered 0 to 10. The mean of the first four is the Pain Severity Score; the mean of the last six is the Pain Interference Score.^[11] The University of Pennsylvania School of Veterinary Medicine distributes it free, as a PDF and as an online form.^[4]

What makes it worth doing is not the questionnaire. It is that somebody has already worked out what counts as a real change. Analysing 150 dogs across two carprofen trials, the definition of treatment success with the most consistent statistical power was a drop of at least 1 point in the Pain Severity Score together with a drop of at least 2 points in the Pain Interference Score, measured from day 0 to day 14.^[5]

That is a threshold you can borrow. Score your dog before the treatment starts. Write the target on the same page. Then score him again on the agreed date, and let the numbers rather than the afternoon decide.

Two honest qualifications. Even in a properly designed trial, that definition detected a real carprofen effect only about 63% of the time, so a single dog missing the threshold is weak evidence rather than a verdict. And the CBPI was validated as an owner questionnaire, not as a lameness meter — see the table above for what it is and is not tracking. If your dog's problem is chronic joint disease rather than a single injured leg, the Liverpool Osteoarthritis in Dogs questionnaire is a good alternative, is also free, and holds together well internally (Cronbach's α 0.88).^[3]

2. A video, weekly, filmed the same way every time#

This is the step that turns a bad instrument into a decent one, and there is a study that shows exactly why.

Six observers scored eighteen dogs with radiographically confirmed elbow osteoarthritis from video, at a walk and a trot, the clips randomly sequenced, each observer scoring every one twice at least two weeks apart. Agreement, both between the six of them and within each of them across the two sittings, came out above 0.9.^[6]

Be careful about what that does and does not prove. It shows that high agreement on canine lameness scoring is achievable — Quinn's three vets, working from the same twenty-one dogs, did not get near it. It does not prove the video is why. Aulakh's study had no comparison arm scoring the same dogs any other way, and the two studies differ in almost every other respect: different dogs, different joints, different observers, different statistics. The inference that holding the surface, the speed, the angle and the dog's day still is what raised the agreement is ours, not the paper's. It is a reasonable inference and it is the only one available for free.

The protocol below. Some of it is sourced and some of it is simply our own choice of a fixed number — the point of a specification here is that it does not move, not that it is optimal:

1. Same route. Ten to fifteen metres of flat, even, non-slip ground. Not a lawn, not tile, not a slope. A dog scrabbling on a hard floor looks lame whether or not he is.
2. Same camera position. Side on, roughly at your dog's shoulder height, at a spot you have physically marked. Put a bit of tape on the paving.
3. Same handler, lead slack. A dog being pulled is a dog whose gait you are altering.
4. Same pace. This is not fussiness. When ground reaction forces were analysed in twenty clinically normal dogs, the sources of variance in order of size were the individual dog, then velocity, then trial repetition, then which limb.^[7] The dog is not a variable you can change. Of the ones you can, speed is the biggest.
5. A walk and then a trot. In the elbow study the same dogs scored as less lame at a walk than at a trot, though how big that gap looks depends on which scale you read it on — 0.17 of a point on the numerical rating scale, 3.54 on the much longer visual analogue scale.^[6] Film both, which costs you thirty seconds, rather than treating either gait as definitive.
6. Same time of day, at the same point relative to rest and to the last dose of anything.
7. Name the file with the date, and never delete one. In March you will want to compare against January, not against last week.

One thing we cannot tell you, because nobody has tested it: whether an owner filming at home produces readings as consistent as six trained clinicians scoring a research video. Almost certainly not. Do it anyway — the alternative you are comparing it against is memory.

dogorthopedic.com

3. Nine counted items with a definition#

In 2022 a group at Tennessee built a stifle function score for dogs with cruciate disease: fourteen items, 100 points, combining what a clinician finds on examination with functional tasks the dog performs. Scored against force-plate symmetry in 27 dogs, the complete score correlated at −0.863 (p < 0.0001).^[8]

Set that beside Quinn's finding that a single eyeballed lameness grade had no surviving relationship with force plate data once the very lame dogs were excluded. Different dogs, different study, so this is a comparison rather than a controlled contrast — but the shape of it is hard to ignore. One glance tracks the plate poorly. A structured checklist of fourteen ordinary things tracks it closely.

The paper splits its fourteen items into eight functional tests worth 65 points and six clinical measures worth 35. All eight functional tests are things you can observe at home. We have added a ninth, thigh circumference, which the paper files under clinical measures — that reclassification is ours, and we make it because a tape measure is a tape measure and the home protocol for using one is published. That gives you nine items and 75 of the 100 points. The five you cannot do are pain response on palpation, joint effusion, stifle extension, stifle flexion, and cranial drawer or tibial thrust.

The scoring criteria below are the study's own, condensed to the anchor points:

Three notes on using that table honestly.

The gift in it is the phrase "weight-bearing on more than 95% of strides." It converts an adjective into a count. Play the video back, count twenty strides, and count how many of them the foot properly loads. That is a number a different person can get the same answer for, which is the only property that matters here.

The two "lameness" rows are the weak ones, by this article's own argument. They ask for a graded impression — barely perceptible, mild, moderate — which is exactly the judgement Quinn's three orthopedic vets could not agree on. Score them from the video rather than from the doorstep, expect them to wobble, and weight the counted rows more heavily in your own reading of the trend.

Two of the items are physical manoeuvres, not observations. "Dancing" means lifting the front legs off the ground, supporting the dog's chest, and walking him forward and back on his hind legs. Thigh circumference is a tape measure; the full protocol, including where on the thigh to measure and why the same person has to do it every time, is in why is my dog's muscle wasting on one back leg. Ask your vet before trying either on a dog whose diagnosis is not yet established.

Now the caveats, and one of them matters more than the study's own. Twenty-seven dogs is a small sample, the score was built for stifles rather than hips, elbows or feet, and the authors state plainly that they lacked adequate repeated measurements to establish whether the score works for monitoring a dog over time — which is precisely what you would be using it for. And the one the paper does not raise, because it had no reason to: the −0.863 and the band table above belong to the whole fourteen-item score, examination findings included. Nobody has published a correlation for the nine-item, owner-only subset, so you cannot claim it inherits the parent score's accuracy. What you inherit is the definitions, which is still the most useful thing available.

One finding worth acting on directly: of the fourteen items, two had no significant correlation with force-plate symmetry. One was stifle flexion. The other was pain response on palpation — pressing the joint to see whether he flinches. That is a common weekly home test, and on this evidence it is the item that tells you least about how much weight your dog is actually putting through the leg.

dogorthopedic.com

What about counting sit-to-stands?#

It is the protocol most often suggested online, and the one piece of evidence that has tested it directly is not encouraging.

Ten dogs with unilateral cruciate rupture and eighteen sound dogs were measured on a pressure-sensitive walkway during both a walk test and a sit-to-stand test. Using peak vertical force symmetry, the walk test identified 100% of the lame dogs and the sit-to-stand 90%. Using vertical impulse, the walk test again identified 100% and the sit-to-stand 50%. The sit-to-stand was quicker to perform, and the authors concluded it has relatively limited clinical utility in the form they tested.^[9]

So keep sit-to-stand as one scored item out of five, because the stifle function score includes it and it earns its place there. Do not build your whole tracking system on it. The plain walk, filmed properly, is the better test — which is convenient, because it is the thing you were going to do anyway.

How long before you should expect to see anything#

The single most useful thing you can do is agree the review date with your vet before the treatment starts, not after. Otherwise the date drifts, and the caregiver placebo effect grows with time.^[1]

Rough anchors, with the honest caveat that each comes from one study population and your dog is not in it:

That surgical study followed twenty dogs after tibial tuberosity advancement rather than TPLO, on a pressure-sensitive walkway. It is a small study of one procedure. Our TPLO recovery guide has the week-by-week version of what to expect after the commoner operation.

What counts as worse#

Short of that, these are the patterns that should move your appointment forward rather than wait for the review date:

• A step down on the same counted item across two consecutive readings. One bad week is weather. Two in the same direction is a trend.
• A limp that is worse at rest than after exercise — stiffness that loosens with movement is the usual arthritic pattern, and a limp that behaves the other way round is worth a different conversation.
• A new limb. A second leg joining in is not the first one worsening; it is a different event, and a dog compensating on three legs is a dog set up for the next injury. We cover that pattern in is my dog limping, or is it something else.
• Something he did last week that he now declines. The car, the sofa, the last flight of stairs. A dropped behaviour often shows up before anything the tape or the video catches, and it is the observation most likely to be filed as him slowing down.
• A thigh that is losing bulk. A limp is a behaviour and a dog can mask it; muscle is a record. The tape-measure protocol lives in why is my dog's muscle wasting on one back leg, and it is the one home measurement that does not depend on your judgement at all.

What your vet does with the file#

dogorthopedic.com

Bring the whole thing. The questionnaire scores with their dates, the counted items, and two or three videos chosen to span the period rather than to show the worst day.

The reason this is worth more than it sounds is buried in the Quinn paper's own advice: observers must stay the same over the course of a study, because each one is running a private scale.^[2] A grade of 2/5 from a locum in April and a grade of 2/5 from your regular vet in August are not two readings of the same instrument. Your file is the only continuous observer in the story, and nobody else can generate it retrospectively.

Expect the consult to add what you cannot: joint-by-joint palpation, drawer and thrust testing on a stifle, range of motion, and imaging if indicated. If the picture is one of weakness or unsteadiness rather than a limp, say so explicitly and ask for a neurologic examination — that is a different appointment, described in telling a limp from weakness.

The gear question, and the one way gear corrupts the measurement#

Everything above costs nothing. You need a phone, a tape mark on the ground, a free PDF and somewhere to write numbers down, and if you buy nothing else after reading this we will consider the article to have worked.

There is one genuine purchase, and it is not for the measuring. It is for the weeks the trend goes the wrong way. A dog who has stopped being able to get up unaided stops going out, and a dog who stops going out loses condition, and lost condition shows up across the counted items — which means the bad fortnight can write itself into the record as a decline that was partly logistics. What a support harness buys is the ability to keep the prescribed short walks happening on the days when the person doing the lifting is tired. We covered the same argument from the exercise side in how much walking is right for a dog with arthritis.

Before you spend anything: a rolled bath towel under the belly, held at both ends, does one or two lifts a day for nothing, and it tells you within an afternoon whether your dog needs holding at the front, the back or both. It stops being enough when the lifts get frequent enough to hurt your own back, or when there is no longer enough muscle on the hindquarters for the towel to grip.

Now the part that belongs specifically in this article: never film him in it. A supported dog's gait is not his gait. The harness changes how much weight goes through the bad leg, which is the exact quantity the counted items exist to track — so a video taken with a hand on the hip handle will read as improvement, and the improvement will be yours rather than his. Take it off, or slacken it and drop the handles, for the ninety seconds of filming. We have not seen it mentioned anywhere, and it is the difference between a record that means something and one that flatters everybody.

Two other things worth knowing before you buy. We re-checked the pricing today: the manufacturer sells the standard matched-size harness at $80.00 for every size and both hip-lift styles, while Amazon runs $95 for the X-Small up to $155 for the X-Large U-Band, and the Amazon buy box is currently held by a third-party reseller rather than by the brand. We link Amazon because that is the programme that pays for this site, and we would still rather you paid the $80. Separately, the Mix & Match option — ordering a smaller rear half for a dog whose hindquarters have wasted, which describes most dogs this harness is bought for — starts at $95, not $80. That $15 is real, and we have not seen it stated anywhere on the ordering path. It is usually still the right choice for a dog with a wasted back end, because a rear half that hangs does not support anything.

The known downside stands: the rear straps can chafe the inner thigh if the harness is left tight while the dog lies down, and the manufacturer's own instruction is to loosen them every time he settles.

What to do this week#

Tonight, download the Canine Brief Pain Inventory, score it, and write today's date on it. It takes about five minutes and it is the baseline that everything else gets compared against. Without a day-0 reading, a day-14 reading tells you nothing.

Tomorrow, put a piece of tape on the ground where the camera goes, film ninety seconds — a walk out and back, then a trot out and back — and name the file with the date. Then work down the nine counted items from the table and write the nine numbers under it.

Then put a recurring reminder in your phone: the video weekly, the counted items fortnightly, the questionnaire monthly. And before you start any new treatment, ask your vet for two things in writing: the date you will both review it, and what would count as it having worked.

The reason to do this is not that your judgement is poor. It is that the thing you are trying to detect is a five per cent change in a body you see every day, and no observer in the literature does that well, including the ones holding a diagnosis. A dated file is not a substitute for your attention. It is the instrument your attention has been missing, and it has been sitting in the clinical literature the whole time, waiting for somebody to move it into a kitchen drawer.

Sources#

1. Conzemius MG, Evans RB. "Caregiver placebo effect for dogs with lameness from osteoarthritis." Journal of the American Veterinary Medical Association. 2012;241(10):1314-1319. doi:10.2460/javma.241.10.1314 — 58 dogs in the placebo arm of an FDA-approved trial; caregiver placebo effect 39.7% for owners, 44.8% for veterinarians at a walk, 44.8% at a trot, 43.1% on joint palpation; effect significantly enhanced with time; mean ground reaction forces unchanged; individually 5 dogs worsened by ≥5%, 7 improved by ≥5%, and 46 were unchanged over 42 days.
2. Quinn MM, Keuler NS, Lu Y, Faria ML, Muir P, Markel MD. "Evaluation of agreement between numerical rating scales, visual analogue scoring scales, and force plate gait analysis in dogs." Veterinary Surgery. 2007;36(4):360-367. doi:10.1111/j.1532-950x.2007.00276.x — 21 dogs, 3 small-animal veterinarians with orthopedic training; agreement generally low among observers for both scales; only impulse had a significant relationship with some observers' scores; no significant relationships between any observer's scores and force plate data existed if very lame dogs were omitted; "each observer uses an individually unique scale"; observers must stay the same for the duration of a study. Note the model: these were mixed-breed research dogs with a right-limb tibial osteotomy repaired with an external fixator, not client-owned dogs with naturally occurring disease.
3. Walton MB, Cowderoy E, Lascelles D, Innes JF. "Evaluation of construct and criterion validity for the 'Liverpool Osteoarthritis in Dogs' (LOAD) clinical metrology instrument and comparison to two other instruments." PLoS ONE. 2013;8(3):e58125. doi:10.1371/journal.pone.0058125 — 222 dogs cross-sectional, 79 longitudinal; LOAD Cronbach's α 0.88; LOAD vs HCPI r_s 0.766; against symmetry index for peak vertical force, LOAD total r_s 0.232 and LOAD lameness index r_s 0.297, CBPI pain severity r_s 0.281 and pain interference r_s 0.276, all p < 0.01, while HCPI came in at r_s 0.128 and was not among the significant correlations — the authors write that it is "not certain why HCPI did not correlate with SI-PVF"; no significant correlations between changes in any instrument's scores and changes in peak vertical force or any activity parameter.
4. University of Pennsylvania School of Veterinary Medicine. "PennChart — Canine Brief Pain Inventory." vet.upenn.edu — the CBPI as an owner-completed questionnaire, distributed free as a PDF and as an online form. This page is cited for availability only; the instrument's structure is sourced separately at note 11.
5. Brown DC, Bell M, Rhodes L. "Power of treatment success definitions when the Canine Brief Pain Inventory is used to evaluate carprofen treatment for the control of pain and inflammation in dogs with osteoarthritis." American Journal of Veterinary Research. 2013;74(12):1467-1473. doi:10.2460/ajvr.74.12.1467 — 150 dogs across two identically designed carprofen studies, day 0 compared with day 14; defining treatment success as a reduction of ≥1 in the pain severity score and ≥2 in the pain interference score had consistently robust power (62.8% where baseline scores were ≥2, 64.7% where ≥3). Note: the published abstract states this criterion in the Results section and reverses the two scores in the Conclusions sentence. We have used the Results-section version, which is the one in general use.
6. Aulakh KS, Dongaonkar KR, Barnes K, Gines AJ, Bordelon JT, Hulse D, Aulakh HK, Liu CC. "Influence of orthopedic examination on lameness scores and interobserver and intraobserver agreement in dogs with naturally occurring elbow osteoarthritis." Veterinary Surgery. 2020;49(3):455-462. doi:10.1111/vsu.13390 — 18 client-owned dogs with radiographically confirmed elbow osteoarthritis; 6 observers each scored randomly sequenced videos twice at least 2 weeks apart; intraclass correlation coefficient for both scores for all observers >0.9; numerical rating and visual analogue scores at a walk were lower than at a trot before orthopedic examination (by 0.17 and 3.54 respectively); an orthopedic examination did not exacerbate scores.
7. Hans EC, Zwarthoed B, Seliski J, Nemke B, Muir P. "Variance associated with subject velocity and trial repetition during force platform gait analysis in a heterogeneous population of clinically normal dogs." The Veterinary Journal. 2014;202(3):498-502. doi:10.1016/j.tvjl.2014.09.022 — 20 clinically normal dogs trotted at habitual velocity; the magnitude of variance effects from largest to smallest was dog, velocity, trial repetition, and limb; a velocity range of 1.5–2.2 m/s captured 84.2% of trials with no significant effect on peak vertical force or vertical impulse.
8. Gundersen K, Millis D, Zhu X. "Development and testing of a stifle function score in dogs." Frontiers in Veterinary Science. 2022;9:895567. doi:10.3389/fvets.2022.895567 — 27 dogs (21 with unilateral cranial cruciate ligament disease, 6 controls); 14-item score out of 100, divided by the authors into eight functional tests (65 points) and six clinical measures (35 points); correlation with force-plate symmetry index −0.863 (p < 0.0001); mean score by symmetry index band 97.8, 85.2, 65.4, 63 and 56.4; all items significantly correlated with the symmetry index except pain response and stifle flexion; stated limitations include the small number of dogs and a lack of adequate repeated measurements to assess the score's utility for monitoring.
9. Triviño A, Davidson C, Clements DN, Ryan JM. "Objective comparison of a sit to stand test to the walk test for the identification of unilateral lameness caused by cranial cruciate ligament disease in dogs." Journal of Small Animal Practice. 2024;65(1):24-29. doi:10.1111/jsap.13679 — 10 dogs with unilateral cranial cruciate ligament rupture and 18 non-lame dogs on a pressure-sensitive walkway; hindlimb peak vertical force symmetry indices were 100% sensitive on the walk test and 90% on the sit-to-stand; vertical impulse symmetry indices 100% and 50% respectively; the sit-to-stand test "has relatively limited clinical utility in its tested form."
10. Guadalupi M, Crovace AM, Monopoli Forleo D, Staffieri F, Lacitignola L. "Pressure-Sensitive Walkway System for Evaluation of Lameness in Dogs Affected by Unilateral Cranial Cruciate Ligament Rupture Treated with Porous Tibial Tuberosity Advancement." Veterinary Sciences. 2023;10(12):696. doi:10.3390/vetsci10120696 — 20 dogs measured before surgery and at 30, 90 and 120 days after porous TTA; lameness score and total pressure index increased significantly compared with pre-operative values from 90 days onward; a statistically significant difference between the operated limb and the healthy contralateral limb remained at every time point, including 120 days.
11. Wells JR, Young AL, Crane A, Moyaert H, Michels G, Wright A. "Linguistic Validation of the Canine Brief Pain Inventory (CBPI) for Global Use." Frontiers in Veterinary Science. 2021;8:769112. doi:10.3389/fvets.2021.769112 — describes the instrument's structure: "four items pertaining to the severity of pain evident in a dog (pain severity domain) and six items relating to pain interference with daily activities (pain interference domain)," each on a 0–10 numerical rating scale, with the mean of each set of responses forming the two scores.
12. Olby NJ, Moore SA, Brisson B, Fenn J, Flegel T, Kortz G, Lewis M, Lewis M, Tipold A. "ACVIM consensus statement on diagnosis and management of acute canine thoracolumbar intervertebral disc extrusion." Journal of Veterinary Internal Medicine. 2022;36(5):1570-1596. doi:10.1111/jvim.16480 — panel of 8 experts; the majority of available literature described observational studies and most recommendations were supported by a low or moderate level of evidence; areas identified as needing further study include "better understanding of the ideal timing for surgical decompression."

This article is educational and does not replace veterinary advice. A tracking protocol is a way of bringing better information to an appointment, not a substitute for one. Any limp that is sudden, non-weight-bearing, accompanied by heat and swelling, or accompanied by weakness or incoordination needs to be seen rather than measured.