Celecoxib

Limits on what this shows

• Null findings: Lal 2009 [8] compared celecoxib, diclofenac, ketorolac, bromfenac and nepafenac in isolated rat liver mitochondria under tert-butyl hydroperoxide stress and reported drug-specific protective or detrimental effects; the direction for celecoxib is not stated in the reconstructed abstract. Concentrations used were not retrieved, so no comparison with human plasma levels was possible. Protein binding of about 97% [1] lowers free drug.
• At-risk subgroups: None established. Hepatic impairment raises exposure (label dose reduction) [1].

Direction

Direct mitochondrial toxicity of the drug (not a nutrient deficiency).

Regulatory status

Not mentioned in the label sections retrieved [1][2].

Population evidence

No human study. Tatematsu 2018 [7]: isolated rat liver mitochondria showed dose-dependent swelling not reversed by cyclosporine A, release of matrix proteins (ALDH2, HMGCS2) and cytochrome c, and increased membrane permeability in human erythrocytes and liposomes (reconstructed abstract). Matheny 2022 [5]: primary human myoblasts and myotubes; celecoxib, but not the COX-2 inhibitor NS-398, reduced mitochondrial membrane potential and cellular respiration, indicating COX-2-independent action. Syed 2016 [6]: celecoxib among five coxibs tested for mitochondrial toxicity in rat liver. Guo 2026 [9]: high-dose celecoxib (above the 20 micromolar or lower range that was anti-inflammatory) caused mitochondrial stress in nucleus pulposus cells.

Mechanism

Named: membrane-permeabilizing swelling distinct from classical permeability-transition-pore opening [7]; loss of membrane potential [5].

Counter-evidence held on record

Counter-evidence and confounding All sources are Tier B in vitro or animal work that used celecoxib itself, so no class extrapolation is involved. Test concentrations may exceed free plasma levels. Abstracts were reconstructed by OpenAlex; full texts were not opened. Framing caution for the consumer surface Mechanistic signal only. It is not evidence of mitochondrial damage in people taking celecoxib.

Limits on what this shows

• Null findings: Lal 2009 [8]: effect of celecoxib on mitochondrial dysfunction during oxidative stress was drug-specific; direction not stated in the abstract reconstruction. No human serum MDA, isoprostane or glutathione measurement on celecoxib was retrieved.
• At-risk subgroups: None established; the rat work involved renal, hepatic and gastric tissue and juvenile animals [10].

Direction

Mixed. Pro-oxidant toxicity in rats; reverse direction (protective) in a human trial and endothelial cells.

Regulatory status

Not mentioned in the label sections retrieved [1][2].

Population evidence

Sozer 2011 [10]: young Wistar rats on celecoxib showed histopathologic changes with raised MDA and lower GSH, SOD and catalase (reconstructed abstract; doses not retrieved). Gupta 2007 [11]: male Wistar rats, chronic celecoxib, interstitial nephritis, higher lipid peroxidation and nitrite, lower antioxidant enzymes, higher urea and creatinine. Chenevard 2003 [12]: double-blind placebo-controlled crossover, 14 men with severe CAD, celecoxib 200 mg twice daily for 2 weeks; oxidized LDL 43.6 vs 47.6 U/L (P=0.028), hs-CRP lower, flow-mediated dilation higher. Hamdulay 2010 [13]: in endothelial cells celecoxib activated PI-3K/Akt and mitochondrial redox signaling, inducing heme oxygenase-1 via Nrf2.

Mechanism

Named on both sides: oxidative injury in tissue toxicity models [10][11]; Nrf2/HO-1 induction through mitochondrial ROS signaling [13].

Counter-evidence held on record

Counter-evidence and confounding Rodent doses are not in the retrieved abstracts and may exceed human exposure. Chenevard [12] is small (n=14), 2 weeks, in a CAD population, and oxidized LDL is a vascular marker, not systemic antioxidant capacity. Circulating consumer lists say NSAIDs deplete glutathione [28][29]; the only celecoxib-specific glutathione data are rat tissue values [10]. Framing caution for the consumer surface Neither a pro-oxidant nor an antioxidant label is supported for people taking celecoxib. No antioxidant supplement is implied.

Limits on what this shows

• Null findings: No ferritin, serum iron, TIBC or transferrin saturation was reported in any retrieved abstract. PubMed listing searches aimed at celecoxib and iron indices (iron/ferritin/hepcidin, serum iron) were rate-limited (429) and never completed, so this is a tooling-limited null, not a confirmed absence of data.
• At-risk subgroups: Older age, prior gastritis, H. pylori infection, high CRP, anticoagulant, antiplatelet, SSRI or SNRI co-use [1][23].

Direction

Blood loss (possible iron loss through the GI tract). Not an absorption or acid-suppression effect. Celecoxib does not suppress gastric acid. Anemia itself is an outcome, not iron deficiency (Hard Rule 2).

Regulatory status

Label 5.12: anemia has occurred in NSAID-treated patients, "may be due to occult or gross blood loss, fluid retention, or an incompletely described effect on erythropoiesis"; incidence 0.6% celecoxib vs 0.4% placebo; NSAIDs including celecoxib may increase bleeding risk. The sections read do not state a lower blood-loss claim for celecoxib; comparative direction comes from trials below [1].

Population evidence

CONDOR elderly subgroup [22]: 6-month, double-blind, 2,446 patients aged 65 or older with OA/RA at elevated GI risk; hemoglobin fall of 2 g/dL or more in 23 of 1,219 (1.9%) on celecoxib 200 mg twice daily vs 76 of 1,227 (6.2%) on diclofenac SR 75 mg plus omeprazole 20 mg. CONDOR post-hoc [23]: 3,774 OA patients, 64 (1.7%) had Hb fall of 2 g/dL or more and/or hematocrit fall of 10% or more; risk factors were higher CRP, gastritis history, H. pylori positivity, age and BMI. Fujimori 2015 [24]: 14-day double-blind RCT, 150 healthy volunteers; hemoglobin fell 2.1% on celecoxib 200 mg/day vs 5.1% on loxoprofen plus lansoprazole (P=0.006). CLASS [21]: 8,059 patients, 6 months; fewer chronic GI blood loss events on celecoxib 400 mg twice daily than on ibuprofen or diclofenac.

Mechanism

Present: GI mucosal injury and occult bleeding, less with COX-2 selectivity (small-bowel mucosal breaks in 10% on celecoxib vs 49% on loxoprofen plus lansoprazole [24]).

Counter-evidence held on record

Counter-evidence and confounding Every comparison above is against an active nonselective NSAID plus PPI, which tests relative GI safety, not celecoxib against non-use; a hemoglobin fall on celecoxib is present in both arms and cannot be attributed to the drug without a placebo arm. The pooled 51-trial analysis [25] included placebo arms but its abstract does not stratify by arm. Underlying arthritis (anemia of inflammation) is a competing explanation. Framing caution for the consumer surface A hemoglobin fall is not iron deficiency and does not mean iron should be taken. New anemia on any NSAID warrants a bleeding work-up, not self-supplementation.

Limits on what this shows

• Null findings: PubMed listing for "celecoxib copper" (about 19 records) returned no study of serum copper, ceruloplasmin or copper absorption in treated people; other hits were chemical-synthesis papers and cancer cell work. Cisplatin-uptake papers in the listing concern platinum, not copper.
• At-risk subgroups: None established.

Direction

Excess/accumulation, intracellular, cell-type specific. Not depletion.

Regulatory status

Copper not mentioned in the label sections retrieved [1][2].

Population evidence

Guo 2026 [9] (Advanced Science): in nucleus pulposus cells and rat models, low-dose celecoxib (20 micromolar or less) suppressed inflammation and preserved matrix, while high-dose celecoxib upregulated HSP90, stabilized RBX1, promoted degradation of the copper exporters ATP7B and COMMD1, and led to copper accumulation, mitochondrial stress and cuproptosis (reconstructed abstract).

Mechanism

Named: HSP90/RBX1-mediated degradation of ATP7B and COMMD1 [9].

Counter-evidence held on record

Counter-evidence and confounding A single recent preclinical report, concentration-dependent and tissue-specific, and the same mitochondrial-stress observation overlaps with the Mitochondrial burden entry. It was not independently replicated in any source retrieved. Framing caution for the consumer surface This is not evidence of copper deficiency or toxicity in patients, and no copper supplement or restriction is implied.

Limits on what this shows

• Null findings: CLASS EXTRAPOLATION, cohort test: Lazar 2019 [15] tested rofecoxib only in rats (4 weeks, 16S rRNA) and found no significant change in jejunal microbiota composition or diversity. Celecoxib was NOT in that cohort, so this is Tier C, capped at yellow, and reverse or null in direction. Listing for "celecoxib microbiome 16S" otherwise returned an HCC study of the non-COX derivative 2,5-dimethylcelecoxib and other models, none a human celecoxib microbiome study.
• At-risk subgroups: None established.

Direction

Altered microbial function (butyrate down, acetate up) in vitro. Net clinical direction unknown.

Regulatory status

Not mentioned in the label sections retrieved [1][2].

Population evidence

Hernandez-Sanabria 2020 [14]: batch culture of human fecal samples in a simulated gut ecosystem model; celecoxib decreased butyrate production by 50%, shifted community composition (donor-dependent), increased acetate, and the treated microbiota preserved epithelial barrier function and lowered IL-8 and CXCL16 in a co-culture inflammation model; microbiota explained about 5% of variance in the epithelial response.

Mechanism

Named in outline: drug-microbe interaction altering short-chain fatty acid output [14].

Class cohort test

Rofecoxib (not celecoxib) in [15]; mechanism relevance limited because rofecoxib was chosen as a non-acidic, non-mucosal-injury COX-2 inhibitor and the authors note other coxib-specific properties may matter. Used only as a counterpoint to dysbiosis claims.

Counter-evidence held on record

Counter-evidence and confounding The in vitro model lacks host, bile and diet; one donor-dependent culture study with no replication. The authors describe the treated community effect on the epithelial cells as anti-inflammatory, not harmful. Framing caution for the consumer surface Not a claim of dysbiosis in celecoxib users; a prompt for future human 16S work only.

Limits on what this shows

• Null findings: Searches for celecoxib hemolytic anemia and methemoglobinemia returned no celecoxib case or study (the methemoglobinemia query returned only generic and other-drug papers). No measurement of arterial oxygenation, red-cell mass or tissue oxygen delivery was retrieved.
• At-risk subgroups: As in the Iron entry; also anyone with a history of blood dyscrasia, since the label postmarketing list includes agranulocytosis, aplastic anemia and pancytopenia [1].

Direction

Reduction in hemoglobin (blood loss, with fluid retention named as a possible contributor in the label). Outcome entry (Hard Rule 2). Distinct from the Iron entry, which concerns the iron nutrient.

Regulatory status

Label 5.12: anemia incidence 0.6% vs 0.4% placebo; monitor hemoglobin or hematocrit if signs or symptoms of anemia. Postmarketing: agranulocytosis, aplastic anemia, pancytopenia, leucopenia (frequency not given) [1].

Population evidence

Sands 2012 [25]: pooled analysis of 51 blinded controlled celecoxib trials of 4 weeks or longer, 51,048 participants, 932 (1.8%) with Hb fall of 2 g/dL or more and/or Hct fall of 10% or more across celecoxib, NSAID and placebo arms; affected patients had more GI events, withdrawals (16.7% vs 10.4%) and cardiac events. Per-arm rates are not in the abstract. CONDOR [22][23] and Fujimori [24] rates as in the Iron entry.

Mechanism

Present, shared with the Iron entry (GI blood loss) plus possible hemodilution from fluid retention named by the label [1].

Counter-evidence held on record

Counter-evidence and confounding No placebo-stratified Hb estimate was retrieved, so a causal excess over non-use is not established; the label anemia difference (0.6% vs 0.4%) is small. Disease-related anemia and aspirin or anticoagulant co-use are competing explanations. Framing caution for the consumer surface This is a monitoring prompt about blood count, not a claim that celecoxib impairs oxygen delivery or that any nutrient is deficient.

Limits on what this shows

• Null findings: No measurement of cellular ATP content, lactate or exercise energetics in people or intact animals was retrieved. The reconstructed Tatematsu abstract also reports increased oxygen consumption, which the authors read as uncoupling.
• At-risk subgroups: None established.

Direction

Direct inhibition of mitochondrial ATP synthesis by the drug (not a nutrient deficiency).

Regulatory status

Not mentioned in the label sections retrieved [1][2].

Population evidence

No human data. Syed 2016 [6]: celecoxib, etoricoxib, rofecoxib, valdecoxib and lumiracoxib inhibited oxidative phosphorylation (ATP synthesis) in rat liver mitochondria; IC50 values were compared in the paper but are not in the reconstructed abstract. Tatematsu 2018 [7]: State 3 (ADP-stimulated) respiration reduced. Matheny 2022 [5]: lower respiration and lower cis-aconitate, alpha-ketoglutarate, succinate, malate, pyruvate and fumarate in myotubes.

Mechanism

Named: inhibition of electron transport and ATP synthesis in isolated mitochondria; metabolite formation and lipophilicity were proposed modifiers [6].

Counter-evidence held on record

Counter-evidence and confounding Shares its underlying studies with the Mitochondrial burden entry (see Overlap map), so the two are not independent confirmation. Hepatotoxicity from selective COX-2 inhibitors is discussed in [6] as a possible mitochondrial outcome, but that is an organ toxicity outcome rather than an energy-nutrient finding. Framing caution for the consumer surface Not a claim of low energy or fatigue caused by celecoxib.