{
 "verdict": "VALID_WITH_MINOR_FIXES",
 "claims": [
  {"claim": "Psyllium 2.1 kg (1.6 to 2.6), MA of 6 RCTs, 354 overweight/obese adults; authors employed by P&G", "status": "confirmed", "evidence": "PMID 37163454 abstract: MD -2.1 kg (95% CI -2.6 to -1.6), n=354, 6 studies; COI: employees of Procter & Gamble (Metamucil).", "fix": ""},
  {"claim": "Vegetarian diet 2.0 kg (1.2 to 2.8), 12 RCTs, 1,151 people", "status": "confirmed", "evidence": "PMID 26138004: WMD -2.02 kg (-2.80 to -1.23), 12 RCTs, 1151 subjects.", "fix": ""},
  {"claim": "Pomegranate 2.0 kg (1.0 to 2.9), MA of RCTs 2024", "status": "confirmed", "evidence": "PMID 38370057 (Food Sci Nutr, online Nov 2023, issue Feb 2024): WMD -1.97 (-2.91 to -1.03). Full text (PMC10867489) Table 3: body-weight pool = 22 effect sizes, matching CSV.", "fix": ""},
  {"claim": "Brown rice instead of white 1.6 kg (1.1 to 2.2), weight pools 6 trials", "status": "confirmed", "evidence": "PMID 33905269: -1.63 kg (-2.15 to -1.11), I2=97%, n=6 trials; 13 RCTs overall; GRADE low/very low.", "fix": ""},
  {"claim": "Vinegar 1.1 kg (0.5 to 1.6); umbrella review 2026; body-weight pool of 13 RCTs, 918 people; retracted ACV trial not in weight pool", "status": "not_confirmed", "evidence": "PMID 42110393 abstract: -1.06 kg, high certainty. Full text (PMC13149749): Table 2 lists Weight '13 | 918 (466/452)' but section 3.4.11 says 'A total of 12 trials' and names 12 (Abid, Golzarand 2008 and 2009, Halima, Jafarirad, Jasbi, Kausar, Khezri, Kondo, Park, Bashiri, Thinathayalan); Abou-Khalil 2024 (retracted, PMID 38966098) is cited only in the introduction, not in the weight pool, so that part is confirmed. Article says 13 RCTs and CSV says 12 RCTs.", "fix": "Article table: 'Umbrella review, 2026; body-weight pool of 12 RCTs (13 comparisons), 918 people'. Use the same wording in the CSV."},
  {"claim": "Salmon about 1 kg, men only, after 4 weeks, SEAFOODplus RCT 324 adults, four arms, not specific to salmon", "status": "confirmed", "evidence": "PMID 17502874: 324 adults, 4 arms, average man lost 4.50 kg (salmon) vs 3.55 kg (control) at 4 weeks; no difference in women; cod and fish oil similar.", "fix": ""},
  {"claim": "Oats 0.9 kg (0.1 to 1.8), MA 2022, includes beta-glucan products", "status": "confirmed", "evidence": "PMID 34977959: weight -0.94 kg (-1.84 to -0.05); includes oat beta-glucan-rich extracts/avenanthramides.", "fix": ""},
  {"claim": "Cinnamon 0.9 kg (0.3 to 1.5), 21 RCTs, 1,480 adults", "status": "confirmed", "evidence": "PMID 31800140: WMD -0.92 kg (-1.51 to -0.33), 21 RCTs, 1480 participants.", "fix": ""},
  {"claim": "Olive oil-enriched diets 0.9 kg (0.7 to 1.2), 11 RCTs of 12 weeks or more", "status": "confirmed", "evidence": "PMID 30461730: -0.92 kg (-1.16 to -0.67), 11 studies, >=12 weeks.", "fix": ""},
  {"claim": "Turmeric/curcumin 0.8 kg (0.4 to 1.3), 60 RCTs, 2023", "status": "confirmed", "evidence": "PMID 36882287: WMD -0.82 kg (-1.30 to -0.35), 60 RCTs, n=3691.", "fix": ""},
  {"claim": "Higher-protein diet 0.8 kg (0.1 to 1.5), 24 RCTs, equal-calorie diets", "status": "confirmed", "evidence": "PMID 23097268: -0.79 kg (-1.50 to -0.08), 24 trials, isocaloric energy-restricted diets.", "fix": ""},
  {"claim": "Green tea supplement 0.7 kg (0.5 to 1.0), dose-response MA 2025", "status": "confirmed", "evidence": "PMID 40326418: WMD -0.74 kg (-0.97 to -0.51).", "fix": ""},
  {"claim": "Sparkling water 0.7 kg (0.1 to 1.4), single open-label RCT, 45 adults, exploratory; two authors employed by Asahi Soft Drinks; paper declares no competing interests", "status": "confirmed", "evidence": "PMID 42106408: 46 randomised, 45 completed; -0.72 kg (-1.37 to -0.07); authors say exploratory, no multiplicity adjustment; Hashimoto and Mizuno affiliated with Asahi Soft Drinks; COI 'no competing interests'.", "fix": ""},
  {"claim": "Magnesium 0.7 kg (0.1 to 1.3), MA 2026", "status": "confirmed", "evidence": "PMID 42588058: WMD -0.70 kg (-1.30 to -0.09), 78 RCTs.", "fix": ""},
  {"claim": "Flaxseed 0.6 kg (0.3 to 1.0), 64 trials, 2024", "status": "confirmed", "evidence": "PMID 38992480: WMD -0.63 kg (-1.00 to -0.27), 64 trials.", "fix": ""},
  {"claim": "Almonds 0.5 kg (0.1 to 0.9), 37 RCTs, 2024", "status": "confirmed", "evidence": "PMID 38351580: WMD -0.45 kg (-0.85 to -0.05), 37 RCTs; significant only against nut-free control diets (noted in CSV).", "fix": ""},
  {"claim": "Viscous fibre 0.3 kg (0.1 to 0.5), 62 RCTs, 3,877 adults", "status": "confirmed", "evidence": "PMID 31897475: -0.33 kg (-0.51 to -0.14), 62 trials, n=3877.", "fix": ""},
  {"claim": "Apple cider vinegar SMD -0.39 (-0.63 to -0.15), 10 RCTs, 789 adults, 2025; included trial PMID 38966098 retracted 23 Sept 2025; pooled result not recalculated", "status": "confirmed", "evidence": "PMID 41010525: SMD -0.39 (-0.63 to -0.15), 10 RCTs, 789. Full text (PMC12472926) includes Abou-Khalil 2024 as ref 44 in its included-study table. PMID 38966098 record: 'Retraction in BMJ Nutr Prev Health. 2025 Sep 23'. No erratum or comment is linked to 41010525 in PubMed.", "fix": ""},
  {"claim": "Median of the 17 kg effects is 0.92 kg; counts 18+20+12+13=63", "status": "confirmed", "evidence": "Sorted CSV kg values (2.1, 2.02, 1.97, 1.63, 1.06, 0.95, 0.94, 0.92, 0.92, 0.82, 0.79, 0.74, 0.72, 0.70, 0.63, 0.45, 0.33): the 9th value is 0.92. CSV label counts: 18 WEIGHT_LOSS, 20 NO_SIGNIFICANT_EFFECT, 12 NOT_TESTED, 13 INDIRECT = 63.", "fix": ""},
  {"claim": "A label was kept only where both readers agreed (54 of 63); the other 9 are in 'tested only indirectly'", "status": "wrong", "evidence": "CSV: the readers agreed on 54 items and disagreed on 9 (coffee, ginger tea, lemon juice, pineapple, spinach, sweet potato, vitamin B12, whole grain bread, yogurt), which is correct. But 4 of the 54 agreed items did not keep the agreed label: banana, pasta and tuna (both readers NOT_TESTED) and collagen (both readers SMALL_EFFECT) were all moved to TESTED_ONLY_INDIRECTLY. The indirect group is 9 + 4 = 13. Pasta's move has no note in the CSV.", "fix": "Replace the sentence with: 'Both readers agreed on 54 of 63 items. The 9 disagreements went into the \"tested only indirectly\" group. After independent review, 4 agreed items were also moved there: banana and tuna, because reviewers found trials of banana products and tuna nuggets, and collagen and pasta, because the only evidence was on fat mass or on pasta within whole diets.' Add a note in the CSV row for pasta."},
  {"claim": "Independent reviewers found seven trials our own reviews had not counted", "status": "wrong", "evidence": "The methods paragraph lists 7 items, but the cited PMIDs make 10 trials: tuna 1, Greek yogurt 2, banana 2, prawns 1, tofu 1, corn 1, coconut water 2.", "fix": "Limits section: 'found trials for seven items (ten trials) that our own reviews had not counted'."},
  {"claim": "Creatine: no meaningful effect", "status": "not_confirmed", "evidence": "The best evidence in the CSV (PMID 41433021, 61 trials) found body mass rose significantly by 0.89 kg (0.76 to 1.01), mostly fat-free mass. That fits neither part of the article's definition of 'no meaningful effect' (no significant change, or a change judged not clinically meaningful).", "fix": "Change the definition to 'no weight loss: no significant reduction ...' and add a line under 'Where other reviews disagree' or the item list: 'Creatine: with resistance training it raised body weight by about 0.9 kg, mainly lean mass and water; it did not lower fat mass.'"},
  {"claim": "Coconut oil: 0.04 kg per extra 5 mL a day, not clinically meaningful", "status": "confirmed", "evidence": "PMID 40481535: MD 0.04 kg (0.01 to 0.08), so body weight was higher, not lower; authors judged it not clinically meaningful. PMID 34786908: -0.75 kg against other oils (9 trials).", "fix": "Give the direction: '0.04 kg higher weight per extra 5 mL a day'."},
  {"claim": "Ginger conflict: 36-RCT 2025 MA null; 27-RCT 2024 MA -1.52 kg", "status": "confirmed", "evidence": "PMIDs 41101746 (BW and BMI not improved) and 38261398 (WMD -1.52 kg, 27 RCTs, 1309 participants).", "fix": ""},
  {"claim": "Whey conflict PMID 29087242 significant reduction", "status": "confirmed", "evidence": "PMID 29087242: 9 RCTs, body weight MD 0.56 kg favouring whey. PMID 33971994: 10 studies, 596 participants, no change in body mass.", "fix": ""},
  {"claim": "Grapefruit: early 91-adult trial positive; PMIDs 22304836, 21288350 null; pooled null", "status": "confirmed", "evidence": "PMID 16579728: fresh grapefruit -1.6 vs -0.3 kg placebo, P<.05. PMIDs 22304836 and 21288350: no between-group difference. PMID 25880021: MD -0.45 (-1.06 to 0.16).", "fix": ""},
  {"claim": "Chia conflict PMID 39299649 vs GRADE MA PMID 39225983", "status": "confirmed", "evidence": "39225983: no significant effect on BW (8 studies). 39299649: abstract concludes a possible benefit on body weight but reports no pooled weight estimate.", "fix": ""},
  {"claim": "No-effect items checked: avocado, beet juice, dark chocolate, fish oil, fruit juice, kefir, nuts, peanut butter, quinoa, saffron, water, watermelon, whole wheat bread, vitamin B6", "status": "confirmed", "evidence": "35405602 (8 studies, 657, null); 42130877 (8 RCTs, 199, null); 29553824 (35 RCTs, -0.108 kg ns); 26571503 (21 studies, SMD -0.07 ns); 38227336 (19 RCTs, MD -0.53, -1.55 to 0.48); 37102491 (6 RCTs, 314, p=0.439); 34494363 (full text: 105 comparisons, +0.09 kg ns); 41632974 (120 adults 65+, weight ns); 36847233 (7 trials, 258, -0.54 ns); 35866520 (25 RCTs, -0.32 ns); 37369281 (9 RCTs, BMI ns); 22357746 (79 women, P=0.11); 34386442 (44 women; between-group differences only for fat mass, VAI, insulin, HOMA-IR, TG, not weight).", "fix": "CSV water row: change '7 RCTs' to '8 RCTs'. PMID 38612997 says 'meta-analysis of eight RCTs'."},
  {"claim": "Tofu: 4-week crossover lipid trial in 42 men, tofu vs lean meat (PMID 11103227)", "status": "not_confirmed", "evidence": "PMID 11103227 (Habito 2000) reports sex hormones from the 42-man tofu-vs-meat crossover. The lipid report of the same trial is PMID 10694766 (Ashton & Ball 2000).", "fix": "Write 'a 4-week crossover in 42 men, tofu against lean meat, reporting lipids and sex hormones (PMIDs 10694766, 11103227)'."},
  {"claim": "Other cited 'tested for other purposes' and 'indirect' PMIDs (Greek yogurt 31114790, 41470847; shrimp 20196970; corn 40944189; coconut water 12056182, 17883020; banana 30887937, 20623003; collagen 35658958; pasta 29615407; tuna 42369918)", "status": "confirmed", "evidence": "Abstracts match the article's descriptions: corn cereal was the comparator to barley in 30 overweight Koreans; the tuna trial was 120 ICM patients with -1.8 kg vs standard care; pasta had 0 pasta-alone comparisons and -0.63 kg within low-GI diets; collagen fat mass -1.21 kg with body weight not pooled.", "fix": ""},
  {"claim": "Plain words: '12 have never been tested ... No randomised trial has given them to people trying to lose weight, or to people with overweight, and measured their weight.'", "status": "wrong", "evidence": "As written, this is literally false. Granola: a 12-week RCT gave granola bars twice daily to 55 overweight/obese adults and measured body weight (PMID 29490721); both arms ate granola bars, so granola was not the thing tested. Corn: overweight adults ate corn cereal daily for 6 weeks with weight measured (PMID 40944189), as the comparator. The journalist definition excludes these through the 'as the intervention, not the comparison' clause, but the plain-language sentence leaves that clause out and states an absolute negative, while the Limits section correctly says 'we ... did not find'.", "fix": "Replace with: '12 have not been tested as a way to lose weight. We found no randomised trial that gave them, as the thing being tested, to people trying to lose weight or with overweight and then measured their weight.' Also add to the list of other-purpose trials: 'Granola: a 12-week trial in 55 overweight adults compared two granola bars, one with added fibre; neither was compared with a non-granola snack (PMID 29490721).'"},
  {"claim": "CSV consistency for relabelled items", "status": "wrong", "evidence": "The tuna row's best_evidence and supporting_sentence still say 'No RCT or systematic review of tuna and body weight' and the pmid field is empty, which contradicts the TESTED_ONLY_INDIRECTLY label and the article's description of PMID 42369918. Banana is the same: 'No RCT ...' with an empty pmid, although the article cites 30887937 and 20623003.", "fix": "Tuna row: set pmid=42369918 and rewrite best_evidence to describe the tuna-nugget RCT. Banana row: set pmid=30887937 and rewrite best_evidence as 'green banana biomass and banana starch trials; no trial of fresh fruit'. Mark supporting_sentence as superseded by the independent review."}
 ],
 "missed_trials": [
  {"item": "granola", "pmid": "29490721", "why_it_matters": "A 12-week triple-blind RCT gave granola bars (with or without oligofructose) twice daily to 55 overweight/obese adults and measured body weight; there was no difference. It does not meet the article's definition, because granola was in both arms. But it contradicts the plain-language sentence 'No randomised trial has given them to people ... with overweight, and measured their weight', and a journalist searching 'granola weight RCT' finds it first. The CSV mentions it; the article does not."},
  {"item": "Greek yogurt (borderline)", "pmid": "31987116", "why_it_matters": "An 8-week RCT in 70 women with overweight/obesity: a low-energy diet with 50 g/day Kashk (condensed, strained, high-protein probiotic yogurt) gave significantly greater falls in BMI and fat % than the same diet without it. Kashk is not marketed as Greek yogurt, so the label can stand, but it is the closest trial of a strained high-protein yogurt for weight loss and is worth naming. PMID 41176279 (6-week mixed dairy including Skyr, 30 OW/OB adults, no weight change) also exists; it is excluded under the 'wider group' rule."},
  {"item": "tofu (borderline)", "pmid": "36615709", "why_it_matters": "A 4-week double-blind crossover of Kori-tofu (freeze-dried tofu) against whey protein in 45 adults with mildly raised cholesterol (BMI 18.5-35, mean 25.4). Body weight was measured and rose similarly in both arms (+0.4 vs +0.3 kg). The population was not overweight by design, so the label stands, but it is a second tofu RCT with weight data that journalists will find. Worth adding to the tofu bullet."},
  {"item": "soursop (other purpose)", "pmid": "31462727", "why_it_matters": "A 3-month RCT of soursop juice (2 x 100 g/day) in 143 prehypertensive adults; the abstract reports no weight outcome. It does not affect the label, but it is the main soursop RCT and could be listed in 'trials for other purposes' so the soursop entry does not look unsearched."}
 ],
 "wording_issues": [
  {"text": "12 have never been tested as a way to lose weight. No randomised trial has given them to people trying to lose weight, or to people with overweight, and measured their weight.", "problem": "An absolute negative that is literally false for granola (PMID 29490721) and corn (PMID 40944189, as comparator). It is stronger than the journalist definition and the Limits section ('we ... did not find').", "fix": "'12 have not been tested as a way to lose weight: we found no randomised trial that gave them, as the thing being tested, to people trying to lose weight or with overweight and measured their weight.'"},
  {"text": "No meaningful effect: ... the best available evidence found no statistically significant change, or a change its authors judged not clinically meaningful", "problem": "Creatine's cited best evidence shows a significant weight increase (+0.89 kg), so it does not fit the stated definition. The category really means 'no weight loss'.", "fix": "Rename or redefine the category as 'no weight loss shown', or add an explicit creatine note (weight gain from lean mass/water)."},
  {"text": "(coconut oil: 0.04 kg per extra 5 mL a day)", "problem": "No direction is given; a lay reader may read it as loss. The pooled change was +0.04 kg (weight higher).", "fix": "'(coconut oil: 0.04 kg higher weight per extra 5 mL a day)'"},
  {"text": "A label was kept only where both agreed, which happened for 54 of 63", "problem": "Four agreed labels (banana, tuna, collagen, pasta) were changed, and the CSV shows it.", "fix": "See the claims fix above."}
 ],
 "required_changes": [
  "Plain-language section: replace 'never been tested ... No randomised trial has given them ...' with the hedged wording ('we found no randomised trial that gave them, as the thing being tested, ...'), and add the granola-bar trial (PMID 29490721) to the list of trials for other purposes.",
  "Fix the creatine inconsistency: its best evidence shows a significant weight GAIN (+0.89 kg, PMID 41433021), which does not meet the stated 'no meaningful effect' definition. Redefine the category as 'no weight loss' or add an explicit creatine note.",
  "Correct the methods sentence 'A label was kept only where both agreed (54 of 63)': 4 agreed items (banana, tuna, collagen, pasta) were relabelled to 'tested only indirectly'. Say so, and add a CSV note for pasta.",
  "Limits section: 'seven trials' should read 'trials for seven items (ten trials)'.",
  "Vinegar row: the article says 13 RCTs and the CSV says 12. The umbrella review's text names 12 trials; its table shows 13 comparisons and 918 participants. Use '12 RCTs (13 comparisons), 918 people' in both places.",
  "Coconut oil: give the direction ('0.04 kg higher').",
  "Tofu bullet: PMID 11103227 is the sex-hormone report; the lipid report of that trial is PMID 10694766. Cite both or describe correctly.",
  "CSV: the tuna and banana rows still say 'No RCT' and have empty pmid fields despite their relabelling. Update best_evidence and pmid (tuna 42369918; banana 30887937/20623003). Water row: '7 RCTs' should be '8 RCTs' (PMID 38612997)."
 ],
 "optional_suggestions": [
  "Add a 'where other reviews disagree' note for positive items as well as null ones. For green tea, the CSV notes a Cochrane review of trials outside Japan found 0.04 kg (not significant). For psyllium, the CSV says psyllium alone was not significant in the larger viscous-fibre MA (PMID 31897475). Without these notes, conflicting evidence is disclosed only for items in the null group.",
  "Mention borderline trials journalists may find: Kori-tofu 4-week crossover with weight measured (PMID 36615709); Kashk (strained high-protein yogurt) 8-week RCT in overweight women (PMID 31987116); mixed dairy including Skyr in OW/OB adults (PMID 41176279); soursop juice 3-month BP trial (PMID 31462727).",
  "Whole wheat bread is labelled 'no meaningful effect' on the basis of a trial of whole-wheat foods in general (PMID 22357746), while whole grain bread is 'indirect' because the evidence is 'whole grains in general'. Consider harmonising, or explain that the Kristensen trial's wheat foods were mostly bread.",
  "Vitamin B6 rests on a single 44-woman trial whose authors conclude B6 'may be effective in reducing BMI' and which found lower fat mass vs placebo. A one-line caveat would pre-empt a journalist's objection.",
  "State the search cut-off date (month/year) for the Bioma reviews and the reviewer rounds, so readers know how current 'not found' is.",
  "Salmon's ~0.95 kg is a model-estimated difference for an 'average man' with no CI for the contrast. Consider excluding it from the median or showing the median with and without the two single-trial items (sparkling water, salmon)."
 ]
}