Celikel 1998.

Methods	Country: Turkey. Design: Single‐centre prospective randomised controlled parallel‐group trial conducted between March 1993 and November 1996 Study site: Single university hospital in Istanbul, Turkey Method of analysis: Mann‐Whitney U test, ANOVA, log rank test, Chi2 test Aim: To compare the efficacy of standard medical therapy and NIV in patients with AHRF due to AECOPD
Participants	Eligible for study: Not stated Recruited: 30 adult patients (15 in each group) with AHRF due to AECOPD Completed: Data from 30 participants available for some outcomes Age: Not stated Gender: Not stated Criteria used to define COPD: Previous PFTs (FEV/FVC < 75% and < 12% bronchodilator response) or clinical history, physical examination, chest radiography, and ABGs (arterial CO2 retention, elevated bicarbonate) Inclusion criteria: Known to have COPD diagnosed on the basis of previous PFTs (FEV/FVC < 75% and < 12% bronchodilator response) or clinical history, physical examination, chest radiography, and ABGs (arterial CO2 retention, elevated bicarbonate), as well as (1) PaCO2 > 45 mmHg and pH < 7.35; and (2) evidence of respiratory muscle fatigue (RR > 22 breaths/min, accessory muscle use, and respiratory distress via direct observation of ICU staff) Exclusion criteria: Need for urgent intubation due to respiratory arrest, haemodynamic instability (systolic BP < 90 mmHg), severe cardiac arrhythmia, abundant secretions, myocardial infarction or cardiac arrest within last 3 months, and unwillingness to participate in the study
Interventions	Intervention description: Usual care plus continuous NIV. Pressure support ventilation (PSV) was delivered via mechanical ventilator (Model 720; Puritan‐ Bennett; Carlsbad, CA) and full face mask (Dryden, Clear Comfort Face Mask; Gibeck Respiration; Uplandsvasby, Sweden; and 9000; Vital Signs Corp; Totowa, NJ). Initial settings: PSV 15 cmH2O; PEEP 5 cmH2O; sensitivity 1 cmH2O; FiO2 0.5; active apnoea backup. Setting adjustments: PS to achieve 5 to ‐7 mL/kg expired TV, FiO2 to maintain SpO2 90% to 92%, and sensitivity as low as possible with no auto‐triggering Control description: Oxygen (min 1 L/min to keep SpO2 90% to 2%), aminophylline infusion (to keep serum theophylline levels 8 to 15 mg/L), atropine (1 mg 4‐hourly), salbutamol nebuliser (2.5 mg 4‐hourly), IV methylprednisolone (40 mg 6‐hourly), antibiotics if indicated (cefuroxime or sulbactam‐ampicillin until culture results available) Duration of intervention: Mean duration of NIV was 26.7 hours (SD 16.1) Intervention delivery by: Not clear Setting: Pulmonary medicine directed critical care unit at a university hospital
Outcomes	Method of outcome data collection: Systolic and diastolic BP, heart rate, RR, ABGs (on room air), and PaO2/FiO2 ratio measured upon admission; at 30, 60, 90, 120, and 180 minutes; then every 3 hours thereafter Prespecified primary outcomes: Not clearly defined Prespecified secondary outcomes: Systolic and diastolic BP, heart rate, RR, ABGs (on room air), PaO2/FiO2 ratio, complications (abdominal distension, nasal bridge abrasion, aspiration), duration of mechanical ventilation, expired tidal volume, and minute ventilation. Mortality, treatment failure, and intubation were also reported Follow‐up period: Until hospital discharge
Notes	Study author contacted and additional information requested, without reply Funder: Not stated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Insufficient information available
Allocation concealment (selection bias)	Unclear risk	"Patients were randomised ... by the envelope method". Unsure if opaque or clear
Blinding of participants and personnel (performance bias) All outcomes	High risk	Owing to nature of the intervention, blinding was not possible. No sham NIV was used. However, outcomes reported were objective outcomes and were unlikely to be affected "This prospective, randomized, controlled but unblinded study"
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	An ICU physician on call, who was not participating in the study, assessed treatment failure according to participant progress. Effects of outcome blinding on other study outcomes less clear
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants' outcome data appear to be reported
Selective reporting (reporting bias)	Unclear risk	Study protocol not available. Unclear which outcomes were prespecified
Imbalance of outcome measures at baseline All outcomes	Low risk	No evidence of statistically significant differences at baseline; however PFT data not available for all participants
Comparability of intervention and control group characteristics at baseline	Unclear risk	No consideration of baseline factors in statistical analyses but no significant baseline differences reported
Protection against contamination All outcomes	High risk	Mortality data very likely to have been influenced by rescue cross‐over to NIV intervention. Participants in the standard therapy group with treatment failure were switched to NIV, then to mechanical ventilation if needed
Selective recruitment of participants	Unclear risk	Insufficient information available. Difficult to tell how many other potentially eligible participants may have been excluded from the study
Other bias	Low risk	No other sources of bias identified