Compounds with a graded row

Ordered by rung, strongest first. The rung rates the study design; the chip beside it says what the study measured. Both are needed.

1. Bactericidal/Permeability-Increasing Protein (BPI)

   human RCT Direct outcome Not approved for this condition

   Who was studied

   children aged 2 weeks to 18 years with severe meningococcal sepsis, n=393 (190 rBPI21, 203 placebo), 22 UK and US centres, outcomes to day 60

   A recombinant N-terminal fragment, rBPI21, was given intravenously on top of standard care and the trial measured mortality, amputations and 60-day functional outcome. Mortality was 7.4% with rBPI21 and 9.9% with placebo, a difference that did not reach statistical significance (odds ratio 1.31, 95% CI 0.62-2.74, p=0.48). No regulator has approved BPI for any infection.

   Source thelancet.com 1 primary source read for this row

2. Magainins

   human RCT Direct outcome Not approved for this condition

   Who was studied

   adults with mildly infected diabetic foot ulcers; randomised, double-blind, multicentre, 14 days of twice-daily topical treatment. Enrolment total not confirmed from the abstract read.

   Pexiganan, a synthetic magainin analogue, was compared as a 0.8% cream against oral ofloxacin, with clinical improvement, microbiological eradication and wound healing as endpoints; the two arms were equivalent, and later placebo-controlled phase 3 work did not show superiority over placebo. The FDA has not approved pexiganan, and no naturally occurring magainin has been tested in humans.

   Source PubMed 2 primary sources read for this row

3. LL-37 (Cathelicidin)

   human RCT Surrogate marker Not approved for this condition

   Who was studied

   adults aged 18-60 with uninfected or mildly infected diabetic foot ulcers, n=25 (13 LL-37, 12 placebo), 4 weeks

   The primary endpoint was granulation tissue formation, a wound-healing measure; aerobic bacterial colony counts on the ulcer surface were a secondary measure and LL-37 did not significantly reduce them. No trial on file measured resolution of an established infection, bacterial cure, or survival with LL-37 in humans.

   Source PubMed Central 1 primary source read for this row

4. Lactoferricin

   human RCT Mechanistic only In registered trials

   Who was studied

   healthy volunteers (sequential randomised double-blind ascending single doses 0.005-5 mg and multiple intravenous doses 0.5-5 mg), plus open-label single 5 mg intravenous doses in autologous haematopoietic stem cell transplant recipients

   Recorded, and not evidence for this condition

   The lactoferrin-derived peptide hLF1-11 was taken into humans, but this trial measured only safety and tolerability: no infection outcome - not cure, not bacterial clearance, not survival - was measured in any living subject, and reversible transaminase elevations were the main finding. No study on file shows hLF1-11 or lactoferricin treating an infection in a person.

   Source NCBI 1 primary source read for this row

5. Bacteriocins

   animal in vivo Direct outcome Not approved for this condition

   Who was studied

   lactating Holstein cows with subclinical mastitis, n=90 (46 nisin Z, 44 untreated control), intramammary infusion of 2,500,000 IU once daily for 3 days

   Bacteriological cure was 65.2% (30 of 46) with nisin Z against 15.9% (7 of 44) spontaneous recovery in controls, with falls in milk NAGase activity and somatic cell count. Nisin Z is the only bacteriocin on file with an infection outcome measured in a living subject, and the subject was a cow; no bacteriocin has an infection outcome measured in humans.

   Source PubMed Central 1 primary source read for this row

6. Cecropins

   animal in vivo Direct outcome Not approved for this condition

   Who was studied

   female C57BL/6 mice aged 6-8 weeks, lethal intraperitoneal Escherichia coli ATCC 25922 challenge; n=6 per group for survival, n=4 per group for bacterial load

   DAN2, a designed cecropin-like peptide, was given 30 minutes after challenge and measured survival and bacterial counts in blood and peritoneal fluid: all mice given 20 mg/kg survived five days against no controls surviving 12 hours. The peptide tested was an engineered analogue rather than a natural cecropin, and no cecropin has been given to a person in any study on file.

   Source PubMed Central 1 primary source read for this row

7. Dermcidin

   animal in vivo Direct outcome Not approved for this condition

   Who was studied

   female BALB/c mice aged 6-8 weeks, 18-22 g, n=5 per group; 5 mm dorsal wounds infected with Acinetobacter baumannii, topical DCD-1L at 8x MIC once daily for 10 days

   Wound bacterial counts fell by 1.15, 2.80 and 5.32 log10 CFU/mL at days 1, 5 and 10 with faster wound closure - an infection endpoint, but in mice, with the peptide applied to the wound. Dermcidin is normally an endogenous sweat peptide; the human literature describes its presence and its in vitro killing, and no study on file has administered it to a person.

   Source PubMed Central 1 primary source read for this row

8. Melittin

   animal in vivo Direct outcome Not approved for this condition

   Who was studied

   immunosuppressed BALB/c mice with peritoneal infection by extensively drug-resistant Acinetobacter baumannii, MRSA, or KPC-producing Klebsiella pneumoniae; repeated sub-lethal melittin doses of 2.4 mg/kg

   Melittin inhibited all three pathogens in culture (MIC 8-32 ug/mL), but in infected mice repeated dosing at the highest tolerated level showed no benefit on survival or peritoneal bacterial load. The limit was toxicity: haemolysis HD50 was 0.44 ug/mL and the murine LD50 was 4.98 mg/kg, so the concentrations that kill bacteria in a dish are not reachable in a living animal.

   Source PubMed Central 1 primary source read for this row

9. Nisin

   animal in vivo Direct outcome Not approved for this condition

   Who was studied

   lactating dairy cows with clinical mastitis, intramammary nisin compared with gentamicin

   Clinical cure was 90.2% with intramammary nisin and 91.1% with gentamicin, with bacteriological cure of 60.8% and 44.6% respectively - an infection endpoint, but in cattle udders, by direct infusion. Nisin's long safety record as a food preservative is a food-additive record and is not evidence that it treats infection in a person; no human infection trial is on file.

   Source PubMed 1 primary source read for this row

10. Proline-rich Antimicrobial Peptides (PrAMPs)

    animal in vivo Direct outcome Not approved for this condition

    Who was studied

    mice with systemic Salmonella typhimurium infection treated with Bac7(1-35) at 30 mg/kg

    The bovine proline-rich peptide Bac7(1-35) measured survival and organ bacterial load, raising mean survival from 10 days in untreated controls to 24.5 days, with rapid renal clearance limiting the effect. This is a mouse result; no proline-rich antimicrobial peptide has an infection outcome measured in humans on file.

    Source NCBI 1 primary source read for this row

11. Temporins

    animal in vivo Direct outcome Not approved for this condition

    Who was studied

    female BALB/c mice aged 8-10 weeks, systemic Staphylococcus aureus A170 and Salmonella enterica serovar Paratyphi infection models, immediate and delayed (day 7) treatment

    Temporin A combined with a modified temporin B measured survival and bacterial load in kidneys, liver and gut: treated mice survived to day 28 while untreated mice died within 4-6 days, and organs were cleared within 3-6 days. All of this is in mice; no temporin has been given to a person in any study on file.

    Source PubMed Central 1 primary source read for this row

12. Defensins

    in vitro Mechanistic only Not approved for this condition

    Who was studied

    multidrug-resistant nosocomial clinical isolates of Staphylococcus aureus, Enterococcus faecium, Pseudomonas aeruginosa, Stenotrophomonas maltophilia and Acinetobacter baumannii; bacterial cultures only, no living subject

    Recorded, and not evidence for this condition

    Human beta-defensin 3 killed these isolates in culture within about 20 minutes, and the work also documents how salt in physiological buffers blunts defensin killing. This was measured in cultures; no study on file administered a defensin to an animal or a person and measured an infection outcome, so the link to treating an infection is an argument from mechanism, not a result.

    Source PubMed Central 1 primary source read for this row

13. PGLa

    in vitro Mechanistic only Not approved for this condition

    Who was studied

    Escherichia coli K12 cultures and model lipid bilayers; no living subject

    Recorded, and not evidence for this condition

    Equimolar PGLa with magainin 2 lowered the minimum inhibitory concentration against E. coli K12 by roughly tenfold, and the work traces this to a heterodimer forming in the membrane. Everything here was measured in culture and in synthetic membranes; no study on file gave PGLa to an animal or a person and measured an infection outcome.

    Source PubMed Central 1 primary source read for this row

14. Psoriasin (S100A7)

    in vitro Mechanistic only Not approved for this condition

    Who was studied

    cell-free growth-inhibition and microdilution assays against Escherichia coli, Pseudomonas aeruginosa, Staphylococcus aureus, Listeria monocytogenes and Lactobacillus plantarum; no living subject

    Recorded, and not evidence for this condition

    Oxidised S100A7 inhibited growth of a zinc-uptake-deficient E. coli mutant, L. monocytogenes and L. plantarum but not wild-type E. coli K-12 or P. aeruginosa, consistent with metal sequestration rather than lysis. This is an endogenous epithelial protein studied where it already occurs; nothing was administered, and no study on file measured an infection outcome in an animal or a person.

    Source PubMed Central 1 primary source read for this row

15. RNase 7

    in vitro Mechanistic only Not approved for this condition

    Who was studied

    cultures of Enterococcus faecium and other skin organisms, plus human skin extracts treated with an anti-RNase 7 antibody; no living subject

    Recorded, and not evidence for this condition

    Skin-derived RNase 7 killed E. faecium in culture independently of its ribonuclease activity, and blocking it with an antibody reduced the killing capacity of skin extracts. That is observational biology about a protein the skin already makes; no study on file administered RNase 7 to an animal or a person, and no infection outcome was measured in a living subject.

    Source PubMed Central 1 primary source read for this row

Checked, and nothing recorded

1 compound was checked against bacterial infection and produced no gradeable row.

• Histatins — checked, no row on file

A compound is here because someone looked and the sources did not support a row — not because nobody looked. Under the rubric, uncertainty resolves to no row, and a row resting on a vendor page or a clinic blog is worse than no row at all.