Attari 2017.

Methods	Allocation: randomised, stratified by baseline PANSS scores Blinding: triple Design: parallel Duration: 10 weeks
Participants	Diagnosis: schizophrenia (DSM‐IV‐TR, clinical interview) n = 60 Age: mean 33 years, SD 8 Gender: 39 men, 21 women History: 2 years since onset of the disease. Patients with acute symptoms referred to psychiatric clinic or emergency department. Inclusion criteria: aged 18–65 years, 2 years since onset Exclusion criteria: unwillingness to participate; failure to follow‐up for whatever reason; unstable medical illness and medical history; contraindications for use of aspirin: asthma or seasonal allergies, ulcers, kidney disease, active bleeding or clotting of blood disorders such as haemophilia or bleeding, gout, nasal polyps, chronic use of NSAID, concomitant use of corticosteroids for any reason; maternity. Setting: psychiatric clinic, single centre Country: Iran
Interventions	1. Aspirin 325 mg/day + regular antipsychoticsa, omeprazole 20 mg/day. n = 20b 2. Aspirin 500 mg/day + regular antipsychoticsa, omeprazole 20 mg/day. n = 20 3. Placebo oral tablet daily + regular antipsychotics, omeprazole 20 mg/day. n = 20
Outcomes	Mental state: PANSS total, negative, positivec Adverse events: any Leaving the study early: any reason Unable to use: Mental state: PANSS general psychopathology (not validated subscore)
Notes	a Chlorpromazine, risperidone, olanzapine, haloperidol bGroup receiving aspirin 325 mg is assumed to have a typo in Table 1: it says 20 male participants, while all other numbers (total, percentages in gender distribution) point to only 12 male participants. We assumed 12 participants for this group throughout this review. cPANSS subscale scores did not always add up to the printed total in the tables. We contacted authors for clarification but at the time of writing this report have not received a reply.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The patients, through a number table, were randomly divided", "stratified randomization was used", "the randomization sequence was computer‐generated, with the randomization itself conducted through SPSS20 software", "Randomization was done by one of the researchers, who did not have a role in the treatment". Comment: Table 2 indicated that baseline PANSS scores differed significantly between the control and treatment groups. However, the SDs of the groups overlapped.
Allocation concealment (selection bias)	Unclear risk	Quote: "Allocation concealment was done by the researcher, who was responsible for the randomization. For this purpose, the numbered envelops that contained the name of the drugs … were used", "participants were referred to the hospital's pharmacy to obtain their drugs", "A resident of psychiatry generated the random allocation sequence and enrolled participants, and a co‐worker psychologist assigned participants to their interventions". Comment: unclear if opaque, sealed envelopes used.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote: "placebo tablets had the same shape and color of the effective aspirin", "Drug and placebo were coded A, B and C", "Neither the examiner nor the clinician and the patient were aware of the drug compounds". Comment: see Materials and methods section.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "The PANSS scale was administered on the first day, the end of the sixth week and one month after cessation of aspirin or placebo", "The outcomes of the study were recorded by the psychiatrist of the psychiatry ward, who made no other contribution to the study". Comment: no indications that adverse effects led to detection of allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No participants lost to follow‐up.
Selective reporting (reporting bias)	Unclear risk	Outcomes reported as stated in protocol, but retrospective registration of protocol.
Other bias	High risk	Reporting bias: inconsistent reporting of values between number tables, minor mistakes with numbers not adding up correctly (see 'Notes' in Characteristics of included studies table). Authors were contacted to clarify the typographical errors, however, there was no response. No adverse events recorded for aspirin or antipsychotics, which is not impossible, but unlikely.