Sulfamethoxazole/trimethoprim

Limits on what this shows

• Key negative findings: Comparator antibiotics (ciprofloxacin, norfloxacin, amoxicillin) were not associated in Antoniou 2010 [5]; norfloxacin was not significant in 2011 [6]. In Gentry 2013 the low-dose group still had 1.05% hyperkalemia, so the baseline risk is not zero without the drug but is lower than at high dose [8]. No contradicting Tier A evidence on direction was found.
• At-risk subgroups: Older adults (66 or older in the Ontario studies; over 58 in the US cohort), users of ACE inhibitors, ARBs, spironolactone, beta-blockers or NSAIDs, renal impairment or elevated baseline creatinine, high-dose treatment, underlying potassium disorders [1][5][6][7][8].

Direction

EXCESS / RETENTION. Reduced renal potassium excretion. Sulfamethoxazole is not implicated; the label attributes the effect to the trimethoprim component.

Regulatory status

Label (read via a summarizing tool, two repackager copies): high dosage of trimethoprim "induces a progressive but reversible increase of serum potassium concentrations in a substantial number of patients"; "Even treatment with recommended doses may cause hyperkalemia" in patients with underlying potassium disorders, renal insufficiency or hyperkalemia-inducing drugs; close monitoring of potassium is warranted; ACE inhibitors: avoid concurrent use; hyperkalemia is among the most frequently reported severe reactions in the elderly [1][2].

Population evidence

Antoniou 2010 [5]: nested case-control, Ontario adults aged 66 or older on continuous ACE inhibitor or ARB therapy; 4,148 hyperkalemia admissions in 14 years, 371 within 14 days of an antibiotic; trimethoprim-sulfamethoxazole versus amoxicillin adjusted OR 6.7 (95% CI 4.5 to 10.0); the other antibiotics (ciprofloxacin, norfloxacin, nitrofurantoin) showed no such association. PubMed abstract through a summarizer. Antoniou 2011 [6]: nested case-control in patients aged 66 or older on chronic spironolactone; 6,903 hyperkalemia admissions, 306 within 14 days of an antibiotic, 248 cases matched to 783 controls; trimethoprim-sulfamethoxazole adjusted OR 12.4 (7.1 to 21.6) versus amoxicillin; nitrofurantoin 2.4 (1.3 to 4.6); norfloxacin 1.6 (0.8 to 3.4, not significant); single-strength tablet OR 9.7 (4.7 to 19.9) and double-strength 13.5 (7.5 to 24.2); population attributable fraction 59.7%. 10.8% of spironolactone users (17,859 of 165,754) received at least one prescription. BMJ page read via a summarizer; hyperkalemia defined by admission diagnosis codes (unvalidated), serum potassium not available. Weir 2010 [7]: 299,749 beta-blocker users aged 66 or older; 189 admitted for hyperkalemia within 14 days of a study antibiotic; trimethoprim-sulfamethoxazole versus amoxicillin adjusted OR 5.1 (2.8 to 9.4). Abstract only; the full text returned HTTP 402 and was not opened. Gentry and Nguyen 2013 [8]: retrospective electronic-record cohort of 6,162 US outpatients with measured potassium; hyperkalemia 3.06% on high-dose versus 1.05% on low-dose (P <.0001). Independent predictors: high-dose OR 2.92 (1.85 to 4.60), concomitant ACE inhibitor 3.27 (2.06 to 5.14), age over 58 years 3.44 (1.86 to 7.0), NSAID 1.71 (1.02 to 2.79), baseline elevated serum creatinine 45.1 (21.7 to 93.2). PubMed abstract via a summarizer (citation first located through a Tier D encyclopedia lead [40]); numbers confirmed against a second record (OpenAlex reconstructed abstract), which adds acute renal injury 1.99% versus 0.700% (P =.0001) and, for that outcome, high-dose OR 3.70 (1.70 to 8.12); Oklahoma City VA, outpatients, dose threshold not in the abstract. Velazquez 1993 [9]: 30 consecutive AIDS patients on high-dose trimethoprim (20 mg/kg/day); serum potassium rose by about 0.6 mmol/L (interval as reported by the summarizing tool: 0.29 to 0.95) and exceeded 5 mmol/L in 15 of 30. Second record (OpenAlex reconstructed abstract) gives the interval 0.29 to 0.95 and adds a 39-rat arm: intravenous trimethoprim reduced renal potassium excretion by 40% (CI 21% to 60%), tubular trimethoprim at 1 mmol/L reduced distal potassium secretion by 59% (CI 26% to 92%) and made the lumen-negative transepithelial voltage 66% less negative (CI 46% to 85%); the reconstruction tool noted word order may be imperfect. Patients had normal adrenocortical function and glomerular filtration. The design is uncontrolled in people; the transtubular potassium gradient was calculated but no value was in the abstract. Preyra 2025 [42]: systematic review and meta-analysis of adult outpatient trials (24 randomized trials in the pooled analysis, 4,458 participants as reported in the abstract) plus a FAERS pharmacovigilance analysis; versus azithromycin, hyperkalemia was reported about 56-fold more often (abstract through an OpenAlex reconstruction). FAERS disproportionality reflects spontaneous reports, not incidence, and no CI or report counts were retrievable (PMC copy returned a bot-verification page and the Wiley pages failed). It shares no database with the Ontario studies and is used as supporting, not deciding, evidence. Smaller prophylaxis-dose signal [12]: Coppens 2019, 63 glioma patients on low-dose co-trimoxazole prophylaxis (28 on RAAS inhibitors, 35 not); laboratory disturbance (sodium, potassium, creatinine) OR 3.17 (P = 0.03) for the RAAS group, generally moderate. No CI or per-lab rates in the abstract.

Mechanism

Present and named. Trimethoprim is structurally similar to amiloride and blocks the epithelial sodium channel in the distal nephron, impairing renal potassium elimination [10][11]; the Antoniou 2011 introduction states it reduces urinary potassium excretion by about 40% in a dose-related way [6]. The human tubular study in AIDS patients is consistent with this [9]; a rat study exists [13] (listing only, not opened).

Downstream outcome (not a nutrient)

Sudden death is a clinical outcome, listed for context. Fralick 2014 [10]: 1,601,542 ACE inhibitor or ARB users aged 66 or older; 39,879 sudden deaths; within 7 days of an antibiotic, 1,027 matched to 3,733 controls; co-trimoxazole versus amoxicillin adjusted OR 1.38 (1.09 to 1.76), within 14 days 1.54 (1.29 to 1.84), about 3 versus about 1 sudden deaths per 1,000 prescriptions; potassium not measured, hyperkalemia is the proposed explanation. Antoniou 2015 [11]: spironolactone users, 328 matched cases, adjusted OR 2.46 (1.55 to 3.90), sensitivity analysis 1.94 (0.93 to 4.04). These support clinical importance but are not used as potassium measurements.

Why red

Repair-pass check (2026-10-08): kept red. The two directly measured-potassium sources, Gentry 2013 [8] (US VA outpatients) and Velazquez 1993 [9] (AIDS patients plus a rat tubular experiment), are independent of the Ontario database and of each other, point the same way, have a named and experimentally demonstrated mechanism, and no Tier A source contradicts them. The Ontario triplet is counted as one infrastructure and is not needed to reach red. Honest limits: Gentry compares high-dose with low-dose co-trimoxazole, not exposure with non-exposure; Velazquez is uncontrolled in people and used trimethoprim-containing drugs at 20 mg/kg/day; Preyra [42] is reporting data. Original reasoning follows. At least two independent Tier A sources with the combination in the cohort and a hyperkalemia-related measure: the US outpatient cohort with measured potassium [8] and the human tubular study [9] are independent of the three Ontario studies [5][6][7]. The Ontario studies use distinct drug-defined cohorts (ACE inhibitor or ARB, spironolactone, beta-blocker) but share one province-wide administrative database and may overlap in patients, so they are counted as one infrastructure, not three independent confirmations. Direction is consistent, the effect is large (odds ratios 5 to 12 for admission), the mechanism is named, and no Tier A source contradicts it. The Ontario outcome is a diagnosis code, which is why the measured-potassium cohorts carry the "directly measured" criterion.

Counter-evidence held on record

Counter-evidence and confounding The Ontario studies are administrative, could not see serum potassium, renal function or adherence, and used amoxicillin as the active comparator, so they test co-trimoxazole against another antibiotic, not against no antibiotic; confounding by infection type is possible but the odds ratios and the dose gradient are large. Hospital-coded hyperkalemia has not been validated. The Gentry cohort is retrospective, single-system and adjusted for a limited set of factors. Framing caution This is an excess/retention risk. Do not add potassium supplements, potassium-based salt substitutes or high-potassium foods in response to this entry; doing so could worsen the risk. Anyone on an ACE inhibitor, ARB, spironolactone or with kidney disease should have potassium checked by their prescriber during a course. Never stop a heart or kidney medicine without clinical advice.

Limits on what this shows

• Key null findings and limits: No study read compares sodium on co-trimoxazole with sodium on a non-exposed or active-comparator group. Illness itself (Pneumocystis pneumonia, hospitalization, fluids) can lower sodium. Hyponatremia at standard doses is documented as case reports only [16].
• At-risk subgroups: High-dose treatment, hospitalized patients, age 65 or older, concurrent RAAS inhibitors (Coppens) [1][12][14].

Direction

DISTURBANCE, not a dietary deficiency (Hard Rule 5). Urinary sodium loss with water retention. It does not imply that salt or sodium intake should increase.

Regulatory status

Label: "Severe and symptomatic hyponatremia can occur in patients receiving sulfamethoxazole and trimethoprim", particularly for treating Pneumocystis pneumonia; evaluation and correction are needed in symptomatic patients [1][2].

Population evidence

Tsapepas 2016 [14]: single-center retrospective review of hospitalized adults, 235 screened and 76 analyzed (trimethoprim more than 8 mg/kg/day for at least 3 days; excluded heart failure, cirrhosis, eGFR below 30, baseline hyponatremia and other sodium-lowering drugs); mean baseline sodium 138.4; 55 of 76 (72.3%) developed sodium below 136 mEq/L, 24 of those 55 (43.6%) below 130; mean nadir 131.6 mEq/L, onset on average 5.5 days; mean urine sodium 104.8 mEq/L in the 22 cases with a value; usually resolved within 3 weeks of stopping. Abstract through a summarizer; no comparison group and no odds ratios. Nagai 2024 [15]: three-center retrospective cohort of non-HIV Pneumocystis pneumonia, 81 patients in total, 55 low-dose (below 12.5 mg/kg/day) and 26 conventional-dose (12.5 to 20 mg/kg/day), overlap-weighted for background characteristics (group sizes corrected on 2026-10-08 from an OpenAlex abstract reconstruction; the first reading had summed the groups wrongly); adverse events of grade 3 or higher, "especially nausea and hyponatremia", 29.8% versus 59.0% (P =.005). The per-arm hyponatremia rate is not in the abstract and the open-access full text could not be opened (publisher pages failed, the OpenAlex PDF copy returned HTTP 429), so no sodium-specific figure is claimed. Stephens 2020 [43]: retrospective review of children under 18 given intravenous sulfamethoxazole-trimethoprim, 61 patients and 66 courses; serum sodium below 135 mEq/L in 20 courses (30.3%); median nadir 133 versus 138 mEq/L in the non-hyponatremic courses (P < 0.001); cumulative dose did not differ (P = 0.104); concomitant diuretic use 80.0% versus 50.0% (P = 0.023), furosemide the drug most associated. Abstract via an OpenAlex reconstruction; no unexposed comparator, so it is a second independent measured-sodium cohort but cannot size excess risk. Preyra 2025 [42]: FAERS comparison with azithromycin, hyponatremia reported about 10-fold more often (abstract only; reporting data, no CI retrieved). Supporting, not deciding. Lei 2022 [16]: case series of 24 patients (median age 67) with co-trimoxazole hyponatremia, median nadir 118 mmol/L (range 101 to 128.1), median sodium normalized 4 days after withdrawal. Listing-level read. Rechallenge reports [16]: an 82-year-old woman, sodium 132 to 121 mmol/L on starting and 138 to 129 on rechallenge (Naranjo score 9); an 86-year-old man with recurrence on inadvertent rechallenge (Naranjo 7). Case reports, Tier B. Prophylaxis dose : Coppens 2019 [12] found more laboratory disturbance (sodium, potassium, creatinine pooled) with RAAS inhibitors on low-dose prophylaxis; no sodium-specific rate was in the abstract.

Mechanism

Proposed and named: the same amiloride-like ENaC blockade by trimethoprim that causes hyperkalemia also promotes urinary sodium loss, consistent with the high urine sodium in the Tsapepas cases [14]. Velazquez 1993 [9] supports this directly in rats: intravenous trimethoprim increased renal sodium excretion by 46% (CI 9% to 83%) while reducing potassium excretion by 40% (the reconstruction tool warned the sodium direction was paired from an unordered index, so confirm against the original); human tubular sodium handling on the combination was not measured in any source read. A case report discusses mechanism in more detail (title-level only) [16].

Why orange, not red

One cohort with directly measured sodium and a large incidence (72.3%) plus a rechallenge-confirmed case literature support a real, bounded effect (high dose, hospital setting). Repair-pass re-check (2026-10-08): a second independent measured-sodium cohort now exists (Stephens, children, 30.3% [43]) and a pharmacovigilance signal against azithromycin [42], and the direction is consistent, so the entry is close to red. It stays orange because (a) no study compares sodium on co-trimoxazole with an unexposed or active-comparator group using measured sodium, (b) both cohorts are high-risk inpatients with strong confounders (Stephens: 80% of hyponatremic courses on diuretics), (c) the established effect is bounded to high-dose or intravenous hospital therapy, with standard-dose data limited to case reports, and (d) the second cohort in Nagai [15] reports sodium only inside a composite. A comparator-controlled, measured-sodium cohort at standard dose would move it to red.

Counter-evidence held on record

Counter-evidence and confounding Confounding by indication is presumed: pneumonia, vomiting, diuretics and antidiuretic stimuli also lower sodium. Tsapepas excluded several such patients but had no control group. Most reports are of high-dose therapy. Framing caution This is a sodium-loss and fluid-balance disturbance. It is not a reason to eat more salt or take sodium tablets. New confusion, severe nausea, headache or seizures during a course need urgent assessment, and a prescriber decides about correction.

Limits on what this shows

• Key null findings: Smulders 2001 [21]: 34 HIV-positive adults (29 men, mean age 36.8) on primary prophylaxis at 80/400 mg once daily, homocysteine sampled twice before and twice during the first year (means 2.6 and 5.3 months); 13.9 +/- 3.7 versus 14.4 +/- 5.0 umol/L (difference 0.5, 95% CI -0.5 to 1.4, P = 0.34); folate 13.1 +/- 6.5 versus 13.3 +/- 5.3 nmol/L; baseline folate, age, sex and creatinine did not predict response (PubMed abstract re-read 2026-10-08; vitamin B12 is not in the abstract and the full text is closed access with the publisher page blocked to automated reading). Uncontrolled before-and-after design. This contradicts a folate effect at the standard prophylactic dose over about 1 year and is the main reason the entry is bounded. Poskitt also found no hematologic folate deficiency [19]. The Bateson radioisotope assay did not show a fall [20].
• At-risk subgroups: Per the label: elderly, preexisting folate deficiency, kidney failure, chronic alcohol use, anticonvulsant therapy, malabsorption, malnutrition, pregnancy, and concurrent pyrimethamine, methotrexate or other antifolates [1][2]. Higher doses and longer courses [18].

Direction

DEPLETION-like (folate antagonism and lower serum folate). Sulfamethoxazole acts on a bacterial enzyme and no source read shows it lowers human folate; the human data are for trimethoprim.

Regulatory status

Label: trimethoprim "binds to and reversibly inhibits" dihydrofolate reductase (bacterial mechanism text); "Hematological changes indicative of folic acid deficiency may occur in elderly patients or in patients with preexisting folic acid deficiency or kidney failure", "reversible by folinic acid therapy"; avoid use with possible folate deficiency (elderly, chronic alcoholics, anticonvulsant therapy, malabsorption, malnutrition); contraindicated with documented megaloblastic anemia due to folate deficiency; "may interfere with folic acid metabolism" in pregnancy; avoid coadministration with leucovorin during Pneumocystis pneumonia treatment because of treatment failure and excess mortality [1][2].

Population evidence

Meidahl Petersen 2016 [17]: double-blind placebo-controlled randomized trial in healthy young men, trimethoprim alone 200 mg twice daily for 7 days; 58 analyzed (30 trimethoprim, 28 placebo); mean serum folate fell 1.95 nmol/L on trimethoprim (within-group P = 0.018) versus a rise of 0.21 on placebo (between-group P = 0.040); eight participants were folate-deficient at baseline and the share who were deficient rose significantly on trimethoprim (P = 0.034). Abstract through a summarizer; homocysteine and B12 not reported. Naderer 1997 [18]: randomized 10-day trimethoprim at 10 versus 20 mg/kg/day in 20 healthy volunteers (10 per group in period I; 9 and 8 analyzed for creatinine after dropouts; crossover period II had only 9 subjects, 10 completing moderate-dose and 7 high-dose phases); journal PDF read through a summarizing tool: serum folate on day 10 fell from 8.25 +/- 5.7 to 5.5 +/- 4.3 (P < 0.01) on 20 mg/kg/day, with no significant reduction at 10 mg/kg/day; the unit was printed as mg/dl in the text returned, which is almost certainly a typographical error, and the folate assay is not stated in the paper. Adverse reactions 75% at high dose versus 11% (P < 0.02). Poskitt 1972 [19]: 10 children with severe protein-calorie malnutrition given a 10-day course of the combination; formiminoglutamic acid (FIGLU) excretion after a histidine load was significantly higher than in a control group, with "no haematological evidence of folate deficiency". Abstract only; dose, control size and values not available. Bateson 1976 [20]: 13 inpatients given the combination for 6 to 14 days versus 8 untreated controls; serum folate by the Lactobacillus casei microbiological assay fell significantly in the treated group and not in controls, but a radioisotope assay showed no significant change, and the authors concluded the combination does not truly lower serum folate and that the microbiological readings probably reflect interference with the assay organism (PubMed record opened through a summarizing tool, 2026-10-08; values not in the abstract). This is the only direct serum-folate measurement of the fixed combination found, and it is null on the non-microbiological assay. Smulders 1998 [49]: a Lancet letter (PMID 9851389) on trimethoprim and fasting plasma homocysteine in healthy volunteers; the PubMed record has no abstract, and the 2001 paper [21] describes it as showing a substantial rise in fasting homocysteine on a short trimethoprim course. Numbers not retrieved; supportive of a functional folate-pathway effect of trimethoprim alone at high dose, not used for grading. Pregnancy outcomes (clinical outcomes, not folate measurements) : Hernandez-Diaz 2000 [23] (folic acid antagonists including trimethoprim: cardiovascular defects RR 3.4, 1.8 to 6.4; oral clefts 2.6, 1.1 to 6.1) and 2001 [24] (neural tube defects: any folic acid antagonist adjusted OR 2.8, 1.7 to 4.6; trimethoprim 4.8, 1.5 to 16.1, 5 exposed cases) versus Hansen 2016 [25] (7,615 exposed infants, no statistically significant increase in anomalies against penicillin or cephalosporin or no antibacterial exposure). Listing-level reads. Case reports of megaloblastosis and pancytopenia [22]: Chanarin 1972 (4 patients with megaloblastic hematopoiesis, trimethoprim suppressed reticulocyte response in 3, pancytopenia in 2), Blackwell 1978 (3 patients, 2 died), Hill 1973, Chan 1980, Kahl 2005 (low folate, normal B12, ferritin and iron). Title or listing level.

Mechanism

Present and named: reversible inhibition of dihydrofolate reductase blocks tetrahydrofolate production; the label states this for the bacterial enzyme, and the human serum-folate fall with trimethoprim alone [17][18] and the label's reversal by folinic acid [1] support a human-cell effect at sufficient exposure.

Why orange, not red

Two randomized, directly measured serum-folate studies of trimethoprim alone [17][18] and a functional-marker study of the combination [19] agree in direction, with a named mechanism. Red was held back because the measured change is small (1.95 nmol/L) and dose-dependent, a Tier A study at the standard prophylactic dose found no folate or homocysteine change [21], one assay-dependent result is discordant [20], and the one direct serum-folate measurement of the combination product (Bateson 1976 [20], now read at abstract level) was null on its radioisotope assay and its authors concluded the microbiological fall was assay interference. Repair-pass re-check (2026-10-08): Naderer now has retrieved numbers (8.25 to 5.5, P < 0.01, 20 mg/kg/day only), but both randomized serum-folate studies used trimethoprim alone at 400 mg/day or above, and the two Tier A studies that used the combination at or near standard dosing (Bateson, Smulders) did not show a folate fall, so the red criterion of no substantial contradicting Tier A evidence is not met. Overt deficiency is documented as case reports in already-deficient patients, which is a bounded subgroup, not a population effect.

Counter-evidence held on record

Counter-evidence and confounding HIV infection, malnutrition, alcohol use and other antifolate drugs confound the patient reports. The volunteer trials used trimethoprim alone, not the fixed combination. Pregnancy defects are clinical outcomes, with confounding by infection, and one large cohort found no significant increase [25]. Framing caution This entry does not recommend folic acid or leucovorin for people taking co-trimoxazole. The label says folinic acid reverses the hematologic changes but also warns against leucovorin during Pneumocystis pneumonia treatment because treatment failure and excess mortality were seen. Pregnancy and folate questions belong with the prescriber. No supplement dose is given here.

Direction

REVERSE (Hard Rule 4): zinc reduced sulfamethoxazole toxicity in the animal model. The drug is not shown to deplete zinc. This is not a supplementation recommendation.

Evidence

Wei 2023 [34]: zinc reversed sulfamethoxazole-induced mitochondrial and lysosomal dysfunction and lipotoxicity in a freshwater teleost, with SOD1 required; Tier B animal study. Human search [35]: PubMed "cotrimoxazole zinc" (10 records) found none measuring zinc as an outcome of cotrimoxazole; the records were Cochrane diarrhea and IMCI guidance where zinc is an intervention, and unrelated chemistry.

Mechanism

Named only in the fish model (SOD1, STAT3 and lysosomal pathways); no human mechanism.

Why yellow

One Tier B animal study with this drug; no human or label support. Kept as a reverse-direction context flag.

Counter-evidence held on record

Counter-evidence and confounding Ecotoxicology concentrations and a non-mammalian species; no relevance to treatment dosing is shown. Framing caution Do not infer a need for zinc from this entry. No dose is given.

Limits on what this shows

• Why no grade: Dual condition met on the channels that worked: the closest records are folate-pathway outcome reports covered in the Folate entry, and no Tier B or C source measures B12. This is a disclosed gap, not a finding of safety.

What was searched

PubMed "trimethoprim vitamin B12 cobalamin" (10 of 23 records) [38]: Hill 1973 and Chanarin 1972 concern megaloblastic patients where B12 was treatment, not a measured exposure outcome; Bernstein 1975 is a review whose indexing mentions blood B12 but the abstract reports no values; Poskitt 1972 measured folate-dependent FIGLU, not B12. Kahl 2005 [22] found normal cobalamin alongside low folate in one patient. Label pages: no B12 statement [1][2]. Volunteer and prophylaxis studies read [17][18][21] did not report B12 in their abstracts. Repair pass 2026-10-08 : PubMed "cotrimoxazole vitamin B12 serum levels" (3 records) [48]: Malik 2009 (103 HIV-infected children, 16 on cotrimoxazole prophylaxis, all with B12 within or above normal; B12 is not reported by exposure and cotrimoxazole was not in the multivariate model for elevated B12); McAdam 1993 (lymphoma chemotherapy with co-trimoxazole prophylaxis, normal B12 and folate, no numbers); De Schoenmakere 1998 (dialysis erythropoietin trial where B12 and co-trimoxazole intake did not differ between groups). None reports B12 stratified by co-trimoxazole exposure. PubMed "trimethoprim homocysteine" [51] found Smulders 2001 [21] (homocysteine and folate, no B12) and a 1998 Lancet letter without abstract [49]. The Smulders 2001 abstract does not report B12 and its full text is closed access (publisher page blocked to automated reading). Label term search for B12 or cobalamin: none [1].

Channel outcomes

Web search exhausted; no consumer claim could be sought, so a "B vitamins lost through gut flora" claim, if it exists, was not located. Tried: label term search, three PubMed phrasings, OpenAlex lookups for Smulders 2001 and Meidahl Petersen 2016 (no abstract text for the former, abstract only for the latter). The full texts of Smulders 2001 and Meidahl Petersen 2016 remain unopened and might report B12. Tooling-limited as much as a literature gap.