Skip to main content

Table 1 Variables, measures, and methods of analysis

From: Short-course antimicrobial therapy for paediatric respiratory infections (SAFER): study protocol for a randomized controlled trial

Variable/outcome

Hypothesis

Outcome measure

Methods of analysis

Primary

 Clinical cure (per-protocol)

Short-course (5-day) treatment with high-dose amoxicillin is non-inferior to long-course (10-day) treatment

At visit (day 14–21)

composite of:

1. Defervesced on or before day 4

2. No more than 1 further fever spike until visit

3. No tachypnoea and decreased work of breathing

4. No additional antibacterials given

Dichotomous

Logistic regression

Secondary, between-group comparisons

 Absenteeism (caregiver, from work)

Short-course treatment is non-inferior to long-course treatment

Daily symptom diaries:

self-reported number of caregiver-days absent from work (count)

Poisson regression

 Absenteeism (child, from daycare/school)

Short-course treatment is non-inferior to long-course treatment

Daily symptom diaries:

reported number of child-days absent from daycare/school (count)

Poisson regression

 Mild drug adverse reactions

Fewer in short-course arm than in long-course arm

Daily symptom diaries:

Reported number of child-days with diarrhoea, rash, abdominal pain, yeast infection (count)

Poisson regression

 Anaphylaxis and other severe drug adverse reactions

Fewer in short-course arm than in long-course arma

Daily symptom diaries, SAE reports. Dichotomous

Descriptive statistics

 Adherence to study medications

Better in short-course arm than in long-course arm

Daily symptom diaries, RA interview:

“adherence” defined as > 80% amoxicillin doses given (<3 doses of initial 15 doses in short-course arm, < 6 doses in reference arm) Dichotomous

Logistic regression

 Recurrence of respiratory illness after primary outcome visit but before 30-day follow up

Short-course treatment is non-inferior to long-course treatment

RA interview. Dichotomous

Logistic regression

 Development of antibiotic-resistant organism (ARO) colonization

Less frequent in short-course arm than in long-course arm

Enteric swab testing:

“new colonization”’ defined as 3–6 month swab ARO positivity in context of baseline swab ARO negativity. Dichotomous

Logistic regression

 Disruption of gut microbiome

Less marked in short-course arm than in long-course arm

Enteric swab testing:

comparison of gut microbiome at 3–6 months to baseline gut microbiome. Continuous

Linear regression

Tertiary, entire-cohort

 Distribution of salivary C-reactive protein in cohort

Mean will be greater than that observed in children with bronchiolitis but less than that observed in children with empyema

Salivary swab testing:

continuous

Descriptive statistics expressed as mean (95% CI)

 Prevalence of high-level S. pneumoniae nasopharyngeal colonization

Majority of cohort will be positive

Nasopharnygeal swab (NPS) testing:

Dichotomous

Percentage, with 95% CI

 Prevalence of Mycoplasma detection

Minority of cohort will be positive

NPS testing:

Dichotomous

Percentage, with 95% CI

Subgroup analyses

 Older (age 5–10 years) vs. younger (age <5 years)

Older age group will have higher salivary CRP values, lower rates of S. pneumoniae high-level colonization, more Mycoplasma positivity, and decreased rates of clinical cure

Clinical cure (as defined above)

Logistic regression with an interaction term between subgroup and treatment variables

 Higher vs. lower salivary CRP

Higher salivary CRP will be associated with decreased rates of clinical cure

Clinical cure

Logistic regression with an interaction term between subgroup and treatment variables

 Virus/Mycoplasma detected in baseline NPS vs. no virus detected (for primary outcome)

Detection of a virus or Mycoplasma will not affect observed rates of clinical cure

Clinical cure

Logistic regression with an interaction term between subgroup and treatment variables

Sensitivity analyses – for primary outcome only

 Intention-to-treat

Results will remain robust

Clinical cure

Logistic regression

 Strict per-protocol (those with adherence >80% and radiologist-verified pneumonia)

Results will remain robust

Clinical cure

Logistic regression

  1. SAE serious adverse event, RA Research Assistant, ARO antibiotic-resistant organism, CRP C-reactive protein
  2. aToo few events expected to be able to observe differences in treatment arms