# Notes for the third reader: points the screening codebook left open

The two coders raised the same questions again and again in their reports. They were written down here before any trial on which the coders differed was decided, and the third reader decides those trials by the codebook and these notes. Where the two differ, these notes govern.

1. **Knowledge of products.** Use what is generally known of a product's composition whether or not it is licensed, when you are confident of it (the record names the product or its maker's code, and its platform and adjuvant are published). Otherwise give platform `other` and adjuvant `unknown`.
2. **Platforms.** A live organism given as the vaccine against itself is `live_attenuated` even when it is not attenuated or cannot replicate (oral adenovirus types 4 and 7, radiation-attenuated sporozoites, single-replication influenza virus, replication-defective cytomegalovirus). A replicating virus that carries another pathogen's antigen (measles-vectored chikungunya, parainfluenza virus carrying an RSV or metapneumovirus protein) is `viral_vector`. Killed whole bacteria given by mouth or under the tongue to prevent infections are `inactivated` vaccines. Outer-membrane vesicles are `protein`.
3. **Adjuvants given beside the vaccine.** An adjuvant mixed with or given alongside the vaccine (CpG, GLA, a cytokine plasmid) is the vaccine's adjuvant, coded `other`, not a separate active product.
4. **One group, two possible comparisons.** When a group could serve in two comparisons, take the comparison that tests the trial's study vaccine, the one its title and conditions name, and leave the other out.
5. **Two kinds of placebo.** When a trial has both an inert placebo group and a placebo group holding adjuvant for the same vaccine groups, take the inert placebo groups and leave the adjuvant ones out. When the vaccine groups divide with them (adjuvanted vaccine with adjuvant placebo, plain vaccine with plain placebo), make two comparisons.
6. **The same participants in several groups.** Take the groups that begin at the first vaccination and count each participant once: the group spanning the whole follow-up if there is one; else, for groups by period, the first period; for groups by dose, the first dose; for a challenge trial, the groups before the challenge.
7. **Vaccination of the placebo arm inside a group.** If the only groups run past the point where placebo recipients received the study vaccine, the trial is ineligible (`groups_not_randomized_arms`). A later dose that only the vaccine arm received, inside its randomized group, changes nothing.
8. **Controls.** A skin-test antigen or any other immunologically active product is not a placebo. A placebo "matching" the vaccine with no composition given is `not_stated`. A "vaccine diluent" or "formulation buffer" is `inert` unless the record says it holds adjuvant. A placebo arm given both a placebo of unstated composition and saline is `not_stated`.
9. **Active-control arms.** In a trial that also has a placebo arm, an arm given a licensed vaccine as an active control is a vaccine group if it differs from the placebo arm only in that vaccine: this study is of vaccines of every kind.
10. **Preventive or not.** A vaccine given to people who already have the infection or lesion it targets, to control it or stop it coming back, is not preventive (HPV vaccine around the treatment of lesions or warts, a cytomegalovirus vaccine in seropositive transplant patients, a vaccine against recurrent candidiasis, a tuberculosis vaccine given to stop relapse after treatment). A vaccine given to people at risk who do not have the infection is preventive.
11. **A drug, supplement, antibody or stimulant tested for its effect on a vaccine every arm gets:** `vaccine_in_every_arm`.
12. **Targets.** `targets` is the disease the trial gave the vaccine to prevent, with `off_target` true when that is not the vaccine's own disease.
13. **Patients.** `patients` is true when the trial, or one of its randomized strata, was designed to enroll people with a medical condition (HIV infection, a transplant, a chronic disease as an entry criterion, prematurity). It is false when such people were merely allowed in, and false for a past infection or a behavioural risk.
14. **Pregnant and other women in one trial.** Make two comparisons if the tables report them apart; `pregnant` is true if any comparison is of pregnant women. Infants of vaccinated breast-feeding mothers count as offspring.
15. **Titles and descriptions that disagree.** Go by the descriptions; if that doesn't settle which group got placebo, the trial is ineligible (`other`).
16. **Follow-up records** of an earlier randomized trial are eligible when their groups are the randomized arms and the placebo arm had not been vaccinated by the end of the period they cover.
17. **Cohorts with no placebo group of their own** are left out.
18. **Unequal follow-up.** If the record says the placebo groups' adverse events were collected over a much shorter time than the vaccine groups', the trial is ineligible (`other`).
