The SAGE Trial Correction, And What It Left Alone

The SAGE trial correction: the short answer

In March 2026, six months after publishing the SAGE yoga falls trial, The Lancet Healthy Longevity issued a formal correction to it. So when was the SAGE trial corrected, and what did the Lancet correction actually change? It changed five things, including how the programme is described and how many of the four instructors specialised in the Iyengar style. It left untouched the one sentence that travels furthest, which is the Findings line in the summary.

The whole article is below in short form.

• The correction of 24 March 2026 made five changes, all quoted here.
• The Findings sentence in the summary was not one of them.
• The study’s own appendix still contradicts the corrected paper.
• When the teacher point entered the record, with every date.
• The trial was funded with $930,103.20 for three things. One was not done.
• The 2020 protocol planned for about eight teachers. The paper describes four.
• It was not Iyengar yoga. It was yoga, and as yoga it did some good.
• What we would rather see happen, and what we are not saying.

What was corrected, in the journal’s own words

The correction is titled “Correction to Lancet Healthy Longevity 2025; 6: 100749”. It was published on 24 March 2026, it carries its own citation as The Lancet Healthy Longevity 2026; 7(3): 100825, and PubMed records its publication type as a published erratum. It closes with the line that these corrections were made to the online version as of 24 March 2026.

It makes five changes. They are set out below one at a time, because the difference between the old wording and the new one is the whole story, and it is easy to miss when the five are run together in a paragraph.

One note before you read them. A correction notice does not reprint the sentence it is replacing. It says what the sentence “should have read”. So the first line of each block below describes what the paper said before in substance, which in every case was the same thing: the programme was called Iyengar, with nothing attached to soften it. The second line is the journal’s own corrected wording, quoted.

Change one: how the programme is described in the summary

BEFORE	The programme was described as Iyengar, with no qualifier.
NOW	“supervised yoga exercise programme based on some elements of the Iyengar style”
WHAT THAT CONCEDES	The paper no longer describes the programme as Iyengar. It says the programme borrowed parts of one.

Change two: how the programme is described in the introduction

BEFORE	The same unqualified description, in the last paragraph of the introduction.
NOW	“supervised yoga exercise programme based on some elements of the Iyengar style”
WHAT THAT CONCEDES	The same concession, made in a second place.

Change three: a brand new sentence about what was left out

BEFORE	Nothing. The paper did not say what the programme left out.
NOW	The programme used standing poses common in Iyengar practice “but it did not include other traditional elements such as hands-on teacher adjustments, due to the online programme delivery, or individualised sequencing that accounts for psychological and philosophical factors influencing balance”
WHAT THAT CONCEDES	Two of the three things we said were missing are now named as missing, in the paper’s own words.

Change four: who taught it

BEFORE	The paper did not divide the four instructors by style.
NOW	“and two of the four instructors specialised in the Iyengar style of yoga“
WHAT THAT CONCEDES	The two-of-four arithmetic is now the publisher’s, not just ours.

Change five: how the programme is described in the discussion

BEFORE	The discussion called it an Iyengar programme.
NOW	“the SAGE yoga programme’s foundation in some elements of the Iyengar style”
WHAT THAT CONCEDES	The same concession again, in the part of the paper where the authors explain their own result.

Read those five together and a picture emerges. The paper now describes a programme that borrowed standing poses from a yoga style, was taught partly by people who specialise in that style, and left out at least two of the things that define how that style is taught.

What the correction left alone

Here is the same paper, part by part.

Part of the paper	Corrected in March 2026?
The title	No
Methods, in the summary	Yes
Findings, in the summary	No
Interpretation, in the summary	No
The research in context panel	No
Procedures, on the instructors	Yes
Procedures, on what was left out	Yes
Discussion, the third explanation	Yes
The supplementary appendix	No

The Findings sentence is the one that matters, and this is not a subtle point about academic etiquette. It is about where words end up.

A summary is what a database copies. It is what appears in a search result. It is what an AI assistant reads when somebody asks it whether yoga causes falls in older people. The full text sits behind a page most readers never open, and the appendix sits behind a download almost nobody opens.

So one page of that paper now has four sentences about what the programme was, three of them carefully corrected, and the fourth one is the one that travels furthest.

In our view, asking a journal to make one sentence agree with another sentence on the same page is a small ask, and a reasonable one. It does not require anybody to agree with us about yoga.

When the teacher point entered the record

One of the five changes has no obvious parent. Most of them answer complaints that were already on the table. The teacher sentence does not.

Here is the sequence.

Late 2025. The Australian Iyengar Yoga Association put its concerns about the trial to The Lancet and asked for the reporting to be revised. Its concerns were about how the programme was designed and delivered. The teachers were not among them.

23 January 2026. Flametree emailed its subscriber list and pointed readers to a blog post. It raised, in public, that no more than half of the teachers involved were Iyengar certified. As far as we can establish, and we have looked, that is the first time the point was raised anywhere.

Why Chris went looking at the teachers in the first place. Chris Lalor is a Certified Senior Iyengar Yoga Teacher, a member of the board of the Australian Iyengar Yoga Association, and a moderator of the independent panels that certify Iyengar teachers across Australia. As Iyengar teachers and students understand it, if you are not taught by a teacher certified and trained in the method, you are not getting Iyengar yoga. The certification is not a badge. It is the thing that makes the description true.

And in our experience, and in the experience of the students we teach, what you get from Iyengar yoga is substantially different from what you get from other styles. The precision, the props, the sequencing and the hands-on correction are not decoration. They are the method.

That difference also shows up in the research. Iyengar yoga has been studied in trials on a range of conditions around the world, and those trials have found benefits. The same research team’s own 2013 pilot used Iyengar yoga, with a certified teacher, in person, and found improvements in balance and mobility. So the method had something to offer this question before this trial started.

Then, on 24 March 2026. The Correction was published. It contains the sentence “and two of the four instructors specialised in the Iyengar style of yoga”.

What we say about that, and what we do not

We are not going to tell you why that sentence was added. We do not know, nobody has told us, and a guess dressed up as a finding would be exactly the thing we have spent two articles objecting to.

What we will say is what the dates show. The question about the teachers was not part of what anybody was asking for in 2025. It was raised in January 2026. It is in the Correction published in March 2026.

In our opinion, on those facts, raising it publicly was worth doing. That is an opinion, it is ours, and you can check the two published dates in it for yourself.

The part of our own email we got wrong

Since we are asking you to check our dates, you should check our wording too.

That January email listed the teacher problem like this: “In the fine print, reports show more than half of the teachers involved were Iyengar teachers, and possibly less.”

A word fell out. It should have read no more than half. As written, the sentence says close to the opposite of what was meant, and the “and possibly less” at the end is the only thing that gives the game away.

We are pointing it out because the whole argument in these articles is that wording matters, that a summary is what travels, and that people should fix a sentence when it does not say what the evidence says. That applies to us as well, and it costs nothing to say so.

The substance was right, and it is the wording the published record now uses: of the four teachers the trial named, no more than two held current Certified Iyengar Yoga Teacher credentials.

The study’s appendix now says something different from the study

This is the cleanest point in the whole argument, because it requires nobody to take our word for anything.

The supplementary appendix was not revised in March 2026. It is still published on the article page, and the corrected paper still refers readers to it for the sequence of poses.

Supplementary Table 1 of that appendix describes the yoga type as “Iyengar yoga, predominantly standing postures”. Under instructor qualifications it says the programme was “delivered by experienced, certified Iyengar yoga instructors”. It gives that same line for both groups.

So the corrected paper says two of the four instructors specialised in the Iyengar style. The appendix hanging off the same web page says all of them were certified Iyengar yoga instructors.

Both cannot be right.

A note on the sequence itself. We took our own copy of the published pose sequence early on. That turned out to be worth doing, and it is how we were able to look at the sequence rather than at a description of it. In our view, as teachers of the method, it is not an Iyengar sequence.

We want to be careful about what we are claiming here. We are not saying anybody did anything improper. We are saying that two documents published together now describe the same thing differently, that anybody can open both and see it, and that the appendix is the one that was left behind.

And here is the claim we do make, in the exact form we make it. Of the four teachers the trial named, no more than two held current Certified Iyengar Yoga Teacher credentials. That is precise about the four who were named, it claims nothing about any teacher who was not named, and nobody has disputed it. As of March 2026 the arithmetic in it is the publisher’s as well as ours.

Because the published record does not say which teacher taught which student, there is no way to know how much of the teaching came from a certified teacher.

That is not a small gap, and it is worth saying why. Reporting guidelines for this kind of trial ask authors to say who delivered the intervention, with what training, and to describe the intervention in enough detail for somebody else to repeat it. The reason is replication. A trial earns its authority by being repeatable, and a programme can only be repeated if the record says what was actually delivered and by whom.

Here it does not. So the portion of the teaching that came from a teacher certified in the method cannot be identified, cannot be repeated, and cannot be tested by anybody else. Which means no conclusion specific to Iyengar yoga can be drawn from this trial in either direction, including a favourable one. We are not saying the teaching was poor. We are saying the record does not allow anybody to find out. See more on this below.

What the trial was funded to do

The trial was funded by the National Health and Medical Research Council, under application APP1163067. The award is on GrantConnect, the Commonwealth’s public register of grants, as GA29797. The recipient is the University of Sydney and the term ran from 1 January 2019 to 31 December 2022.

The value is $930,103.20. One funder, all of it Commonwealth money. There is no private funder on any public record.

We state that figure as a fact and we attach no adjective to it. We make no suggestion at all about how the money was spent. A trial of 700 people over several years is expensive to run, and that is a normal cost for one.

What is worth reading is the grant’s own title: “Preventing falls in older age with yoga-based exercise: effectiveness, cost-effectiveness and implementation.”

Three things. Effectiveness, cost-effectiveness, and implementation. The published paper reports the first. It states that the second was not conducted.

The economic evaluation that was planned and not conducted

The published protocol, which came out in BMJ Open Sport and Exercise Medicine in 2020, describes the cost-effectiveness work at length.

It set out a health and community care funder perspective. It planned to count falls prevented and quality adjusted life years gained, to calculate an incremental cost per quality adjusted life year, to plot a cost-effectiveness plane, to bootstrap the confidence intervals, to run one-way and probabilistic sensitivity analyses, and to build a cost-effectiveness acceptability curve. It planned to collect intervention cost data including staff costs.

In plain English, an economic evaluation answers one question: should a health department pay for this?

The paper states that this analysis was not conducted.

We want to give the authors their due here. They said so. They did not bury it. A paper that discloses what it did not do is behaving properly, and it would be both wrong and self-defeating for us to suggest otherwise.

But the question of whether a health department should fund yoga for falls prevention is the question a lot of people thought this trial had answered. It is not the question the published paper answers.

Why that missing piece is the expensive one

A quality adjusted life year is the unit health economists use to put one number on how long somebody lives and how well they live while they are doing it. It is the unit that picks up the things falls take away and money does not measure: independence, walking, confidence, staying in your own house.

That distinction is not small, and here is the scale of it.

Fall-related injuries cost the Australian health system about $5.4 billion a year. The Australian Institute of Health and Welfare also counts what falls take from people directly, and for Australians aged 65 and over that comes to 48,034 healthy years every year. Value a year of healthy life the way the Australian Government values one when it assesses its own regulations, and the second number is worth several times the first.

So the economic evaluation was not a tidy appendix to the main result. It was the side of the ledger where most of the value sits, and it is the side nobody counted.

Add to that the trial’s own subgroup finding, where falls went down 39% among the least active participants. That is the group with the most falls to avoid and the most independence to lose. Nobody costed it, because the analysis that would have costed it was not conducted.

We are not saying what the answer would have been. Nobody knows, and that is the point. The grant was for three things, the paper reports the first, and the second was funded and not run.

What a protocol is, and how binding it is

A protocol is the plan. It is written and published before the trial runs, and it says what will be done, to whom, by whom, and how the results will be analysed. Publishing it in advance is what stops any research team from looking at the results first and then deciding what it was they had been testing.

The SAGE protocol was published in 2020. It was approved by the University of Sydney’s ethics committee, registered in advance on the Australian trials registry, endorsed by a national clinical trials network, and backed by a statistical analysis plan lodged publicly before the analysis began.

How binding is it? Professionally and ethically, not criminally. The obligation comes from the ethics approval, from the reporting standard the paper says it follows, from the Australian Code for the Responsible Conduct of Research, and from the policies the protocol paper binds itself to. A research team can change its plan, and teams often do for good reasons. What is expected is that the change is recorded and reported.

And the paper does report several of its own changes. It reports a change in the statistical model it used. It reports recalculating how many people it needed. It reports that one analysis was not planned in advance. It reports that the economic evaluation was not conducted. It also says that the protocol and subsequent amendments were approved by the ethics committee, so there is an approved trail of changes. That trail is simply not published.

Anybody who told you the changes were hidden would be wrong, and we are not going to tell you that.

The teacher numbers, as a question rather than a claim

This is the part where we are going to lay out the published numbers and then stop, because we do not know the answer and neither does anybody reading this.

Here is what the published documents say.

The 2020 protocol planned for “approximately eight yoga teachers”. That is in its table describing who would provide the intervention.

The 2025 paper describes four instructors.

The paper says the intervention ran through 26 yoga classes, twice weekly, over 40 weeks. That is 52 classes a week in total. Shared between four instructors it works out at about 13 classes a week each, every week, for 40 weeks. The comparison group’s workshops had to be run as well.

Thirteen classes a week is a heavy teaching load. It is not an impossible one, and we are not going to pretend that it is. Plenty of teachers have done more.

What we would say is that anybody who knows how yoga teaching actually works, and how it is rostered, reads those three published facts and has a question. The published record does not answer it. That is all we are saying, and we are saying it as a question because a question is all we have.

Two other things the protocol promised

A balanced allocation by teacher. The protocol said randomisation would use stratification to ensure balance by yoga studio and instructor. That is the standard way of making sure differences between teachers do not end up contaminating the result. That sentence does not appear in the 2025 paper, and no replacement for it is described.

An analysis of the people who actually did the programme. The protocol planned a secondary analysis to estimate what happened to people who attended at least 80% of the supervised yoga sessions. This is the analysis that would have told you what the programme did for people who actually did it, as opposed to people who were offered it and did not go. It does not appear in the paper.

That second one is worth pausing on. The headline finding counts everybody who was offered the programme, including those who barely attended. That is the correct way to run the main analysis, and nobody disputes it. But it does mean the headline is a statement about being offered the programme, not a statement about doing it.

What the reporting standards ask for

There are two agreed checklists for reporting what an intervention in a trial actually was. One is called TIDieR, published in the BMJ in 2014. The other is CLARIFY, published in BMJ Open in 2021 and written specifically for exercise trials.

Both ask for who delivered the intervention, what training those people had, how much of it each participant received, and how far it was delivered as planned. CLARIFY asks specifically for the style of exercise and the instructor’s training in that style.

The published SAGE reporting does not supply several of those items.

In our view this is the point most worth raising with a journal, and it is the reason we keep coming back to it. It is a point about reporting, not about yoga. It does not require anybody to agree with us that the programme was or was not Iyengar yoga. It only requires the reader to compare a published paper with a published checklist.

What we would rather see happen

We have spent two articles being careful about what this trial was and was not. It would be easy to read that as wanting somebody told off. That is not it, and it is worth saying what we would actually like instead.

One. Drop the label, and keep the finding

The correction has already gone most of the way. The paper no longer claims the programme was Iyengar yoga. It says the programme was based on some elements of the style, and it names what was left out.

The step still missing is small. Make the Findings sentence in the summary say the same thing as the Methods sentence on the same page. Nothing about the result changes. What changes is that a defined method, with a training and a certification behind it, stops carrying a result it did not produce.

Two. Pick up the good finding, because it belongs to the researchers

Among the least physically active participants, falls were 39% lower. That is in their data, from their trial, and the test of whether the pattern was real came out at p equals 0.0009, which is a far more secure number than the headline.

It also points at exactly the group a falls prevention programme is for. Nobody has to concede a thing to claim it. In our view it is a better finding than the one that travelled around the world, and at the moment nobody has read it.

Three. Finish the part that was funded

The economic evaluation was planned in detail in the published protocol and was not conducted. Running it now would answer the question the public actually has, which is whether any of this is worth paying for.

And the part we will keep saying

It was not Iyengar yoga as it is normally taught, and the published record should say so in every sentence rather than most of them.

But it was yoga. Standing poses, twice a week, with teachers, for a year. And as yoga it produced a result worth having in the group that needed it most. Those two things sit together perfectly well, and we would rather the whole of it was on the record than half.

None of this asks anybody to say they were wrong about yoga. It asks for one label to be corrected and one good result to be picked up. That looks to us like a better position than the one on offer now, which is a headline that has to keep being defended and a finding nobody has seen.

What we are not saying

We are going to set the limits out plainly, because a post like this is worth nothing if the reader cannot tell where the evidence stops.

We make no accusation about the integrity of the trial or of anybody who worked on it, and we do not speculate about why it was designed as it was. Our argument is with how the study was built and how it was reported, and it stops there.

We are not saying the arithmetic is wrong. We recomputed the headline from the paper’s own table and it reproduces. There is no error to attack.

We are not saying the trial harmed anybody. Falls that caused injury did not increase. That is what the evidence shows, and we say that and stop.

And we are not asking for the paper to be withdrawn. We are asking for one sentence in the summary to agree with the sentence three lines above it, and for an appendix to say the same thing as the paper it is attached to.

Why any of this matters if you are not a researcher

Because of where the words end up.

Search engines and AI assistants build their answers out of whatever has been published. Nobody controls what they produce, and no author or publisher is under any obligation to chase down a summary that has already been copied somewhere else.

So the uncorrected sentences are the ones travelling. Nobody is going to go back and fix what an AI assistant says about this trial, or what a news article from October 2025 still says, unless the source documents are made to agree with each other in the first place.

That is the whole reason we keep writing about it.

If you are an Iyengar yoga teacher being asked about this study by a student, this post is here to be sent. If you are somebody over 60 who read a headline and wondered what to make of it, our companion post on what the numbers actually mean is the one to read, and our post on what the trial did and did not test goes into the teaching side of it.

Where to start with us

Flametree Yoga Studio teaches Iyengar yoga across the greater Darwin region and LIVE online across Australia, with small classes, props and structured teaching.

The easiest way in is the half price 10 class pass with 2 classes free.

Frequently asked questions

What did The Lancet correct about the SAGE yoga trial?

On 24 March 2026 The Lancet Healthy Longevity published a formal correction to the SAGE trial paper. It made five changes. The programme is now described as “based on some elements of the Iyengar style” in the methods summary, the introduction and the discussion. The paper now says two of the four instructors specialised in that style. And a new sentence records that the programme did not include hands-on teacher adjustments or individualised sequencing.

Does the correction change the SAGE trial’s result?

No. The correction changed how the programme is described, not what was found. The reported result, that the yoga group had about a third more falls, is unchanged. What the correction does change is the answer to the question of what exactly was being tested.

What does the SAGE study’s appendix say about the teachers?

The supplementary appendix was not revised in March 2026. Supplementary Table 1 still describes the yoga type as “Iyengar yoga, predominantly standing postures”, and under instructor qualifications it says the programme was “delivered by experienced, certified Iyengar yoga instructors”, for both groups. The corrected paper says two of the four instructors specialised in the style. Anybody can open both documents from the same page and compare them.

Who paid for the SAGE yoga trial, and how much was it?

The National Health and Medical Research Council, under application APP1163067. The award appears on GrantConnect, the Commonwealth’s public grants register, as GA29797, with a value of $930,103.20, running from January 2019 to December 2022. The recipient was the University of Sydney. All of it is Commonwealth money and there is no private funder on any public record.

What is a trial protocol, and does it have to be followed?

A protocol is the plan for a trial, published before it runs, setting out what will be done and how the results will be analysed. Publishing it in advance stops a team deciding after the fact what they were testing. It is binding professionally and ethically rather than criminally, through the ethics approval, the reporting standards and the national research code. Plans can be changed, and often are for good reasons. The expectation is that changes are recorded and reported.

How many yoga teachers taught the SAGE trial?

The published record gives two different answers at two different times. The 2020 protocol planned for approximately eight yoga teachers. The 2025 paper describes four instructors. The paper also reports 26 classes running twice weekly over 40 weeks, which works out at about 13 classes a week for each of four instructors. That is a heavy teaching load. It is not an impossible one. We do not know what happened in between, the published record does not say, and we make no claim about it.

When was the SAGE trial corrected?

The correction was published on 24 March 2026, six months after the trial itself appeared in September 2025. It carries its own citation, The Lancet Healthy Longevity 2026; 7(3): 100825, and PubMed records it as a published erratum. It makes five changes and says the corrections were made to the online version as of 24 March 2026.

Who first raised the question about the SAGE trial’s teachers?

The Australian Iyengar Yoga Association put its concerns about the trial to The Lancet in late 2025, but those concerns were about how the programme was designed and delivered, not about the teachers. Flametree raised the teacher point publicly on 23 January 2026, in an email to its subscribers and in a blog post, saying that no more than half of the teachers involved were Iyengar certified. As far as we can establish, that is the first time it appeared anywhere in public. The correction of 24 March 2026 states that two of the four instructors specialised in the Iyengar style of yoga.

What would fix the SAGE trial’s published record?

Three things, and none of them requires anybody to change their view of yoga. Make the Findings sentence in the summary describe the programme the same way the corrected Methods sentence on the same page already does. Revise the supplementary appendix, which still describes both groups as taught by certified Iyengar yoga instructors. And conduct the economic evaluation that the published protocol planned in detail and the paper says was not conducted.

What is an economic evaluation in a clinical trial?

It is the part that works out whether a treatment is worth paying for, usually by comparing what it costs against the health it produces, measured in quality adjusted life years. The SAGE protocol described one in detail, including collecting the cost of running the programme. The published paper states that it was not conducted.

More on the SAGE trial from Flametree

This is one of four articles. Does yoga cause falls? What the numbers actually say explains the statistics for an ordinary reader. Falls and yoga studies reviews the in-person research that came before it. Yoga for seniors: what the SAGE trial did not test looks at the programme itself.

Sources

1. “Correction to Lancet Healthy Longevity 2025; 6: 100749.” The Lancet Healthy Longevity 2026; 7(3): 100825. Published 24 March 2026. Indexed on PubMed as a published erratum.
2. Oliveira JS, Sherrington C, Lord SR, Tiedemann A, et al. “The effect of an Iyengar yoga-based exercise programme versus a seated yoga relaxation programme on falls in people aged 60 years and older (SAGE).” The Lancet Healthy Longevity 2025; 6: 100749, and its supplementary appendix.
3. The published SAGE protocol, BMJ Open Sport and Exercise Medicine 2020;6:e000878.
4. GrantConnect, Grant Award GA29797, Commonwealth of Australia.
5. Hoffmann TC, Glasziou PP, Boutron I, et al. “Better reporting of interventions: template for intervention description and replication (TIDieR) checklist and guide.” BMJ 2014;348:g1687.
6. The CLARIFY reporting standard for exercise interventions, BMJ Open 2021.

Flametree Yoga Studio teaches Iyengar yoga across the greater Darwin region and LIVE online across Australia. This article is general information and is not medical advice. It discusses published research documents and quotes them as published.

\pagebreak