Table 2

Data to extract from publications included in the systematic review

Type of data	How to code the data	Comments
Data to extract on each reason mention
Content	Assign to each reason mention: A broad reason type, for example, avoid exploitation A narrow reason type, for example, avoid exploiting research participants or avoid exploiting the host country.	To minimise bias, it is best to assign reason types to a passage based only on the words in the passage, because different reviewers can agree on which words are used. An exception is when publications use the same word, for example “reciprocity” to express different reasons, such as reciprocity and distributive justice. See the text below this table (Content). One should think about the review's purpose when deciding how broad to make reason types, whether types should be mutually exclusive, and whether mentions should be assigned more than one narrow (or broad) reason type. For example, if one purpose is to prompt examination of the relations between reasons that are considered distinct in the literature, different mentions should be assigned to different types whenever the suspicion arises that the types may differ. Whenever possible, reason types should be given short, self-explanatory names in order to keep lists of reasons accessible to decision-makers.
Alleged implications	Assign to each reason mention a number code depending on whether it is, for example, allegedly for ensuring PTA to the trial drug, or allegedly against ensuring this, or the mention is claimed to have implications for ensuring PTA, but the implications are unspecified or unclear.	Reasons against ensuring PTA, for example, may include reasons why PTA need not, or should not be provided. Most reviews will need to extract data that reflects such a distinction. We did this by extracting from each reason mention the conclusion drawn from the reason mention, as distinct from the all-things-considered conclusion drawn by the publication (see below).
Optional: person(s) expressing attitude to reason mention	Code each reason mention depending on whether the person expressing attitude to reason is The author, or Another party whose view the author reports If 2, consider recording the other party's name if given, and any relevant reference.	Extracting mentions to which other parties express an attitude allows reasons to be collected that are not directly published in the literature reviewed. Recording the other party's name and any relevant reference enables identification of the publications that are more frequently discussed, endorsed or rejected.
Attitude taken	Assign a number code to the attitude taken to the reason mention according to whether the reason is always endorsed the reason is sometimes endorsed it is unclear whether the reason is endorsed or rejected the reason is rejected. Example: assign 2 to a reason mention coded as reciprocity when the author asserts that reciprocity supports giving PTA in some contexts but in other contexts fails to apply.	Where the author's attitude to a reason changes in the course of a publication, context and order should be used to decide which attitude is ‘authoritative’. A limitation is that rejected reasons receive the same attitude code, irrespective of whether they are considered irrelevant or based on incorrect factual or moral claims. The review's purpose should be considered when deciding how to balance removal of these limitations against keeping data accessible.
Any conclusion drawn from the reason mention	For each reason mention, note the conclusion drawn from it (as distinct from the all-things considered conclusion that the publication draws based on all the reasons it considers). Consider developing codes for conclusions that capture parts of the conclusion. Our review used codes to capture: The PTA-related action, for example funding PTA, making a pre-trial plan to ensure PTA Whether the action is said to be permissible, forbidden or required The agent held responsible for the PTA-related action (eg, researchers).	It is particularly important to extract the conclusion drawn from the reason mention when mentions of the same reason differ as to the conclusion drawn. In the literature we reviewed, some conclusions drawn from reasons concerned whether there are moral obligations to ensure PTA, others whether it should be legally required. Reviewers will need to decide how narrow conclusion codes should be. There is a trade-off between ensuring that aggregate statistics are accessible, which requires broader codes, and meaningful. On the one hand, the latter tends to favour narrow codes but, on the other hand, if they become too narrow then the aggregate statistics lose their meaning.
Optional: reference	For each reason, consider noting whether an appropriate reference is given where one is necessary. Criteria will be needed. We considered references necessary except when reasons occur in the titles or abstract, or were clearly original when published. Code on the basis of whether the reason falls under the scope of an appropriate reference, rather than on whether there is a reference in the quotation extracted.	While collection of data on references is optional, it is particularly useful when writing a review that will enable academics to understand the state of a literature, as absence of appropriate references suggests (as in the case of our review) that some proponents of a view are ignorant of relevant publications. Be sensitive to different referencing conventions, and exercise caution when assessing the originality of reasons. Include appropriate disclosures in the limitations section.
Data to extract on the publication
Any conclusion drawn by publication from all the reasons considered	Same as for conclusion drawn from the reason, except that the conclusion extracted is the all-things-considered conclusion that the publication draws on the basis of all the reasons it considers, as distinct from an interim conclusion that it draws from just one of these reasons.	Same as the comments on extracting conclusion drawn from the reason mention (see above).
Publication type	Collect various data on the publications included in the systematic review. For example: Broad (or narrow) type of publication for example, article, monograph (an example of a narrow type of publication is a policy review) Any restrictions on scope (eg, to HIV research) Reference Consider developing a measure for, and collecting data on, authors' conflict of interest	When possible, give eligibility criteria for types. In our first systematic review of reasons we also collected data on: Whether PTA to trial drug is publication's key topic (Y/N) Whether the publication's PTA content exclusively concerns research conducted by resource-rich country sponsors in resource-poor or middle-income countries (Y/N)

PTA, post-trial access.