Davies 1997.

Methods	Double‐blind, placebo‐controlled, parallel group RCT
Participants	17 male, 7 female, age 18 to 90 years Eligibility criteria: community‐acquired pneumonia, sepsis and a parapneumonic effusion as defined by the following criteria: purulent fluid with evidence of bacteria on microscopy or culture, pH < 7.1, lactate dehydrogenase > 1000 IU/L, pleural fluid/blood glucose ratio < 0.25 with loculation/septation of pleural fluid on CT. Exclusion criteria: previous treatment with streptokinase within the previous 2 years, bleeding diathesis, and significant haemorrhage or stroke within the previous 6 months, or disease making survival < 2 months unlikely. Participants all received antibiotics which were appropriate for the organisms cultured, or broad‐spectrum antibiotics to cover gram‐positive, gram‐negative and anaerobic organisms.
Interventions	Daily intrapleural streptokinase 250,000 IU for 3 days versus normal saline. Intervention group: intrapleural streptokinase 250,000 IU via a 14 French van Sonnenberg catheter in 20 mL normal saline on 3 consecutive days. Control group: 20 mL normal saline in the control group. The drain was then clamped for 2 hours. Both groups received a background of 6‐hourly 20 mL normal saline flushes until the intercostal catheter was removed. The catheters were inserted into the most dependent portion of the effusion or into the largest loculation. Continuous suction of −20 cmH₂O was applied. The intercostal catheter was removed after the 5th day and when the amount of fluid drained was < 150 mL for 2 consecutive days. Further aspiration or catheter drainage was at the discretion of the admitting physician. Follow‐up: 3 years.
Outcomes	Days to defervescence Duration of hospital stay Duration of pleural drainage Pleural fluid amount over first 3 day's treatment Improvement in chest radiograph Referral for surgery Death Overall treatment failure Side effects of treatment Haemorrhagic complications
Notes	Parapneumonic effusion only. Funding sources not identified in published manuscript.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation sequence (author correspondence).
Allocation concealment (selection bias)	Unclear risk	Information not available.
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind author correspondence.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Radiographs scored by blinded radiologists. Not specified for other outcomes.
Incomplete outcome data (attrition bias) All outcomes	Low risk	"Withdrawals" were referrals for surgery — an outcome.
Selective reporting (reporting bias)	Low risk	All outcomes described in Methods fully reported. Original protocol unavailable to confirm original primary outcomes.
Other bias	Low risk	Nil identified.