Not yet independently audited
This page was written from verified sources but has not yet passed the independent citation audit. Treat its claims as provisional.
Open questions — osteoarthritis¶
Last audited: 2026-08-30 · sharpened 2026-08-31 (lorecivivint phase 3, methotrexate, Frydendal, Moseng) · sharpened 2026-10-15 sweep (OQ-2, OQ-3, OQ-11, OQ-16, OQ-23) · sharpened 2026-10-29 (OQ-4, OQ-11, OQ-21)
Questions are tiered by potential to change practice and present-day designability. IDs are stable: revise wording or mark a question answered without renumbering later questions.
Sweep 2026-10-29. Direct date-limited PubMed searches were repeated for OQ-1–OQ-12 and OQ-23. No question was closed. SciELO and LILACS could not be searched (HTTP 403), so this sweep cannot establish transportability across world regions. Search strings are in the sweep manifest.
Tier 1 — would change practice and is designable now¶
OQ-1 — Can a prespecified phenotype select an effective treatment?¶
Question. In adults with symptomatic knee OA, does assigning treatment by a prespecified inflammatory, mechanical-overload or sensitization phenotype improve pain/function more than guideline-based stepped care?
Why open. Six recurring phenotype families have been described, but most were derived retrospectively and do not validate treatment selection (Dell'Isola 2016, PMID 27733199). Sensitization prevalence varies by tool and averages 20% with I²=89%, so an enrichment rule must be fixed before randomization (Previtali 2022, PMID 35356833).
Design now. Multicenter strategy RCT; phenotype algorithm locked before enrollment; compare matched pathway with ordinary stepped care; primary outcome 12-month pain/function; interaction and misclassification analyses prespecified.
OQ-2 — Does large weight loss modify joint structure?¶
Question. Does ≥10–15% intentional weight loss reduce MRI cartilage loss, marrow lesions or joint-space narrowing independently of symptom improvement?
Why open. IDEA linked diet/exercise to weight, load, inflammation, pain and function but not durable cartilage preservation (Messier 2013, PMID 24065013). STEP 9 found -13.7% versus -3.2% weight change and greater WOMAC pain improvement with semaglutide, without resolving structural mediation (Bliddal 2024, PMID 39476339). Sharpened 2026-10-15: TRIUMPH-1 repeated the pattern at larger weight loss — retatrutide 9/12 mg gave about −24% to −25% whole-trial weight change and a WOMAC-pain difference of −1.6/−1.8 points (1–10 scale) in 574 people with knee OA, with no structural endpoint reported in the abstract (Jastreboff 2026, PMID 42814954). A GLP-1 RA meta-analysis adds a competing hypothesis: consistent lean-mass loss (low certainty) that could offset load reduction (Beaudart 2026, PMID 42730869). The ACR 2026 summary now conditionally recommends GLP-1 RA for weight in knee OA with obesity, so the structural question is being overtaken by practice (American College of Rheumatology — "2026 Update of the American College of Rheumatology Recommendations for the Management of Osteoarthritis of the Knee, Hip, and Hand", https://assets.contentstack.io/v3/assets/bltee37abb6b278ab2c/bltb3d12c34020da842/oa-guideline-summary-2026.pdf, accessed 2026-10-15). The design should now add body-composition (lean mass, muscle function) arms to MRI outcomes.
Design now. Factorial or mediation-capable trial with repeated weight, gait/load, inflammatory and MRI outcomes plus post-treatment follow-up.
OQ-3 — Which exercise dose is sufficient and sustainable?¶
Question. What combinations of intensity, volume, supervision and behavioral support maximize maintained benefit per unit burden for knee and hip OA?
Why open. Land exercise improves knee pain by about 12/100 immediately but only 6/100 at 2–6 months after formal treatment, and only 35% of included studies met three core bias safeguards (Fransen 2015, PMID 26405113). Program adherence and dose vary too much to identify a universal optimum (Marriott 2024, PMID 38317328). Sharpened 2026-10-15: a dose-response meta-analysis of isolated resistance training (20 RCTs, n=791, low to very low certainty) estimated nonlinear curves — pain benefit largest near 6,000 repetitions, function peaking near 2,600 moderate-intensity repetitions — and larger effects centre-based than home-based (Núñez-Cortés 2026, PMID 42676087); a 12-month blended smartphone-plus-physiotherapy programme did not beat usual care (Weber 2026, PMID 42373005). Volume–response is now estimable across trials; whether it holds within a person is untested.
Design now. Sequential multiple-assignment trial comparing starting dose and step-up rules, with wearable-confirmed exposure and 24-month maintenance.
Sharpened 2026-10-22. Digital-support trial completion is not exercise adherence. A new network review cannot isolate professional contact from exercise content or comparator intensity (Luo 2026, PMID 42876545). A delivery comparison should hold exercise dose constant and measure performed sessions beyond trial completion.
OQ-4 — Can PRP outperform a credible placebo reproducibly?¶
Question. Does a fully characterized PRP formulation produce a clinically important benefit over saline/sham under allocation concealment and blinding?
Why open. RESTORE found a pain difference of -0.4/10 (95% CI -0.9 to 0.2) and cartilage-volume difference -0.2% (-1.9% to 1.5%) at 12 months (Bennell 2021, PMID 34812863), while PRP-versus-hyaluronate meta-analysis reports favor PRP (Tang 2020, PMID 32912243). Sharpened 2026-10-29: a structural-outcome meta-analysis of 12 RCTs (n=1,200) found limited and inconsistent imaging effects and did not support PRP as a disease-modifying treatment (Pasqualotto 2026, PMID 42918921; abstract). Product and comparator heterogeneity still prevent a class conclusion on symptoms; structure is not a rescue argument.
Design now. Independently funded, multi-batch RCT with product release criteria, saline and no-injection contextual arm, prespecified responder threshold and ≥24-month safety/structure follow-up.
OQ-5 — Who benefits enough from intra-articular corticosteroid to justify exposure?¶
Question. Can clinical or imaging synovitis identify short-term responders without encouraging harmful scheduled repetition?
Why open. Repeated triamcinolone every 12 weeks for two years caused 0.11 mm greater cartilage-thickness loss (95% CI 0.03–0.20) without pain advantage (McAlindon 2017, PMID 28510679). Individual-patient meta-analysis suggests greater baseline pain may modify short-term benefit (van Middelkoop 2016, PMID 26836288).
Design now. One-injection enriched RCT with ultrasound/MRI synovitis, aspiration biomarkers, 2–12-week response and explicit prohibition on automatic reinjection.
OQ-6 — Can a DMOAD improve structure and what patients value?¶
Question. Can a target-active therapy produce replicated structural change plus clinically important pain/function or delayed replacement?
Why open. Sprifermin 100 µg q6mo increased 2-year cartilage thickness 0.05 mm (0.03–0.07) without a significant WOMAC difference (Hochberg 2019, PMID 31593273). Lorecivivint phase 3 OA-11 missed Pain NRS at week 12 (LOR −2.24 vs placebo −2.49; p=0.185) (Yazici 2025, PMID 39808286); OA-10 also missed its primary, with only a post-hoc KL 2 signal (Yazici 2025, PMID 39808288). Low-dose methotrexate in effusion-synovitis knee OA was null for pain and MRI synovitis at 52 weeks (Zhu 2025, PMID 40455462). No approved therapy has established the combined standard (Cho 2021, PMID 34848838).
Design now. Biomarker-enriched phase 2b/3 program with hierarchical symptom and structure outcomes, estimands for rescue/crossover and post-trial durability.
OQ-7 — Can potent analgesia be separated from accelerated joint damage?¶
Question. What monitoring, dose or phenotype rule could preserve analgesic benefit while preventing rapidly progressive OA?
Why open. NGF inhibition shows that pain relief can uncouple protective sensation from structure; pooled phase III analyses characterize peripheral-nerve safety but do not remove joint-safety concern (Brown 2023, PMID 37460782).
Design now. Joint-safety registry nested across trials, standardized adjudication, serial imaging and mechanistic analysis of load/activity change after analgesia.
OQ-8 — What is the optimal arthroplasty timing strategy?¶
Question. Among surgery-eligible patients, does early replacement, time-limited optimized nonsurgical care or a trigger-based strategy produce the best two- and ten-year net benefit?
Why open. TKR plus nonsurgical care improved KOOS4 by 15.8 points (95% CI 10.0–21.5) over nonsurgical care but caused more serious adverse events (24 vs 6); 74% assigned nonsurgical care avoided replacement at 12 months (Skou 2015, PMID 26488691). For already surgery-indicated hip OA, THR beat resistance training by 11.4 Oxford Hip Score points (8.9–14.0) at 6 months, with 21% crossover from training (Frydendal 2024, PMID 39476341). That quantifies incremental surgical benefit; it does not identify the optimal timing strategy over two and ten years.
Design now. Pragmatic strategy trial with crossover estimands, waiting-time harms, employment, persistent pain, complications and revision-adjusted quality of life.
OQ-9 — Can preoperative sensitization be treated to prevent persistent postsurgical pain?¶
Question. Does a mechanism-targeted prehabilitation pathway improve replacement outcomes in sensitization-positive patients?
Why open. Preoperative neuropathic-like pain and central sensitization associate with worse knee-replacement outcomes, but association does not show that modifying sensitization changes prognosis (Wluka 2020, PMID 32791103; Kim 2022, PMID 35626402).
Design now. Screen-then-randomize trial of pain-neuroscience/graded activity/duloxetine-compatible pathway with blinded outcome assessment.
OQ-10 — Which medicine sequence maximizes absolute benefit and minimizes harm?¶
Question. Does topical-first, oral-NSAID-first, duloxetine-first or symptom-mechanism sequencing produce different net outcomes in multimorbid knee OA?
Why open. Networks show topical NSAIDs similar to oral NSAIDs for function and fewer GI adverse events, while duloxetine and tramadol trade small benefit against higher adverse-event burden (Zeng 2021, PMID 34174454; Wang 2015, PMID 26176791; Toupin April 2019, PMID 31132298).
Design now. Large pragmatic sequence trial using absolute responder benefit, treatment discontinuation, serious harm and patient preference.
OQ-11 — Can OA trials measure contextual effects rather than hide them?¶
Question. What proportion of injection/procedure improvement is attributable to product, needling, expectation and clinical encounter?
Why open. Placebo/context effects are large in OA (Zhang 2019, PMID 31621561); no hip injection significantly outperformed saline at 2–4 or 6 months in one network, although all groups improved from baseline (Gazendam 2021, PMID 32829298). Sharpened 2026-10-15: across 179 placebo-controlled non-surgical knee-OA RCTs, 60% of placebo arms improved by Hedges' g ≥0.5; heterogeneity was large (I² 88%; prediction interval −0.31 to 1.60), modality did not significantly modify the placebo response, and baseline pain, funding source, active-arm effect size and region together explained only 33% of the variance (Holtedahl 2026, PMID 42657310). In hand OA, about 70% of the overall effect of topical NSAIDs was attributed to contextual effects (34 RCTs; Wen 2026, PMID 42790728). The variance that matters is mostly unexplained, which favours trials that measure expectation directly rather than more meta-regression.
Design now. Balanced placebo or dismantling design with expectation measurement and credible sham assessment.
Sharpened 2026-10-29. A 12-month sham-controlled genicular-artery embolization RCT (n=58) found no KOOS-pain difference (7.46, −13.63 to 28.56) and no synovitis change, with successful blinding (van Zadelhoff 2026, PMID 42009866; full text, Table 2). A review pooling three sham-controlled GAE RCTs reported SMD −0.23 (−0.71 to +0.25) against large within-group VAS reductions (Lanza 2026, PMID 42208168; abstract). Procedure-level placebo is now measured for embolization; GEKO (ISRCTN50216144) is the larger pending replication, not the first sham test.
OQ-12 — Can guideline disagreement be resolved with absolute-benefit thresholds?¶
Question. Would common thresholds for pain benefit, harms, cost and evidence certainty reconcile recommendations on hyaluronate, PRP, acupuncture and braces?
Why open. ACR/AF, OARSI, NICE and AAOS use different scopes and judgments despite overlapping evidence (Kolasinski 2020, PMID 31908149; Bannuru 2019, PMID 31278997; Brophy 2022, PMID 35383651).
Design now. Cross-panel evidence-to-decision experiment using identical effect estimates and explicit patient-derived thresholds.
Sharpened 2026-10-22. Include VA/DoD 2026 in the cross-panel comparison: its neutral knee-hyaluronate and knee/hip-PRP positions differ from ACR 2026 recommendations against. Compare evidence cut-offs and decision thresholds before attributing disagreement to treatment effects (VA/DoD provider summary, Table 1; ACR summary, accessed 2026-10-22).
Tier 2 — important, but enabling methods or longer horizons are needed¶
OQ-13 — What is “early OA” biologically?¶
Early symptomatic disease can exist before established radiographic OA, but classification, prognosis and treatment responsiveness are not aligned (Wang 2024, PMID 37933434; Runhaar 2024, PMID 38862243). Longitudinal multimodal cohorts must define transitions without circularly using the endpoint as the baseline label.
OQ-14 — Which tissue is the dominant pain generator in an individual?¶
Synovitis and marrow lesions associate with pain at group level, but discordance is wide and cartilage is aneural (Dainese 2022, PMID 34968719; Yusuf 2011, PMID 20829200). A clinically useful attribution tool needs perturbation/response validation, not cross-sectional correlation.
OQ-15 — Can biochemical biomarkers become joint-specific?¶
Blood and urine turnover markers integrate multiple joints and systemic tissues; synovial fluid is local but invasive. Candidate-marker literature has not established a routine diagnostic/prognostic test (Watt 2018, PMID 29107060; Saberi Hosnijeh 2019, PMID 30552966).
OQ-16 — Can imaging change qualify as a surrogate endpoint?¶
MRI is sensitive to cartilage, marrow and synovium, but treatment-induced image change must reliably predict a patient-important benefit before surrogacy (Roemer 2016, PMID 27832771). The necessary intervention-level validation is incomplete. Sharpened 2026-10-15: reliability is no longer the limiting step — pooled test–retest RMS-CV is 2.7–3.0% for cartilage volume/thickness and medial tibiofemoral thickness is the most responsive measure (annualized SRM −0.350) (Jia 2026, PMID 42025941) — but a 2026 network meta-analysis of six intra-articular DMOAD RCTs found no credible structural or symptomatic effect to validate against (Wong 2026, PMID 42753488).
OQ-17 — Which genetic findings are therapeutically causal?¶
GWAS across 826,690 individuals identified 100 independent variants, but association-to-target translation requires cell/tissue context and perturbation (Boer 2021, PMID 34450027). Large-effect rare variants and common small-effect loci may imply different strategies (Styrkarsdottir 2018, PMID 30374069).
OQ-18 — How should multi-joint OA be represented?¶
Most trials enroll one index joint, while people often experience several affected sites and systemic treatment burdens. “Generalized OA” definitions remain inconsistent (Nelson 2014, PMID 24461078).
OQ-19 — What prevents post-traumatic OA after ACL/meniscal injury?¶
ACL injury carries near seven-fold higher OA odds, and reconstruction does not eliminate long-term risk (Webster 2022, PMID 33852440). Prevention trials need decade-scale structural and symptomatic outcomes.
OQ-20 — How can work participation become a treatment endpoint?¶
Qualitative research documents disclosure, adaptation and job-demand effects, but intervention trials rarely measure sustained work participation (Alyousef 2024, PMID 37995059; Agaliotis 2018, PMID 30373994).
OQ-21 — What are equitable arthroplasty thresholds?¶
BMI, age, comorbidity and socioeconomic access are used inconsistently. The unresolved task is to separate modifiable risk reduction from categorical exclusion while measuring harms of delay (Caesar 2022, PMID 33988862; Skou 2015, PMID 26488691). Sharpened 2026-10-29: among 2,035 people referred to Alberta hip/knee clinics, females had lower adjusted odds of definite current TKA willingness (OR 0.67, 0.51–0.88), and willingness mapped onto a surgeon recommendation more strongly in males (RR 5.03, 3.04–8.33) than in females (RR 2.15, 1.68–2.75) (Hawker 2026, PMID 42880894; abstract). Equity work has to measure recommendation behaviour, not only formal BMI cut-offs.
OQ-22 — How should patient acceptable symptom state be used?¶
PASS thresholds translate scores into acceptability but depend on instrument, time, baseline and wording; they should not become rigid eligibility cutoffs (Kiadaliri 2024, PMID 38409279; Kunze 2022, PMID 34958538).
OQ-23 — Is inflammatory-phenotype OA methotrexate-responsive?¶
Question. In knee OA with imaging effusion-synovitis, does a conventional synthetic DMARD (methotrexate or another) reduce pain and synovitis versus placebo?
Why open. Zhu 2025 (n=215; NCT03815448) found no 52-week difference in VAS pain (0.3 mm, −6.7 to 7.3) or effusion-synovitis area (Zhu 2025, PMID 40455462). That closes low-dose methotrexate for this enrichment rule; it does not close every anti-inflammatory strategy, nor hand OA, where Kloppenburg 2025 notes promising inflammation-targeted data (Kloppenburg 2025, PMID 39755397).
Correction and sharpening, 2026-10-15. The "closed-negative" framing of 2026-08-31 omitted a conflicting trial. PROMOTE (UK; n=155; oral methotrexate 10→25 mg/week vs placebo over 12 months; not enriched for synovitis) reduced 6-month knee pain by 0.79 NRS points (95% CI 0.08–1.51) (Kingsbury 2024, PMID 39074374), and its 2026 biological substudy (n=87) found hs-CRP the only marker with a consistent treatment-arm interaction for pain response (Liang 2026, PMID 42413648). So the two methotrexate trials point in opposite directions, and the population that responded was selected by neither imaging synovitis nor hs-CRP. Two other anti-inflammatory knee trials were null without phenotype selection: colchicine (CLOAK, n=120; Samuels 2026, PMID 42679472) and tofacitinib in KL 1–2 (n=100; Fasni 2026, PMID 42760593). The ACR 2026 summary conditionally recommends against methotrexate in knee/hip OA and against colchicine, but conditionally for methotrexate in erosive hand OA, flagged as a change of direction since 2019 (American College of Rheumatology — "2026 Update of the American College of Rheumatology Recommendations for the Management of Osteoarthritis of the Knee, Hip, and Hand", https://assets.contentstack.io/v3/assets/bltee37abb6b278ab2c/bltb3d12c34020da842/oa-guideline-summary-2026.pdf, accessed 2026-10-15). The sharpened question: is a systemic-inflammation marker (hs-CRP), rather than imaging effusion-synovitis, the enrichment variable that predicts methotrexate response in knee OA? Design now: a prospectively hs-CRP-stratified methotrexate-vs-placebo trial with a prespecified interaction test, or an individual-patient-data pooling of PROMOTE and Zhu 2025.
Design now (2026-08-31, superseded in part above). Do not re-run the same methotrexate dose/enrichment. Test a different target (for example a specific cytokine or a locked synovial phenotype) or a different joint. Maintenance rule still applies: keep OQ-23 even if later studies close other anti-inflammatory strategies.
Dots not yet connected¶
These junctions were re-searched in PubMed on 2026-08-30. Several build-time absences were stale, so each row now distinguishes a connection found from the narrower residual question. A residual gap means that the specified design was not established after screening the targeted query results on that date; it is not proof of universal nonexistence.
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| 1 | Semaglutide produces large weight and pain change (PMID 39476339) | MRI phenotypes predict progression | The targeted search returned four records, but no repeated-MRI or load-mediation analysis nested in the semaglutide RCT was established as of 2026-08-30 | Distinguishes symptom relief from structure modification |
| 2 | Sensitization affects ~20% by pooled methods (PMID 35356833) | Exercise adherence has been tested cross-sectionally against sensitization, with no significant association reported (PMID 36397008) | Prespecified sensitization-stratified exercise-dose and maintenance RCT | Tests whether treatment durability is mechanistically patterned rather than merely associated cross-sectionally |
| 3 | Bone-marrow-lesion change after opening-wedge high tibial osteotomy has been associated with better symptoms (PMID 42004451) | Osteotomy redistributes compartment load | Prospective causal mediation of alignment correction → lesion change → pain, with a comparator that separates natural fluctuation | Connects mechanical correction to tissue and symptom pathway |
| 4 | PRP has formulation heterogeneity | Context effects are large (PMID 31621561) | Multi-formulation, saline and no-injection factorial trial | Separates product chemistry from procedure/expectation |
| 5 | Post-GWAS, single-cell and functional analyses now prioritize locus-to-cell-state candidates (PMID 42519341) | Human genetics provides target anchors (PMID 34450027) | Prospective genotype-to-cell-state-to-drug-response study in the same participants | Converts target nomination into treatment-predictive enrichment |
| 6 | Occupational exposure raises knee-OA risk (PMID 21382500) | Exoskeletons/braces can change load | Workplace cluster trial with incident symptoms and structure | Moves from attribution to prevention |
| 7 | Preoperative sensitization predicts poorer outcome (PMID 32791103) | Decision aids improve decision quality (PMID 33599773) | Decision aid that includes individualized persistent-pain mechanism risk | Tests whether mechanism information improves consent and outcomes |
| 8 | Canceled surgery can feel like rejection (PMID 33988862) | A six-week exercise-and-education RCT improved pain while patients awaited arthroplasty (PMID 27233479) | Supported-waiting pathway powered for function, trust, opioid exposure and eventual surgery rather than pain alone | Separates therapeutic optimization from abandonment |
| 9 | Hand OA has dexterity/appearance burdens (PMID 33537932) | The targeted hand-OA digital-rehabilitation/co-design query returned zero records on 2026-08-30 | Hand-specific co-designed digital rehabilitation trial | Tests site-specific patient priorities |
| 10 | Clinicians report managing expectations when recommending intra-articular therapy (PMID 33166060) | Patient treatment preferences are measurable (PMID 33376311) | Preference-informed expectation manipulation with blinded product effect | Quantifies preference–context interaction |
| 11 | High BMI is a population risk factor (PMID 24990315) | Weight-loss barriers and stigma are documented, but the targeted query found no OA trial jointly measuring stigma, engagement and clinical response as of 2026-08-30 (PMID 27313449) | Weight intervention with those three prespecified outcome domains | Optimizes benefit without iatrogenic disengagement |
| 12 | Preoperative peripheral-blood cathepsin S has been associated with persistent pain after hip arthroplasty (PMID 37238223) | Patients prioritize participation, sleep and work | External validation plus a biomarker-guided intervention showing incremental clinical utility | Moves from prognostic association to an outcome-improving decision rule |
Registered follow-up — embolization¶
A 12-month sham-controlled GAE RCT is now published and null (van Zadelhoff 2026, PMID 42009866; NCT03884049 COMPLETED). GEKO (ISRCTN50216144) remains recruiting (target 216; blinded placebo-procedure comparison; six-month VAS pain primary). Follow GEKO as a larger replication of that null, not as the first sham test. Uncontrolled and registry improvement still cannot establish a procedure-specific effect (ISRCTN record, accessed 2026-10-29). No result was posted in the retrieved GEKO record.
Maintenance rule¶
At each literature sweep, search every Tier 1 ID directly. If a study answers a question, retain the ID and add an Answer status line with the resolving evidence rather than deleting the question. Re-run absence searches behind every “missing junction” before strengthening its wording.