Caffeine to improve neurodevelopmental outcomes in infants born late preterm (The Latte Trial): study protocol for a randomised controlled trial.

The trial plans to test **caffeine citrate** (a loading dose of 40 mg per kg of body weight, then 20 mg per kg daily) versus an **identical placebo** (same volume, same appearance). The intervention is given **enterally** (by mouth, via a feeding tube or bottle) starting within 72 hours of birth and continuing daily until the infant reaches **40+0 weeks postmenstrual age** (roughly their original due date).

**Primary outcome:** Cognitive score on the Bayley Scales of Infant Development, 4th Edition (Bayley-4) at **2.5 years corrected age** (adjusted for prematurity).

**Secondary outcomes** include: motor and language scores on the Bayley-4, rates of cerebral palsy, visual and hearing impairment, growth parameters, sleep quality, feeding tolerance, heart rate abnormalities, and safety outcomes like hospital readmission.

The trial aims to recruit **478 late preterm infants** (or twin pairs) born at **34+0 to 36+6 weeks' gestation** (i.e., 34 to just under 37 weeks). Infants are recruited from **13 hospitals across Aotearoa New Zealand**, including both secondary and tertiary maternity centres.

**Key inclusion criteria:**

Infants born at 34+0 to 36+6 weeks

No contraindication to caffeine

Parent/caregiver able and willing to provide informed consent

Recruited within 72 hours of birth

**Key exclusion criteria:**

Major congenital abnormalities

Previous caffeine treatment

Renal or hepatic impairment

Tachyarrhythmia (abnormally fast heart rate)

Seizures

Hypoxic ischaemic encephalopathy (brain injury from oxygen deprivation)

Maternal age under 16 years

Triplets or higher-order multiples

**Notable design feature:** The trial intentionally aims to recruit **equal numbers of Māori and non-Māori infants** (approximately 239 each) to achieve "Mana Whakamārama" — equal explanatory power for the Indigenous Māori population, who make up 20% of late preterm births in New Zealand.

**Primary outcome measure:** Bayley Scales of Infant Development, 4th Edition (Bayley-4) **Cognitive Score**. This is a standardised, validated developmental assessment administered by trained assessors who are blinded to treatment allocation. Scores are normed to a mean of 100 with a standard deviation of 15. Higher scores = better cognitive function.

**Secondary outcome measures include:**

Bayley-4 **Language** and **Motor** scores

Diagnosis of **cerebral palsy** (standardised neurological exam)

**Visual impairment** (defined as visual acuity worse than 6/60 in the better eye)

**Hearing impairment** (requiring amplification or cochlear implant)

**Growth** (weight, length, head circumference at term and 2.5 years)

**Intermittent hypoxaemia** frequency (measured by overnight pulse oximetry — the number of desaturation events per hour where oxygen saturation drops ≥10% below baseline)

**Heart rate** (mean and incidence of tachycardia, measured by oximetry)

**Feeding tolerance** (time to full enteral feeds, vomiting episodes)

**Sleep quality** (parent-reported questionnaire)

**Hospital readmission** rates

**Mortality**

**Study design:** Phase III, multi-centre, parallel two-arm, double-blind, placebo-controlled randomised superiority trial.

**Randomisation:** Infants are randomised **1:1** to caffeine or placebo using an online randomisation system. Randomisation is **stratified** by three factors:

1. **Study site** (grouped by hospital)

2. **Gestational age** at birth (34, 35, or 36 completed weeks)

3. **Ethnicity** (Māori or non-Māori)

**Blinding:** Double-blind — neither the parents/caregivers, the clinical staff, nor the outcome assessors know which treatment the infant receives. The placebo is identical in volume, appearance, and taste. Blinding is maintained by an independent pharmacy that prepares and distributes the study drug.

**Duration of intervention:** From enrolment (within 72 hours of birth) until **40+0 weeks postmenstrual age** (roughly the original due date). This means treatment duration varies depending on how early the infant was born — approximately 3–6 weeks.

**Duration of follow-up:** Primary outcome assessed at **2.5 years corrected age** (adjusted for prematurity). This means the trial will take approximately 3–4 years to complete from first enrolment to final outcome assessment.

**Statistical analysis:** Primary analysis will be on a **modified intention-to-treat** basis (all randomised infants who receive at least one dose of study drug and have outcome data). The analysis uses **generalised mixed models** with adjustment for stratification factors, socioeconomic status, and sex, and includes a random effect for non-independence of multiples (twins). A sample size of 478 infants (239 per group) provides **80% power** to detect a difference of **0.3 standard deviations** (approximately 4.5 points on the Bayley-4 cognitive scale) at a two-sided alpha of 0.05.

**What this design can prove:**

**Causality:** Randomisation and blinding mean that any difference in outcomes between groups can be attributed to caffeine (not to confounding factors like parental education, socioeconomic status, or hospital care quality).

**Generalisability:** Multi-centre design across 13 hospitals improves external validity.

**Safety:** Double-blind placebo control allows detection of both benefits and harms.

**What this design cannot prove:**

**Mechanism:** The trial tests whether caffeine works, but not exactly *how* (e.g., via reducing intermittent hypoxaemia, direct neuroprotection, or other pathways).

**Long-term outcomes beyond 2.5 years:** Cognitive scores at 2.5 years are moderately predictive of later IQ, but not perfectly so. Longer follow-up would be needed to confirm effects persist into school age.

**Dose-response:** Only one dose regimen is tested (40/20 mg/kg). The trial cannot determine whether a lower or higher dose would be more effective or safer.

**Subgroup effects:** While stratified by ethnicity and gestational age, the trial may be underpowered for formal subgroup analyses.

**Methodological strengths:**

Double-blind, placebo-controlled design minimises bias

Stratified randomisation ensures balance on key prognostic factors

Modified intention-to-treat analysis preserves randomisation benefits

Active recruitment of Indigenous population addresses health equity

Pre-registered protocol and statistical analysis plan

**Methodological weaknesses (as a protocol):**

No results yet — this is a plan, not findings

Long follow-up (2.5 years) means attrition is a risk

Bayley-4 is a clinical assessment requiring trained assessors; inter-rater reliability must be maintained across sites

The 72-hour enrolment window may miss some eligible infants

Exclusion of triplets and higher-order multiples limits generalisability to these high-risk pregnancies

**This is a study protocol — no results are reported.** The paper describes the planned methodology, not outcomes. The trial is currently recruiting or yet to complete follow-up.

However, the protocol references the earlier **Latte Dosage Trial** (Phase II), which did report findings:

**132 late preterm infants** randomised to caffeine citrate (at doses of 10, 20, 30, or 40 mg/kg loading, with corresponding maintenance doses) or placebo

**Primary outcome:** Frequency of intermittent hypoxaemia (oxygen desaturation events ≥10% below baseline per hour) measured by overnight oximetry at 2 weeks post-randomisation

**Result:** Caffeine citrate **reduced intermittent hypoxaemia frequency**, with **20 mg/kg being the most effective dose** — nearly a **70% reduction** in the geometric mean hourly event rate compared to placebo

**Secondary outcomes:** Caffeine increased mean oxygen saturation (SpO₂) and reduced time spent with SpO₂ < 90%

**Safety:** No adverse effects on growth, feeding, or sleep. Caffeine increased mean heart rate, with a **significantly higher incidence of tachycardia** in all caffeine groups versus placebo

Since the main Latte Trial has no results, the effect magnitude is unknown. Based on the Phase II Latte Dosage Trial:

**Intermittent hypoxaemia:** A ~70% reduction in desaturation events per hour is a large effect — equivalent to going from, say, 10 events per hour down to 3 events per hour

**Oxygen saturation:** Mean SpO₂ increased by approximately 1–2 percentage points, and time below 90% saturation was reduced

**Heart rate:** Tachycardia incidence was higher in caffeine groups, but the absolute rates and clinical significance are not fully detailed in the protocol

For the main trial's primary outcome (Bayley-4 cognitive score at 2.5 years), the trial is powered to detect a **0.3 standard deviation difference** (approximately **4.5 points**). To put this in context:

A 4.5-point difference on a scale with mean 100, SD 15 is considered a **small-to-moderate effect**

This is roughly equivalent to the difference between an average child and a child at the 40th percentile

For population-level outcomes, even small shifts in mean cognitive scores can have large public health implications

**As a protocol, the main limitation is that no results exist yet.** However, the planned trial has several inherent limitations:

**Acknowledged by authors:**

Long follow-up period (2.5 years) risks attrition — families may move, lose interest, or become difficult to contact

Bayley-4 is a clinical assessment requiring trained, blinded assessors; maintaining blinding and standardisation across 13 sites is challenging

The trial is powered for the primary outcome but may be underpowered for secondary outcomes and subgroup analyses

Exclusion of triplets and higher-order multiples limits generalisability to these higher-risk infants

**Critical reader observations:**

**No results:** The paper describes what the trial *will* do, not what it *found*. It cannot inform clinical practice until completed

**Single dose regimen:** Only one caffeine dose is tested (40/20 mg/kg). The Phase II trial suggested 20 mg/kg was optimal for reducing hypoxaemia, but the Phase III trial uses 40/20 mg/kg — a higher loading dose. The rationale for this choice is not fully explained

**Surrogate endpoint concern:** The primary outcome (Bayley-4 at 2.5 years) is a well-validated measure, but it is still a proxy for long-term cognitive function. Effects may not persist into school age

**Generalisability:** Recruited from New Zealand hospitals; results may not apply to other populations with different genetic backgrounds, healthcare systems, or breastfeeding practices

**Industry funding:** Funded by the Health Research Council of New Zealand (public funding) — no industry involvement, which is a strength

**No interim analysis described:** The protocol does not mention planned interim analyses for futility or harm, which would be important for a trial with a long follow-up

**Multiple comparisons:** With many secondary outcomes, there is risk of false positive findings without correction

**Important caveat:** This is a study protocol for a trial in **infants**, not adults. The findings (when available) will not directly apply to self-experimentation in healthy adults. However, the principles of testing caffeine's effects on cognition and the methodology are instructive.

**For someone running their own n=1 experiment on caffeine and cognition:**

### What to test

**Intervention:** Caffeine citrate (or caffeine anhydrous) at a standardised dose. For adults, typical doses range from 100–400 mg. Based on the trial's dose (20 mg/kg maintenance in infants), a proportional adult dose would be ~1.5–3 mg/kg (105–210 mg for a 70 kg adult)

**Comparator:** Placebo (identical capsule or beverage without caffeine) — ideally double-blind (you and the person administering don't know which is which)

**Timing:** Daily dosing at the same time each day (e.g., morning) to control for circadian effects

### Minimum meaningful duration

**Acute effects:** 1–2 weeks per condition (caffeine vs. placebo) to assess immediate cognitive effects

**Chronic effects:** 4–6 weeks per condition to assess tolerance, withdrawal, and sustained cognitive benefits

**Washout period:** At least 1 week between conditions to eliminate carryover effects (caffeine has a half-life of 3–6 hours, but withdrawal symptoms can last 2–9 days)

### What to measure

**Cognitive performance:** Use validated computerised tests (e.g., Psychomotor Vigilance Task for reaction time, N-back task for working memory, Stroop test for executive function). Measure daily at the same time (e.g., 30–60 minutes after dosing)

**Subjective alertness:** Visual analogue scale (0–100) for alertness, fatigue, and mood, completed 3–4 times daily

**Sleep quality:** Actigraphy (if available) or sleep diary (bedtime, wake time, sleep latency, number of awakenings, total sleep time). Measure nightly

**Heart rate:** Resting heart rate measured at the same time each morning (before and after dosing) — caffeine increases heart rate

**Side effects:** Daily log of jitteriness, anxiety, gastrointestinal discomfort, palpitations

**Confounders:** Record caffeine from other sources (tea, coffee, chocolate, soft drinks), alcohol, exercise, stress, and menstrual phase (for women)

### Key confounds to control for

**Caffeine tolerance:** If you're a regular caffeine user, you'll need a longer washout (2–4 weeks) to establish a true baseline

**Circadian effects:** Test at the same time each day; avoid testing late in the day when caffeine may interfere with sleep

**Expectation effects:** Use blinding (e.g., have someone else prepare identical capsules) and randomise order of conditions

**Dietary confounds:** Avoid other stimulants (nicotine, energy drinks) and depressants (alcohol) during the experiment

**Sleep debt:** Even small sleep deficits (1–2 hours) can confound cognitive results. Track sleep quality nightly

**Genetics:** CYP1A2 genotype affects caffeine metabolism. Slow metabolisers may experience stronger effects and more side effects. Consider genetic testing (e.g., 23andMe) if available

**Menstrual cycle:** For women, cycle phase affects caffeine metabolism. Test across at least one full cycle per condition

### What a positive result would look like

**Reaction time:** 10–20% improvement on Psychomotor Vigilance Task (e.g., mean reaction time decreases from 280 ms to 250 ms)

**Working memory:** 5–10% improvement in accuracy on N-back task (e.g., 80% to 85% correct)

**Subjective alertness:** 15–20 point increase on 0–100 visual analogue scale

**Consistency:** Effects should be reproducible across multiple days within the same condition, not just a single "good day"

**Dose-response:** If you test multiple doses (e.g., 100 mg vs. 200 mg), higher doses should produce larger effects (up to a plateau, after which side effects may negate benefits)

**Effect size:** A Cohen's d of 0.3–0.5 (small-to-moderate effect) is realistic based on the literature. This means your performance on caffeine would be better than ~60–70% of your performance on placebo

**Bottom line for self-experiment