Updates undertaken since publication of the original review resulted in the following changes to the protocol.

1. Significant updates to the methods to bring the review in line with current Cochrane (MECIR) standards. This included insertion of a 'Summary of findings' table and updates to the risk of bias methods (removal of the Jadad scale). This process included evaluation of the following additional domains of potential bias: imbalance of outcome measures at baseline, comparability of intervention and control group characteristics at baseline, protection against contamination, and selective recruitment of participants.

2. Following critical feedback regarding this review, we amended the definition of AHRF from the initial review to now include reference to pH < 7.35. We deemed this essential to enhance its consistency with clinical practice.

3. Amendment of primary review outcomes. In the initial version, the primary outcome was 'treatment failure', defined according to any combination of intubation, mortality, or treatment failure. To remove ambiguity regarding overlap between these definitions, the current version lists treatment intolerance as an independent secondary review outcome. We also made clarifications to the definition of our primary outcome 'need for endotracheal intubation' to minimise ambiguity.

4. Exclusion of one study that was included in the original review (Conti 2002) due to inclusion of mechanical ventilation as a comparator intervention; and reclassification of one study originally included in the review (Servillo 1994) to 'Awaiting classification' due to lack of sufficient information pertaining to mean baseline pH to determine eligibility for inclusion in the review in accordance with our revised definition of AHRF.

5. Replacement of one abstract included in the original review with a full‐text version in the update (Khilnani 2002 is now Khilnani 2010).

6. Minor edits to wording of two eligibility criteria to explicitly clarify that (a) studies of patients who commenced NIV before hospital admission are ineligible for inclusion in the review; and (b) 'usual medical care' may not include any form of positive pressure ventilation considered to be 'usual' for that study centre.

7. Thys 2002 was originally included as an 'ICU' study in the subgroup analysis of ward versus ICU care. As this study was based in an emergency department (potentially more intensive than a ward, but not an ICU setting), in keeping with the primary aim of the review, we excluded this study from the ward versus ICU subgroup analysis for the present (and subsequent) updates.

8. Data from Bott 1993 were originally included in some of the meta‐analyses related to blood gas tensions; however owing to lack of sufficient study information regarding the number of participants included in these outcome data, this study was subsequently removed from the meta‐analysis and was reported separately. Additionally, mortality data were removed from the meta‐analysis, as they were identified as relevant to the post‐discharge period, not to in‐hospital mortality.

9. We conducted subgroup analyses of admission pH and location only for the primary outcomes of interest for this review (mortality and need for endotracheal intubation), rather than for all outcomes.

10. One additional table (percentage change in PaCO₂ at one hour post intervention) included in the earlier review was removed.

11. Type of NIV for subgroup analysis was removed from the investigation of heterogeneity section of the protocol, as all studies except one used pressure‐cycled NIV.