The Medication Effect Score escape hatch does not exist, but the prescription counts are recoverable exactly, and a bounded ITT reconstruction cuts the payer headline from -41% to -19%
Numbers-honesty brief for an investing calibration. Nothing here is medical advice or a clinical recommendation.
The question
Verbatim: "Can an ITT-equivalent be reconstructed for the Virta DRCP medication-reduction outcomes (which the paper reports for completers only), e.g. from the Medication Effect Score mixed-effects model on n=169 — or does that require unpublished data?"
ITT = intention-to-treat. MES = Medication Effect Score, a continuous summary of glucose-lowering drug burden. CCI = Continuous Care Intervention, the Virta trial arm. This is follow-up #1 from [[2026-07-30-virta-drcp-5yr-itt-remission-rate]], and it is the one remaining gap in the payer-savings arm of the behavioral-substitution bear vector.
The answer has three parts. The hoped-for route is not available. A different and better-grounded route is available. And the reconstruction, once done, moves the payer number by more than the remission correction did.
What we already know (from the vault)
- [[2026-07-30-virta-drcp-5yr-itt-remission-rate]] read the 2024 five-year paper's full text directly and recorded the medication figures with their basis, verbatim: "Among completers only, we assessed changes in the proportion of participants prescribed diabetes and lipid-lowering medications." The completer figures: any diabetes medication 85.2% to 71.3%, non-metformin 55.7% to 32.8%, insulin 26.2% to 13.1%, sulfonylureas 27.0% to 4.9%. That brief's open follow-up proposed the MES model on n=169 as a possible ITT route, parenthetically giving "1.5 to 0.9, a 40% reduction."
- [[2026-10-12-lookahead-direct-remission-itt-vs-completer]], filed earlier tonight, fixed the denominator tiers reused here: 262 enrolled in the CCI arm, 194 completed two years and were approached, 169 consented to the extension, 122 remained to five years, 120 had five-year HbA1c.
- [[2026-10-12-glp1-discontinuation-cohort-past-18-months]] is the third sibling brief closing the same thread's remaining open vector, the behavioral-substitution-durability question past 18 months.
- [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]] resolved the fate of the missing 93 (68 never approached, being the cumulative two-year attrition; 25 declined) and recommended 9.2% on 262 as the payer-facing remission anchor. It also recorded that the 68 were directionally worse-prognosis but not significantly so, and it explicitly left the medication gap open: "It does not rescue the payer-savings arm, whose medication-reduction figures remain completer-only with no published ITT."
- [[2026-05-25-virta-health-5yr-cohort-reversal-durability]] is the grandparent, which first flagged completer framing as the laundering surface.
- The vault holds no prior coverage of the Medication Effect Score as an instrument. A QMD sweep returned only the parent brief's own follow-up sentence. Everything below is new to the vault.
What the web says
Three primary sources were read in full this run: the one-year paper, the two-year paper, and the five-year paper's PubMed record. The five-year paper's own full text was not reachable; see Sources.
- The MES figure cannot be verified to exist in the published record, on five independent checks. (1) The PubMed abstract for PMID 39433217, fetched this run via NCBI efetch, enumerates exactly what the mixed-effects models covered, verbatim: "Changes in body mass, glycemia, and cardiometabolic markers from baseline to 5 years were assessed using linear mixed-effects models." Its results list body mass, HbA1c, triglycerides, HDL-C, inflammatory markers, LDL-C and total cholesterol. No MES. (2) Neither predecessor paper uses the instrument at all: zero hits for "effect score" or the token "MES" across the full texts of Hallberg 2018 and Athinarayanan 2019. (3) A Europe PMC full-text search for "medication effect score" restricted to five Virta authors returns 0 hits. (4) One web search across Virta's own blog and press surface plus secondary coverage surfaces no MES. (5) The paper is not in PMC and both the publisher HTML and the publisher PDF return HTTP 403.
- And in the parent brief, the MES figure is the single number carrying no verbatim quote. Every other figure in [[2026-07-30-virta-drcp-5yr-itt-remission-rate]] sits in its evidence section with a quoted lead-in. The MES line sits in a parenthetical inside an open follow-up, with no quote, and the assertion that it "is ITT" carries no supporting sentence from the paper. [inferred] The 1.5-to-0.9 pair is plausible in shape for a MES and most likely a genuine read of a table row, but the "on n=169, mixed-effects, therefore ITT" attribution is the prior run's inference rather than a quoted fact.
- The decisive methodological finding: medication was deliberately excluded from the mixed-effects ITT analysis at two years, and the same design choice carries to five years. The two-year paper's main results table is labelled ITT in its own footnote, verbatim: "Ns for continuous care intervention = 262 ... estimates reported were obtained from linear mixed-effects models which provide adjusted means and mean changes ... This maximum likelihood-based approach uses all available repeated data, resulting in an intent-to-treat analysis." I extracted that table's complete row list: 40 outcome rows spanning glycemic, body-composition, cardiovascular, liver, kidney, thyroid and inflammatory markers. Zero medication rows. The medication rows appear only in the baseline-characteristics table.
- The reason is structural, not editorial. Two-year paper, methods, verbatim: "We also examined changes in participants' diabetes medication use. First, we compared the rates of diabetes medication use within groups from baseline to 2 years using McNemar's test with continuity correction when appropriate. Next, we calculated the proportion of participants in each group with each diabetes medication class eliminated, reduced, not changed, increased, or added. Paired t-tests were used to assess within-group changes in insulin dosages." McNemar and paired t-tests require paired observations by construction, so they can only be computed on people present at both timepoints. The five-year abstract confirms the same choice: "diabetes status assessed using McNemars' test."
- So the Virta cohort has never published an ITT medication analysis at any timepoint. The two-year headline "any glycemic control medication (excluding metformin) among CCI participants declined (from 55.7 to 26.8%)" is a completer figure on the 194, matching that paper's own 194-column baseline of 55.67%. The two-year paper did run a parallel completers-only set for its continuous outcomes and put it in a supplement ("We conducted a second set of analyses with 2-year completers only. Results of the completers-only analyses appear in Supplementary Table 3"), which means the ITT-versus-completer gap is quantified in this cohort for biomarkers and never for medication.
- The paper applies two contradictory missing-data assumptions in the same table, and medication sits on the generous side of neither. Per [[2026-07-30-virta-drcp-5yr-itt-remission-rate]], continuous outcomes use maximum likelihood on n=169, which is valid under missing-at-random, while the binary diabetes-status endpoints use worst-observation-carried-forward, which is an explicitly missing-not-at-random convention. The authors therefore do not themselves believe MAR holds for their dropouts on the outcomes they judged dropout-sensitive.
- Baseline medication prevalence is published for every subgroup needed, which is what makes a reconstruction possible. Athinarayanan 2019 Table 1 is structured as three CCI columns, All n=262 / Completers-with-data n=194 / Dropout-or-missing n=68, and carries every drug class. Non-metformin: 56.87% / 55.67% / 60.29%. Insulin: 29.77% / 29.38% / 30.88%. Sulfonylurea: 23.66% / 25.77% / 17.65%. Metformin: 71.37% / 71.65% / 70.59%. GLP-1: 13.36% / 13.40% / 13.24%.
- One sourced anchor that makes baseline-carry-forward a defensible rather than alarmist assumption. Two-year paper, verbatim: "The UC group had no changes in these parameters (except uric acid and anion gap) or diabetes medication use." The trial's own usual-care arm did not escalate medication over two years, so carrying a dropout at baseline burden is close to the observed no-intervention trajectory rather than a pessimistic invention. The usual-care arm was not followed to five years, so this anchor covers two years only.
- Context on effect decay, both figures model-based and ITT-flavoured. The two-year ITT mixed-effects HbA1c change on N=262 is -1.3 at one year and -0.9 at two years from a baseline of 7.7 (two-year paper, adjusted-changes table). The five-year abstract reports -0.3. So the glycemic effect fell to roughly a third of its two-year size even though the five-year figure sits on the more favourable 169 denominator.
My reconstruction
Two layers. Layer 1 recovers the raw numerators from the published percentages, which is close to certain. Layer 2 builds the bounded ITT-equivalent, which rests on labelled assumptions. Everything in this section is RDCO-derived and appears in no paper.
Layer 1: the completer counts are recoverable exactly, and the denominator is 122
All four published medication percentage pairs resolve to integers over n=122, and over no other candidate denominator. Residuals are the gap between the implied count and the nearest integer, in percentage points:
| Class | Baseline % | implies | 5-year % | implies | max residual |
|---|---|---|---|---|---|
| Any diabetes medication | 85.2 | 104/122 | 71.3 | 87/122 | 0.056 |
| Non-metformin | 55.7 | 68/122 | 32.8 | 40/122 | 0.046 |
| Insulin | 26.2 | 32/122 | 13.1 | 16/122 | 0.036 |
| Sulfonylurea | 27.0 | 33/122 | 4.9 | 6/122 | 0.060 |
All eight land within 0.06 pp of an integer. Tested against n=120, seven of eight fail (85.2% would be 85.0% or 86.7%, neither of which rounds to 85.2). Tested against n=194, seven of eight fail. So the medication analysis runs on 122, the count who "remained in the study," not the 120 with five-year HbA1c that the remission analysis uses. The closure check holds with no residual: 122 completers + 47 extension dropouts + 25 decliners + 68 never approached = 262.
Two notes on this recovery. It is independent corroboration that the prior run transcribed the percentages correctly, since a transcription slip would almost certainly break the integer fit. And the baseline non-metformin figure of 55.7% happens to equal the two-year paper's 194-column baseline of 55.67% exactly; I checked whether the 2024 baselines might therefore be on the 194, and they are not, because 2024's insulin baseline of 26.2% and sulfonylurea baseline of 27.0% match neither the 194 (29.38%, 25.77%) nor the 262 (29.77%, 23.66%). The 55.7% match is coincidence.
Layer 2: the bounded ITT-equivalent
Two bounds, no point estimate.
The neutral bound requires no arithmetic, because under missing-at-random the ITT-equivalent equals the published completer figure. If non-completers' year-5 medication distribution matched completers', the completer rate is already the population rate. So the published numbers are the optimistic end of the range, not a neutral midpoint.
The pessimistic bound is baseline-carried-forward, matching the paper's own convention for its dropout-sensitive endpoints. Assumptions, labelled:
- A1. Each non-completer is carried at their baseline medication status. Supported for two years by the usual-care arm's flat medication trajectory; unsupported beyond that.
- A2. The 47 extension dropouts are carried at the 122 completers' baseline rates. This proxy is probably optimistic. The 169 is nowhere split into 122 and 47, so no subgroup baseline exists; and in the one place a completer-versus-dropout split is published (2019, 194 versus 68), dropouts had higher non-metformin use, 60.29% against 55.67%.
- A3. The 68 never approached are carried at their own published 2019 baseline rates. No proxy needed.
- A4. The 25 decliners are carried at the 194-column baseline. They were two-year completers, so the 194 is the right reference class, but they are not separately published.
- A5. "Any diabetes medication including metformin" cannot be taken to the 262 tier at all. The 2019 table publishes the non-metformin composite and metformin separately but not their union, and the overlap is unpublished.
Tier B, the 169 extension consenters:
| Class | 5-year ITT-equivalent | vs baseline | change | relative | published completer relative |
|---|---|---|---|---|---|
| Any diabetes medication | 127/169 = 75.1% | 85.2% | -10.1 pp | -11.8% | -16.3% |
| Non-metformin | 66/169 = 39.1% | 55.7% | -16.7 pp | -29.9% | -41.1% |
| Insulin | 28/169 = 16.6% | 26.2% | -9.6 pp | -36.8% | -50.0% |
| Sulfonylurea | 19/169 = 11.2% | 27.0% | -15.8 pp | -58.4% | -81.9% |
Tier C, the 262 enrolled, against the published 262 baseline:
| Class | numerator build | 5-year ITT-equivalent | vs published 262 baseline | change | relative |
|---|---|---|---|---|---|
| Non-metformin | 40 + 26 + 41 + 14 = 121 | 121/262 = 46.2% | 56.87% | -10.7 pp | -18.8% |
| Insulin | 16 + 12 + 21 + 7 = 56 | 56/262 = 21.4% | 29.77% | -8.4 pp | -28.2% |
| Sulfonylurea | 6 + 13 + 12 + 6 = 37 | 37/262 = 14.1% | 23.66% | -9.5 pp | -40.3% |
Numerator terms in order: five-year completers on the drug, then 47 extension dropouts, then 68 never approached, then 25 decliners.
And the MES, conditional on the unverified figure being real and being a completer mean. If 1.5 and 0.9 are the observed completer means on 122, then baseline-carried-forward gives (122 x 0.9 + 47 x 1.5) / 169 = 1.07 on the 169, a -28.9% reduction, and (122 x 0.9 + 140 x 1.5) / 262 = 1.22 on the 262, a -18.6% reduction, against the -40% the parent brief quoted. If instead 0.9 is already the n=169 model-implied marginal, the 169-tier figure cannot be computed, because that requires the completer-only mean, which is unpublished. The two readings differ by exactly the quantity being estimated, which is why this line is not usable in the thesis file.
What blocks a point estimate with an interval, precisely
Six quantities. The first three are hard blockers and all three are unpublished:
- The SD or SE of MES at baseline and at five years. Without dispersion there is no interval, only a mean. Not published in any reachable source.
- The per-timepoint N for the MES model. Table 1's footnote gives the series for HbA1c, 169 / 165 / ... / 120, but medication ran on a separate completers-only basis, so the MES N series is unknown.
- Whether 0.9 is an observed completer mean or a model-implied marginal on 169. Unresolvable from the figure alone, and it changes the answer by the whole effect.
- Baseline medication status of the 47 extension dropouts as a subgroup. Forces assumption A2.
- Baseline "any diabetes medication including metformin" on the 262. Forces A5 and blocks that row at tier C.
- Any post-dropout medication observation for the 140 non-completers. Non-existent by construction, for anyone including the authors.
Items 1 through 5 are answerable by the authors or by Table 2 of the five-year paper. Item 6 is irreducible, which is why this is permanently a bounding exercise and never an estimate.
Convergences and contradictions
- The question's premise is correct and its proposed solution is not. The medication outcomes are completer-only, as the parent brief said. But the MES mixed-effects route does not exist as described: medication was kept out of the ITT mixed-effects table at two years by an explicit methodological choice I verified row by row, and the five-year abstract shows the same choice. The dispatch's hypothesis that a mixed model on all available observations is already ITT-flavoured is correct in general and is stated verbatim in this cohort's own two-year footnote, so it partly dissolves the question for body mass and HbA1c. It does not dissolve it for medication, because medication was never in that model.
- The backlog note was therefore right for the wrong reason, and the gap is wider than it assumed. It framed this as "unlike the remission rate these have no published ITT figure" at five years. In fact they have no published ITT figure at any timepoint in this cohort's eleven-year publication history, and the two-year paper's much-quoted "55.7 to 26.8%" is a completer figure too. Any model that chained the two-year and five-year medication numbers inherited completer selection twice.
- A genuine convergence against the parent brief's caution, in the paper's favour. The parent brief warned that a payer model built on the completer medication numbers "is implicitly assuming the 47 dropouts behaved like the 120 completers, which the WOCF convention explicitly assumes they did not." That is exactly right, and it is now quantified: assuming the dropouts behaved like completers inflates the non-metformin reduction from -18.8% relative to -41.1% relative, a factor of 2.2.
- The reconstruction is directionally harsher than my own tables show. Three of the labelled assumptions bias the result toward Virta. A2 uses completers' baseline for the dropouts when the only published analogue says dropouts had higher drug burden. The usual-care anchor supporting A1 covers two years, not five, and over five years of untreated type 2 diabetes the plausible drift in drug burden is upward rather than flat. And the sulfonylurea case has a particular wrinkle: the 68 never approached had the lowest baseline sulfonylurea use of any subgroup, 17.65%, so adding them to the denominator dilutes a strong completer result by less than the other classes.
Synthesis for RDCO
The headline for the thesis file is a replacement number, not a caveat. The payer-facing medication reduction is "non-metformin glucose-lowering prescriptions fell from 56.9% to roughly 46% across everyone who started, about a 19% relative reduction over five years," and not "roughly halved." The 41% relative figure that the parent brief flagged as "a fair read of the non-metformin line" is the completer number, and the completer number is identical to the ITT-equivalent only under missing-at-random, which this paper's own authors decline to assume for their dropout-sensitive endpoints. Carry the full range rather than a point: -41% relative under MAR on 122 completers, -30% under baseline-carried-forward on the 169 extension consenters, -19% under baseline-carried-forward on the 262 enrolled. Use the 262 tier for payer economics, consistent with the anchor choice [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]] made for remission, because a payer does not get to re-consent its non-engagers out of the denominator. The insulin line deflates less violently and is the most robust of the four, -50% completer to -28% on 262, which matters because insulin is the expensive line item.
That has a specific effect on the bear case, and it is the same direction as the remission correction but larger in proportion. The remission figure moved from 20% completer to 9.2% on 262, a factor of 2.2. The non-metformin medication reduction moves by a factor of 2.2 as well, from -41% to -19% relative. The payer-savings arm of the behavioral-substitution vector was the one arm still resting on undeflated numbers after the last two briefs, and it now deflates by the same multiple as everything else in this cohort. A program that reduces glucose-lowering prescription prevalence by roughly a fifth across an unselected enrolled population, with no control arm to attribute even that to, is not a mechanism that erodes chronic pharmaceutical revenue. It thins and delays prescriptions, which is the conclusion [[2026-07-30-virta-drcp-5yr-itt-remission-rate]] reached from the remission decay curve, now independently reached from the medication side. The bear vector stays trimmed and this is the third consecutive brief that declines to re-inflate it. One honest counterweight: a 19% relative reduction in non-metformin prescribing sustained for five years is not nothing, and the sulfonylurea result survives deflation better than the others at -40% relative on 262, so the right posture is a small persistent margin drag rather than dismissal.
Do not put the MES figure in the thesis file. It fails verification on five independent checks, it is the one number in the parent brief without a verbatim quote, and the specific claim attached to it, that it is an ITT mixed-effects estimate on n=169, is contradicted by what I could verify about how this cohort analyses medication. It may well be a real table row that the prior run read correctly. But the useful thing to notice is that its "is ITT" label was an inference generated inside our own research chain and then carried forward as if sourced, which is the same failure mode the parent brief caught the paper committing when it wrote "20% of those enrolled." If the 1.5-to-0.9 pair is wanted, it needs Table 2 of the five-year paper in hand, and the specific thing to read off it is not the means but the N and the standard errors.
The reusable gate from this run is about escape hatches. The last two briefs produced gates about denominators: tables are load-bearing and discussion prose is not; re-derive a cited comparator's denominator from its own paper. This one is different. When a completer-only outcome has a tempting model-based sibling that looks like it would rescue it, verify that the outcome is actually inside the model before building on it, by reading the model table's row list rather than its footnote. The footnote here says ITT on 262 and means it; the row list says medication is not in scope. Both are true, they sit a centimetre apart, and only the row list answers the question. That check cost one Europe PMC full-text fetch and a regex. The related discipline: when a figure enters our own chain without a verbatim quote while its siblings all have one, that asymmetry is the signal, and the cheapest moment to catch it is before the next brief treats it as a premise.
Why this is in the vault
This resolves open follow-up #1 of [[2026-07-30-virta-drcp-5yr-itt-remission-rate]] and closes the last undeflated input to the payer-savings arm of the behavioral-substitution bear vector in [[2026-05-21-lilly-glp1-longevity-thesis]]: the medication-reduction figure a payer model should use is roughly -19% relative on the 262 denominator, not the -41% completer figure, and the Medication Effect Score route that was supposed to supply a published ITT number does not exist and should not be cited.
Open follow-ups
- Get Table 2 of the five-year paper and settle the MES question. Three specific reads are wanted: whether a Medication Effect Score row exists at all, its per-timepoint N, and its standard errors. An author email to shaminie@virtahealth.com (the paper's listed corresponding address) or one institutional-network session is the route. The publisher HTML and PDF are both behind a hard automation block and the paper is not in PMC, so this is a test-plan step rather than a desk-research item. Same blocker that [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]] hit.
- Pull the two-year paper's Supplementary Table 3 and compute this cohort's empirical ITT-versus-completer deflation factor. The two-year paper ran both analyses on its continuous outcomes and published the completers-only set in a supplement, so the ratio of completer effect to ITT effect is measurable in this exact cohort for HbA1c, weight and triglycerides. That factor would replace my A1-through-A4 assumptions with an empirically calibrated adjustment for the medication numbers. The supplement is a single PDF, Data_Sheet_1.pdf; the PMC instance URL I tried returned HTML rather than the file, so it needs one more routing attempt. This is the highest-value item on the list.
- Insulin dose, not insulin prevalence. Both papers analyse insulin dosage by paired t-test among insulin users, and dose is closer to what a payer actually spends than prevalence is. The five-year dose figures were not captured in any brief in this thread, and dose has never been ITT'd either. It is also the one medication outcome where a partial responder shows up as a reduction rather than as a binary non-event.
- Does Virta's commercial or real-world evidence report medication reduction on an all-starters denominator? The trial cohort is structurally incapable of answering the payer question without assumptions. NCT05516797 (IGNITE) is the one Virta registration with posted results per [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]], and posted results modules carry a Participant Flow table, which is exactly what the DRCP cohort lacks.
- Is the "any diabetes medication including metformin" union recoverable for the 262? Blocked here by the unpublished overlap between the non-metformin composite and metformin. The one-year paper's Table 1 has the same three-column structure and might carry the union row, which would unblock the only tier-C row I could not compute.
Related
- [[2026-07-30-virta-drcp-5yr-itt-remission-rate]] - parent brief; this closes its open follow-up #1 and corrects the MES line inside it
- [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]] - sibling; supplied the 262/169/122 reconciliation this reconstruction allocates across
- [[2026-10-12-lookahead-direct-remission-itt-vs-completer]] - sibling filed earlier tonight; source of the denominator-tier framework reused here
- [[2026-10-12-glp1-discontinuation-cohort-past-18-months]] - sibling brief closing the behavioral-substitution-durability vector of the same thread
- [[2026-05-25-virta-health-5yr-cohort-reversal-durability]] - grandparent; first flagged completer framing
- [[2026-05-21-lilly-glp1-longevity-thesis]] - the live position this calibrates
- [[2026-05-24-virta-health-payer-contracts-lly-thesis-input]] - the payer-economics arm whose inputs this brief finally deflates
- [[2026-07-18-capital-rx-reverse-lilly-relationship]] - substitution-versus-complement resolution on the payer side
- [[2026-07-18-virta-covered-lives-tracker-data-source]] - scale-side tracker for the same thesis
- [[2026-05-22-tim-ferriss-sami-inkinen-virta-t2d-rowing]] - Inkinen anchor interview
Sources
Primary, fetched and read in full this run:
- Athinarayanan SJ, Adams RN, Hallberg SJ, McKenzie AL, Bhanpuri NH, Campbell WW, Volek JS, Phinney SD, McCarter JP. "Long-Term Effects of a Novel Continuous Remote Care Intervention Including Nutritional Ketosis for the Management of Type 2 Diabetes: A 2-Year Non-randomized Clinical Trial." Front Endocrinol 2019;10:348. PMID 31231311, PMC6561315. Full text via Europe PMC REST: https://www.ebi.ac.uk/europepmc/webservices/rest/PMC6561315/fullTextXML (322 KB). Read directly: statistical-analysis section (McNemar and paired-t medication methods), Table 1 three-column baseline medication rows, Table 2 "Adjusted mean changes over time" complete row list and ITT footnote, abstract.
- Hallberg SJ, McKenzie AL, Williams PT, et al. "Effectiveness and Safety of a Novel Care Model for the Management of Type 2 Diabetes at 1 Year." Diabetes Ther 2018;9(2):583-612. PMID 29417495, PMC6104272. Full text via Europe PMC REST (260 KB). Searched for MES and medication-score terms; zero hits.
- PubMed record for McKenzie AL et al. 2024, PMID 39433217, full structured abstract via NCBI efetch: https://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=pubmed&id=39433217&rettype=abstract - read directly for the mixed-effects outcome enumeration and the McNemar statement.
- Europe PMC search API, query
("medication effect score") AND (AUTH:"Hallberg SJ" OR AUTH:"Athinarayanan SJ" OR AUTH:"McKenzie AL" OR AUTH:"Bhanpuri NH" OR AUTH:"Volek JS")- hitCount 0. - OpenAlex (DOI 10.1016/j.diabres.2024.111898) and Semantic Scholar Graph API - both confirm the publisher page is the only open-access location; no repository or preprint copy exists.
Primary, read in a prior run and relied on here without re-fetch:
- McKenzie AL, Athinarayanan SJ, Van Tieghem MR, Volk BM, Roberts CGP, Adams RN, Volek JS, Phinney SD, Hallberg SJ. "5-Year effects of a novel continuous remote care model with carbohydrate-restricted nutrition therapy including nutritional ketosis in type 2 diabetes: An extension study." Diabetes Res Clin Pract 2024;217:111898. The four medication percentage pairs, the "among completers only" sentence, and the Table 1 footnote denominators are quoted from the direct full-text read recorded in [[2026-07-30-virta-drcp-5yr-itt-remission-rate]]. The integer-recovery test in Layer 1 is independent corroboration that those percentages were transcribed correctly.
Secondary, one search, no figure taken from it:
- Virta Health blog and press surface plus aggregator coverage of the five-year study, via one web search. Checked only for whether a Medication Effect Score figure appears anywhere outside the paper. It does not. Virta's own framing on that surface is "total medications were cut in half," an unsourced completer-flavoured claim that matches none of the published percentage pairs.
Unreachable this run, flagged and not retried:
- Publisher full text, https://www.diabetesresearchclinicalpractice.com/article/S0168-8227(24)00808-8/fulltext and the matching /pdf path - both HTTP 403 with a browser user-agent. Same hard automation block recorded in [[2026-10-11-virta-drcp-68-unapproached-participants-denominator]], which also exhausted headless and headed Playwright on it.
- fatcat / scholar.archive.org release lookup for the DOI - non-JSON response.
- Two-year paper supplementary PDF (Data_Sheet_1.pdf, 791 KB, SM1) via the PMC instance path for PMC6561315 - returned 21 KB of HTML instead of the file. Carried as the highest-value open follow-up.
Own arithmetic (labelled, in no paper):
- Count recovery on n=122: 104, 87, 68, 40, 32, 16, 33, 6; all eight published percentages within 0.06 pp of an integer over 122, versus seven of eight failing over 120 and over 194.
- Closure: 122 + 47 + 25 + 68 = 262.
- Tier B (169), baseline-carried-forward: any 127/169 = 75.1%; non-metformin 66/169 = 39.1%; insulin 28/169 = 16.6%; sulfonylurea 19/169 = 11.2%.
- Tier C (262), baseline-carried-forward: non-metformin 121/262 = 46.2%; insulin 56/262 = 21.4%; sulfonylurea 37/262 = 14.1%.
- Relative changes: completer basis -16.3% / -41.1% / -50.0% / -81.9%; tier B -11.8% / -29.9% / -36.8% / -58.4%; tier C -18.8% / -28.2% / -40.3%.
- Deflation factor on the non-metformin headline: 41.1 / 18.8 = 2.2x.
- MES conditional bounding: (122 x 0.9 + 47 x 1.5) / 169 = 1.067; (122 x 0.9 + 140 x 1.5) / 262 = 1.221.