# Screening codebook: which adverse-event groups compare a preventive vaccine with placebo

Each record is a randomized trial on ClinicalTrials.gov with posted results whose text names a vaccine and a placebo. You see its titles, conditions, phases, design, masking, age limits, arms with their descriptions and interventions, the time frame of its adverse-event tables, and the groups those tables are reported by (`ae_groups`), with no counts. Decide, for each trial, which adverse-event groups form clean comparisons of a preventive vaccine against placebo, and describe them.

Decide from the record and from what is generally known about a named, licensed product's composition (for example that Prevnar 13 holds aluminium phosphate). Don't look up results, publications, or other records, don't use the web, and don't open any file but the codebook and the input you were given.

## What counts

**A preventive vaccine** is a product given to induce immunity that prevents an infectious disease: any platform, licensed or investigational, any age, including vaccines given to pregnant women to protect their infants and vaccines given to prevent a disease other than their own (BCG against respiratory infections, for example).

These are not preventive vaccines, and a trial that tests only these is ineligible: vaccines given to treat an infection or disease the participants already have (therapeutic HIV, hepatitis B, herpes or HPV-lesion vaccines), vaccines against cancer, allergy, addiction, Alzheimer's disease or another non-infectious target, antibodies and immunoglobulins, adjuvants or immune stimulants tested alone, and drugs or supplements tested for their effect on a vaccine that every arm receives.

**A vaccine group** is an adverse-event group whose participants were randomized to receive one or more preventive vaccines that the placebo group did not receive. Any dose, formulation, lot, or schedule counts.

**A placebo group** is an adverse-event group whose participants were randomized to receive, in place of that vaccine, something the record calls placebo, saline, diluent, buffer, vehicle, or adjuvant alone, with everything else the same as the vaccine group: the same co-administered vaccines, the same background care, the same period. A group given a different vaccine as its control (a licensed vaccine against another disease, or an older vaccine against the same one) is an active control, not a placebo group, even if the record's arm type says "Placebo Comparator". A group given nothing, with no placebo, is not a placebo group.

**A comparison** is a set of vaccine groups and a set of placebo groups that differ only in the vaccine. Most trials have one. Pool doses, formulations, lots, schedules, age cohorts, countries and other subgroups of one design into one comparison. Make separate comparisons only when the background differs between sets of groups (study vaccine with a flu shot against placebo with a flu shot, and study vaccine alone against placebo alone), or for offspring (below). Each adverse-event group belongs to at most one comparison.

## Which groups to take

- **Count each randomized participant once.** When the tables report the same participants in several groups (by period, by dose number, or as a whole and again as subsets), take the set of groups that counts each randomized participant once and covers the longest placebo-controlled follow-up. Say what you chose in `note`.
- **Only the placebo-controlled period.** Leave out groups for a run-in, an open-label extension, a booster phase without placebo, a period after placebo participants crossed over to the vaccine, and groups that mix participants of both arms. In a trial with a planned crossover, the groups for the first period, before anyone crossed over, are eligible if the tables report them.
- **Re-randomization.** If participants were randomized again later (to a booster or placebo, say), prefer the first randomization. Take the later one only if the tables report no groups for the first, and say so in `note`.
- **Offspring.** In a trial that vaccinated pregnant women, groups of the women are one comparison and groups of their infants are a second comparison with `"offspring": true`.
- **Vaccine with another active product.** Leave out a group that got the vaccine together with a drug, antibody or immune modulator its placebo group didn't get.
- **Several vaccines.** If a trial tests different vaccines against one placebo, put all the vaccine groups in the comparison, each with its own platform and adjuvant.

When no clean comparison remains, the trial is ineligible: set `eligible` to false and give the reason.

## What to record for each comparison

**`placebo_content`**, what the placebo held, by the record:

- `inert`: saline, sodium chloride, water, a buffer (phosphate-buffered saline, tromethamine, histidine), a sugar solution, or the vaccine's diluent or excipients, with no antigen and no adjuvant.
- `adjuvant`: the record says the placebo holds an adjuvant (an aluminium salt, "alum", AS01, AS03, AS04, MF59, CpG, Matrix-M, ...) or another active part of the vaccine without its antigen (an empty viral vector, lipid nanoparticles alone).
- `not_stated`: called placebo with no composition given.

**For each vaccine group, `platforms`**, a list (usually one item) from:

- `live_attenuated`: a weakened organism that replicates, given as the vaccine against itself or as a chimeric attenuated virus (rotavirus, intranasal influenza, Zostavax, varicella, measles-mumps-rubella, yellow fever, the dengue vaccines CYD-TDV, TAK-003 and Butantan-DV, the chikungunya vaccine VLA1553, BCG, oral typhoid Ty21a, oral polio, oral cholera CVD 103-HgR, MVA or ACAM2000 given against smallpox or mpox).
- `viral_vector`: a virus engineered to carry another pathogen's antigen (Ad26, Ad5, ChAdOx1, MVA carrying inserts, rVSV-ZEBOV, ALVAC).
- `mrna`: messenger RNA or self-amplifying RNA.
- `dna`: plasmid DNA.
- `inactivated`: killed whole organisms or split virus (split or whole-virion influenza, inactivated polio, hepatitis A, whole-virion COVID-19 vaccines, killed oral cholera).
- `protein`: recombinant or purified protein, subunit, virus-like particle, toxoid, or peptide (surface-antigen influenza, hepatitis B, HPV, Shingrix, RSV prefusion F, Novavax, tetanus, diphtheria and acellular pertussis, meningococcal B, malaria RTS,S and R21).
- `polysaccharide`: plain polysaccharide or polysaccharide-protein conjugate (pneumococcal, meningococcal ACWY, Hib, typhoid Vi).
- `other`: anything else, or not clear from the record: say what in `note`.

A group that got two vaccines of different platforms lists both.

**For each vaccine group, `adjuvant`**:

- `aluminium`: the vaccine holds an aluminium salt, alone or with something else (AS04 counts here).
- `other`: an adjuvant with no aluminium (AS01, AS03, MF59, CpG 1018, Matrix-M, and the like).
- `none`: no adjuvant: stated, or a platform that carries none (live attenuated, viral vector, mRNA, DNA, plain polysaccharide, unadjuvanted influenza vaccine).
- `unknown`: not stated and not known for the product.

A group whose arms differ in adjuvant (some doses with, some without) takes the adjuvant most of its participants got if the record shows it, else `unknown`.

## What to record for each trial

- **`targets`**: the diseases the vaccine is meant to prevent, a list from: `covid19`, `influenza`, `rsv`, `zoster`, `varicella`, `pneumococcal`, `meningococcal`, `hpv`, `hepatitis_b`, `hepatitis_a`, `hepatitis_e`, `rotavirus`, `norovirus`, `dengue`, `chikungunya`, `zika`, `ebola`, `malaria`, `tuberculosis`, `hiv`, `cmv`, `herpes_simplex`, `c_difficile`, `s_aureus`, `e_coli`, `group_b_strep`, `tetanus_diphtheria_pertussis`, `polio`, `measles_mumps_rubella`, `japanese_encephalitis`, `tick_borne_encephalitis`, `lyme`, `rabies`, `typhoid`, `cholera`, `shigella_etec`, `anthrax`, `smallpox_mpox`, `plague`, `yellow_fever`, `other` (name it in `note`).
- **`off_target`**: true when the vaccine was given to prevent a disease other than its own (BCG against COVID-19).
- **`pregnant`**: true when the people vaccinated were pregnant women.
- **`patients`**: true when the trial enrolled people because of a medical condition or procedure (HIV infection, transplant, cancer, dialysis, planned surgery, a hospital stay), false for healthy people and the general population at any age.

## Output

One JSON object per trial, in the order given, as a JSON array written to the file named in your task:

```json
{
  "nct": "NCT01234567",
  "eligible": true,
  "exclusion_reason": null,
  "comparisons": [
    {
      "vaccine_groups": [
        {"id": "EG000", "platforms": ["protein"], "adjuvant": "aluminium"},
        {"id": "EG001", "platforms": ["protein"], "adjuvant": "aluminium"}
      ],
      "placebo_groups": ["EG002"],
      "placebo_content": "inert",
      "offspring": false,
      "background": "none"
    }
  ],
  "targets": ["hpv"],
  "off_target": false,
  "pregnant": false,
  "patients": false,
  "note": ""
}
```

- `exclusion_reason` (when `eligible` is false; `comparisons` is then an empty list), one of: `not_a_preventive_vaccine` (the product tested is not a preventive vaccine against an infectious disease), `no_placebo_group` (controls are active vaccines, no treatment, or absent), `vaccine_in_every_arm` (every arm gets the vaccine; the trial compares vaccines, doses or schedules, or tests something else), `groups_not_randomized_arms` (the tables don't report the randomized placebo-controlled arms as separate groups), `vaccine_with_other_active` (the vaccine arm also got an active product its placebo arm didn't), `other` (explain in `note`).
- `background`: a few words on what both sides also received (`"routine infant vaccines"`, `"seasonal influenza vaccine"`), or `"none"`.
- `note`: anything a second reader would need to follow your decision, kept short: which period's groups you took, why a group was left out, what an `other` is.
- For an ineligible trial, still give `targets`, `off_target`, `pregnant` and `patients` if the record shows them, else `[]` and false.
