Small single-centre RCTs of whey protein plus exercise (n=71, baseline imbalance, follow-up only to postoperative day 5), VR-augmented physiotherapy (n=52, retrospectively registered, underpowered, 1 month) and bed pedalling (n=72, 1 month) all report large early effects. However, each bundles several interventions and relies on generic or short-horizon measures. The psychologically informed physical therapy meta-analysis found gains in OKS, catastrophizing and kinesiophobia but no difference on KOOS, WOMAC, PROMIS or pain, with high heterogeneity and no long-term data. Authors call for 2x2 factorial designs, joint-specific instruments, ITT analysis, dose-response data and 12 to 24 month follow-up, so a preregistered multicentre factorial RCT with blinded assessment, fidelity monitoring and cost data is needed.
Open questions
What the latest literature hasn’t answered.
Explore contradictions, missing evidence, and the next questions worth asking. Every research direction leads back to the papers behind it.
24 questions
Paper count indicates supporting literature, not certainty.Two National Joint Registry target trial emulations (666,283 hips and 697,145 knees) found no difference in all-cause or cause-specific revision at a mean 2.5 years, apart from fewer malposition-related revisions after robotic THR (HR 0.53). Neither measured patient-reported outcomes or cost, and complex hips and dysplasia were explicitly untested. Separately, a retrospective cohort reported higher satisfaction and Forgotten Joint Scores with unrestricted kinematic versus mechanical alignment at 1 year, and its authors call for comparing personalised alignment strategies by CPAK phenotype. The next step is a randomised or registry-linked study with a prespecified alignment strategy, measuring PROMs, dislocation, longer-term survival and within-study cost-effectiveness, plus a dysplasia subgroup.
A Danish emulation of 3,161 patients with dementia found internal fixation associated with fewer recorded reoperations and lower mortality than arthroplasty. The authors concede residual confounding and differential outcome ascertainment, and mobility measures were coarse. A Japanese clone-censor-weight emulation of 4,564 periprosthetic hip fractures (PFF) found no clear mortality benefit from surgery within 2 days, but lacked fracture classification, operative complexity, frailty and reasons for delay; only 11% of early-strategy clones were adherent. Norwegian data showing that PFF risk varies by stem design, sex and age suggest the index implant also shapes these outcomes, so the next study should be a prospective or linked-registry emulation capturing frailty, cognition, Vancouver type and fixation, with function, residence and reoperation as outcomes.
Both RCTs found early weightbearing safe (ankle fracture reoperation 2.3% vs 3.4%; no difference in pain or function after first MTP fusion). However, the ankle trial showed no significant return-to-work gain (median 54 vs 69 days, HR 1.1), and the MTP trial did not measure the mobility, independence, or resource benefits it cites as rationale. The MTP trial was too small to assess non-union and could not exclude a functional advantage for non-weightbearing, while the ankle trial excluded syndesmotic fixation. A pragmatic multicentre RCT powered for non-union, including syndesmotic injuries, with prespecified return-to-work, caregiver burden, and cost outcomes and subgroup stratification, is the next step.
In the Inject trial, high-volume injection with corticosteroid beat sham at 6 weeks and was likely cost-effective, yet was worse than steroid-free injection at 26 weeks, and no injection beat sham at 12 months. The Broström meta-analysis found a 6-month AOFAS advantage for arthroscopy that disappeared by 12 and 24 months, with the RCT-only estimate too imprecise to confirm it, alongside fewer wound complications. Neither resolves whether early benefits translate into outcomes patients value (return to sport or work, costs), nor, for the Achilles, why steroid was linked to later worsening. Adequately powered RCTs with prespecified early and late endpoints, economic analysis, and tendon structural or mechanistic measures are needed.
A retrospective single-centre cohort found less symptomatic hardware with headless screws (20.7% vs 34.8%; adjusted OR 0.55) but no difference in hardware removal (23.4% vs 30.8%). The groups differed in age and sex, and symptoms were not captured with a validated instrument, which the authors flag as a limitation. The ankle fracture weightbearing trial already counts hardware failure within its reoperation endpoint, so future fixation trials could test screw type alongside loading protocols. A randomised trial of headless versus headed screws using formal patient-reported outcomes, removal rates, and costs at 1 to 2 years would settle whether the benefit is real.
Paper 9 (retrospective ANSWER cohort) found less ineffectiveness-related discontinuation with upadacitinib and filgotinib, and prior IL-6i or JAKi exposure worsened persistence only for tofacitinib, baricitinib and peficitinib; however, these are observational data open to residual confounding. Paper 6 (BSR-PsA) found only 24% reached MDA at 12 months, PDE4 inhibition doubled discontinuation risk, and men, younger and employed patients responded better; the authors call for better risk stratification. Paper 2 is placebo-controlled in biologic-naive patients and cannot inform sequencing after failure. Pragmatic head-to-head trials of mechanisms after IL-6i/JAKi failure, plus prospective validation of a response prediction model, are needed.
Paper 1 found structured tapering non-inferior for keeping DAS28-ESR LDA at 24 weeks, but DAS28-CRP, DAS28-ESR and SDAI favoured continuation, and the authors explicitly call for longer studies of structural progression and long-term safety. Paper 7 showed that a 2-week hold around vaccination gave no immunogenicity benefit but more than doubled flare risk (OR 2.27), most strongly with JAKi; Paper 1 also cites reduced-dose baricitinib data showing poorer LDA maintenance. In PsA, Paper 2 shows that damage accrues during inadequately controlled periods, so repeated transient flares may matter radiographically. A multi-year randomised tapering trial stratified by mechanism (TNFi, IL-6i, JAKi), with radiographic progression and cumulative flare burden as endpoints, is the next step.
Paper 4 pooled 9 studies (1886 patients): sequential antiresorptive therapy halved rebound vertebral fracture risk (RR 0.50), zoledronic acid gave the most consistent protection while alendronate only trended toward benefit, and early initiation (≤3 months) looked better than delayed (RR 0.32 vs 0.66) although the interaction test was not significant (p = 0.07). The authors warn that cohort participants adhering to sequential therapy may be lower risk than average, inflating benefit estimates, and they call for head-to-head multicentre RCTs of zoledronic acid versus oral bisphosphonates and for sex-specific trials in men. A randomised trial comparing agent and start timing, stratified by denosumab exposure duration and sex, with incident vertebral fracture as the primary endpoint, would settle this.
In pooled RCT data, dual-mobility CMC1 implants gave the largest 3-month pain and DASH improvements, but 12-month outcomes were comparable, implant complications were 24.49%, and heterogeneity prevented a formal meta-analysis; the authors stress weighing this against long-term durability. Percutaneous retrograde nailing improved DASH only through 2 weeks in a level III cohort followed for 12 weeks, and its authors call for longer-term, patient-reported and economic outcomes. The distal radius study likewise asks for cost-effectiveness analyses and extended follow-up. Randomised trials with multi-year follow-up and formal cost-utility analysis are needed to show whether faster early recovery is worth any added cost, complications, or revision risk.
Nonoperative care of spiral or oblique metacarpal shaft fractures left a mean 3 mm more shortening than surgery, yet extension deficit, cosmesis and satisfaction were similar at 4.5 years; the authors note that the link between shortening and hand function is suspected but unproven (n = 33). The distal radius cohort found casting non-inferior to fixation at 12 months, but it lacked granular radiographic data, so severity-stratified analyses were impossible; the authors ask for these parameters to be recorded prospectively. The next step is a prospective multicentre cohort or RCT that uses reliable radiographic measures (such as SH-Abs) at baseline and union and links them to DASH, PRWE and grip outcomes, in order to define deformity thresholds for choosing nonoperative care.
The 430-patient FIRST trial found no difference in PRWHE between long, short and mini splints. Adherence was only 76 to 83%, and some participants with low splint wear had no adverse consequences, which led the authors to call for testing whether no splint is viable for selected patients. They also note that the effect of the wrist component on tendon excursion with the mini splint is unknown. A noninferiority RCT that compares minimal or no splinting with a relative motion splint is needed; it should measure rupture rates and objectively monitor wear, and it should be paired with biomechanical excursion studies to define which patients can safely forgo immobilization.
In paper 1, surgeon-performed SFA was noninferior to ISB on cumulative 48-hour pain, but ISB gave better relief at 4 to 8 hours and was the only arm with complications (two Horner syndromes and one respiratory event requiring intensive care). This came from a single 80-patient trial of arthroscopic surgery in a mostly Middle Eastern population. Paper 4 found that liposomal bupivacaine ISB lowered pain only below the MCID, with no opioid sparing and 92.5% heterogeneity at 24 hours; like paper 1, its authors call for identifying subgroups rather than adopting universal strategies, and paper 1 adds a call for cost-effectiveness data. The next step is a multicentre RCT stratified by pulmonary risk and procedure type (arthroscopy versus arthroplasty), comparing ISB, SFA, and longer-acting formulations, powered for respiratory complications and including cost analysis.
Paper 3 showed that Latarjet cut redislocation (5% versus 26%) and revisions (0 versus 9) compared with Bankart repair, yet 5-year WOSI scores were nearly identical, osteoarthritis progression did not differ, and only 71% of patients were analysed. Paper 5 similarly found that biologic graft augmentation reduced retear risk (RR 0.31) and increased tendon thickness, but PROM gains were inconsistent, graft types were heterogeneous, and reported harms (adhesive capsulitis, synovitis, foreign-body reaction) were unquantified. The authors call for longer follow-up of arthritis, and for RCTs that pair standardised imaging and biomarkers with PROMs. Extending the Finnish cohort to 10 or more years with standardised arthritis grading, and running graft-specific trials that report patient-relevant outcomes and adverse events, would show whether these structural advantages matter to patients.
Paper 2's network meta-analysis of 189 RCTs found only moderate-certainty evidence, and only short term: ESWT for pain relief, and manipulation under anaesthesia (with or without steroid) for flexion. Most trials were at high risk of bias, mid-term evidence was of low or very low certainty, adverse events were inconsistently reported, and applicability to diabetes, secondary disease, prior surgery, and later-stage disease is unknown. Paper 6's finding that early-mobilisation flexion gains after arthroplasty disappeared by 6 months shows how early range-of-motion advantages can fade. The authors call for adequately powered head-to-head trials with 18 to 24 month follow-up, blinded assessment, and a core outcome set; a trial comparing ESWT, manipulation with steroid, and usual care, stratified by diabetic status, would test whether early gains change long-term recovery.
Recent rehab trials mostly report short-term surrogate gains. F-IMT increased ultrasound diaphragm thickness over 10 weeks in 40 patients, but therapist contact was not matched (6). DNS plus VR improved lordosis angle and balance without a significant pain difference, and its authors question whether the ROM and proprioception changes are clinically meaningful (12). VR education cut preoperative STAI scores by 7.4 points, but no postoperative outcomes were measured (3). In AIS, exercise improved Cobb angle but no modality improved SRS-22 scores (4). Cost-effectiveness evidence covers conventional exercise and education but not these newer modalities, and none of it comes from low- and middle-income countries (7). Multicentre RCTs are needed with dose- and attention-matched controls, patient-reported primary outcomes (pain, ODI, postoperative recovery), follow-up of at least 12 months, and embedded economic evaluation.
Each of these advantages rests on weak or indirect evidence. OLIF beat TLIF on recovery and MCID attainment in older multilevel stenosis patients, but only in a 79-patient retrospective cohort (9). UBE's shorter stay and lower complication odds came mainly from observational data and were not confirmed by the randomised subgroup (13). Epidural anaesthesia's apparent edge for PELD comes from a network with no closed loop and no direct EA versus GA trials (10). The TXA route network shows the same problem: trial sequential analysis found the required information size was not reached for complications or haemoglobin (5). The next step is pragmatic multicentre RCTs, for example OLIF versus TLIF in older multilevel stenosis and EA versus GA for PELD, powered for complications, length of stay, and 2-year patient-reported outcomes.
Several studies independently found that sicker patients improve more. Only the high-suffering, high-disability cluster improved meaningfully after lumbar fusion (14). Higher baseline adverse psychosocial scores and disability predicted clinically relevant EQ-5D gains after neck exercise (2). The F-IMT advantage on ODI grew from about 5.8 to 15.1 points as baseline ODI rose (6). In a related finding, crude opioid-related surgical risks largely disappeared after adjusting for mental health (11). None of these designs can separate true treatment-by-phenotype interaction from ceiling effects or regression to the mean: 14 had no nonoperative arm, and 2 reported weak predictive ability. The authors of 14 call for validating these phenotypes across diverse treatments and testing added psychological support.
NMES improved isometric quadriceps strength only at 6 to 12 weeks after ACL surgery, and its authors say no conclusion is possible beyond 3 months. They call for protocols that extend NMES use into the intermediate term. Supervised exercise after ACLR outperformed self-directed care at 4 months, but the difference narrowed by 12 months and cartilage was unchanged. Hybrid MPFL grafts sped recovery at 3 and 6 months, yet the groups were equivalent at 12 months, and band-training strength gains reversed after only 3 weeks of detraining. Trials with 2 or more years of follow-up should test extended or maintenance dosing, and should judge success by reinjury, return to sport, osteoarthritis markers and cost-effectiveness rather than early surrogate gains.
In revision ACLR, lateral augmentation was associated with half the graft failure risk and less residual pivot shift. However, this evidence is observational (IPTW), and the LET versus ALL comparison was explicitly exploratory and not evidence of equivalence. The French registry found better PROMs with nonhamstring grafts for unclear reasons. Its authors call for testing whether graft effects differ with concomitant lateral procedures and medial injury severity, using objective laxity measures. Double-bundle reconstruction improved pivot shift only at 24 to 48 months without reaching MCID, and one quadriceps study found more positive pivot shift with all-soft-tissue grafts. A randomised trial is needed comparing LET, ALL and no augmentation, stratified by graft type, with objective laxity, sex-disaggregated data and 5-year failure endpoints.
The SVF trial reported superiority over placebo at 12 months, but it used unbalanced randomisation (83 vs 31), a one-sided P value, and a composite endpoint that included the contralateral knee; the abstract also reports no structural MRI result. The PRP pathway trial improved pain and function over conventional PRP care, but its authors state that bundled components and the lack of a placebo arm mean it cannot establish PRP efficacy. Meanwhile, PRP, BMAC and MSC augmentation of ACLR showed no PROM or tunnel benefit, and heterogeneous preparations limited conclusions. A multi-arm or factorial placebo-controlled trial is needed, using characterised products, two-sided endpoints, structural imaging and at least 24 months of follow-up.
Paper 4 found operative clavicle fixation improved union (RR 1.12) and cut malunion, yet Constant and DASH differences excluded the MCID, and 12.3% of patients had implant irritation or removal. Paper 2 found function equivalent between nail and plate for distal tibial fractures; its authors note return to work, quality of life and cost are rarely reported and call for a core outcome set. Paper 5's 0.86-point mRUST advantage has clinical relevance its own authors call uncertain. Future trials should adopt a shared core outcome set (PROMs, return to work, HRQoL, cost-utility) and test whether treatment selected by predicted nonunion or malunion risk, as Paper 4 recommends, improves patient-level results.
Paper 3 pooled only 5 RCTs (1,373 patients) and found a 36% relative reduction in fracture-related infection, confined to gram-positive organisms, with no effect on gram-negative or resistant infections. Its authors call the evidence low quality and say use in closed fracture fixation and fracture-related arthroplasty is underexplored. Paper 2 independently identifies a target population: plate fixation of extra-articular distal tibial fractures more than doubled infection odds versus nailing (OR 2.35, moderate certainty). A multicentre pragmatic RCT in closed, high-risk fixations, stratified by implant and site and powered for deep infection, should also track organism profile, resistance, nephrotoxicity, nonunion and cost-effectiveness.
Paper 1 showed casting without reduction was noninferior to reduction under anaesthesia (PROMIS difference -0.03) in children aged 5 to 10. However, most children had only mild displacement, so its authors say results for greater deformity remain uncertain. Paper 5, in a similar age group treated conservatively, linked 25(OH)D below 20 ng/mL to much lower 12-week union (48.9% vs 80.3%), but dosing was not randomised, outcomes were radiographic only, and remodelling beyond early union was not followed. Paper 1's abstract does not report vitamin D status, and Paper 5 did not measure function. A multicentre trial spanning wider displacement bands should measure baseline vitamin D, ideally randomize supplementation dose in deficient children, and follow PROMIS and remodelling longer term.
Synthesised from recently approved papers over a 56-day window, using abstracts or summaries and available future research directions from the full texts. Read the original studies before designing a project.