In total, 722,372 mother–child dyads were identified. Most mothers were 30–34 years old (31.7%), normal weight (49.0%) and nulliparous (66.9%) (Table 1).

Table 1 Descriptive statistics for the cohort, stratified by prenatal exposure to proton pump inhibitors (PPIs) or antibiotics

Altogether, 23,910 (3.3%) women received a PPI prescription at any time during pregnancy, with 11,732 (1.6%) during the first, 10,251 (1.4%) the second and 9117 (1.3%) during the third trimester. For PPI-users, the mean DDD was 86 (interquartile range [IQR] 28–100) and the average number of prescriptions per individual was 1.6 (IQR 1–2).

Overall, 160,703 mothers (22.2%) were prescribed antibiotics at any time during pregnancy, with 63,374 (8.8%) during the first, 69,125 (9.6%) the second and 60,597 (8.4%) in the third trimester. For users, the mean DDD was 10.0 (IQR 6.3–16.0), and the average number of prescriptions per individual was 1.5 (IQR 1–2).

The number of H2RA prescriptions during pregnancy (n = 1080, 0.1%) did not allow for any statistical analyses, and therefore, only PPIs and antibiotics were investigated.

A total of 636,224 children reached at least age 2 years in the cohort and were used in early life exposure analysis. In total, when only considering early life exposure, 3903 (0.6%) got PPIs, 218,998 (34.4%) got antibiotics and 2193 (0.3%) got H2RAs before the age of 2 years (Table 2).

Table 2 Descriptive statistics for the cohort, stratified by early life (before age 2 years) exposure to proton pump inhibitors (PPIs) or antibiotics

More children of nulliparous women got early life PPIs or antibiotics compared with those that did not (70.9% versus 64.4%), while children of multiparous women had higher prevalence of antibiotics exposure (43.1% versus 31.7%) (Table 2).

Children exposed to PPIs in early life (before second birthday) had higher prevalence of being delivered via acute c-section (19.9% versus 10.4% of unexposed) and having a reflux diagnosis before the age of two (35.8% versus 0.2%) (Table 2).

Follow-up time accumulated to 4,145,681 person-years, with mean follow-up of 6 years (IQR 3–8). Follow-up for children exposed to PPIs in early life was IQR 2–7, compared with IQR 3–8 of unexposed children. Children exposed to antibiotics in early life had follow-up with IQR 4–9, compared with IQR 3–8 of unexposed children. No subgroup analyses were performed regarding early life exposure to PPIs due to insufficient power.

3.1 Cancer Diagnoses

In total, 1091 (0.15%) children were diagnosed with malignant cancer during the follow-up (Table 1), with an incidence rate of 151 cases per 100,000 individuals.

Out of those exposed to PPIs prenatally, 0.1% got diagnosed with cancer and 0.2% of those exposed prenatally to antibiotics (Table 1). Regarding early life exposure, 0.1% of those with antibiotics prescription got cancer after the age of 2 years and 0.4% of those with PPI prescription (Table 2).

Considering cancer subtype, ‘eye, brain and other parts of central nervous system (CNS)’ were the most common (n = 271, 24.8%), followed by lymphoid leukaemia (n = 235, 21.5%). The age distribution for all subtypes of cancers was similar (Fig. 1).

Fig. 1

Distribution of age at first diagnosis by cancer type

3.2 Prenatal Exposure and Overall Risk of Cancer

Parity (HR 0.88, 95% CI 0.78–1.00), delivery mode (elective c-section HR 1.35, 95% CI 1.09–1.67), being born preterm (HR 1.36, 95% CI 1.08–1.72) and being large for gestational age (HR 1.48, 95% CI 1.11–1.99) were associated with a higher risk of cancer, (p < 0.10 in a univariable model, Supplementary Table S2) and were, therefore, included in the multivariable model.

In the Cox proportional hazard model (adjusted for preselected variables: maternal age, maternal BMI, sex of child and the variables mentioned above) prenatal PPIs (adjusted hazard ratio [aHR] 1.11, 95% CI 0.79–1.56) and antibiotics (aHR 0.95, 95% CI 0.83–1.11) were not associated with the risk of cancer (Table 3). A sensitivity analysis showed that the missing BMI values did not influence the analysis (results not shown).

Table 3 Prenatal exposure to proton pump inhibitors (PPIs) and antibiotics and the risk of overall cancer, and risk of cancer before and after the age of two

Likelihood ratio tests showed that prenatal exposure to PPIs and antibiotics showed no interaction (p = 0.94).

Furthermore, there was no interaction between prenatal PPIs and delivery mode (p = 0.69).

Cumulative incidence curve can be seen in Supplementary Material (Supplementary Fig. 2).

Estimation of the proportional hazard assumption for the adjusted model showed that all predictors met that assumption (p = 0.57).

3.2.1 Subgroup Analysis

Subgroup analyses on antibiotic classes and cancer locations (insufficient power for PPIs) did not show any statistically significant associations (Supplementary Tables S3 and S4).

No association was observed for PPIs and antibiotics for any trimester (Table 4). Furthermore, no association was seen when the cohort was divided by delivery mode or preterm status (Table 4).

Table 4 Prenatal exposure of proton pump inhibitors (PPIs) and antibiotics and the risk of overall cancer, stratified by delivery mode, trimester of exposure and preterm status3.2.2 Dose–Response

No association was observed for prenatal PPI exposure as DDD or number of prescriptions with increased risk of childhood cancer. For prenatal antibiotics, the association with an increased risk of childhood cancer increased with the number of prescriptions, being the highest for > 3 prescriptions (aHR 1.51, 95% CI 1.02–2.25) (Fig. 2).

Fig. 2

Dose–response analysis using multivariable Cox hazard models (aHR, 95%CI) adjusted for maternal age, BMI, sex of child, parity, delivery mode, preterm status and being large for gestational age, with no prescriptions as reference for all analyses. PPIs (top row) and antibiotics (bottom row) regarding DDD as quartiles (left column) and number of prescriptions (right columns) and the overall risk of childhood cancer

3.3 The Risk of Cancer Before the Age of 2 Years

After univariable analysis, the multivariable model was adjusted for maternal age, BMI, parity, sex of child, maternal Nordic birth country, tobacco use, delivery mode and preterm status (Supplementary Table S2).

Neither prenatal PPIs (aHR 0.56, 95% CI 0.28–1.09) nor prenatal antibiotics (aHR 1.28, 95% CI 0.98–1.67) were associated with an increased risk of cancer before the age of 2 years (Table 3).

No interaction was found for prenatal exposure of both PPIs and antibiotics (p = 0.87).

In subgroup analysis, those exposed to quinolone antibacterials (J01M) prenatally (n = 4 cancer cases) were associated with increased risk of cancer before the age of 2 years (aHR 4.23, 95% CI 1.36–13.21) (Supplementary Table S3). No association was found for other antibiotics or cancer locations (Supplementary Table 4).

3.4 The Risk of Cancer After the Age of 2 Years

After univariable analysis, the multivariable model was adjusted for maternal age, BMI, sex of child, parity, tobacco use, delivery mode, preterm status and being large for gestational age (Supplementary Table S2).

Prenatal exposure to PPIs (aHR 1.26, 95% CI 0.86–1.85) and antibiotics (aHR 0.91, 95% CI 0.77–1.08) did not show an association with increased risk of cancer after the age of 2 years (Table 3).

For early life exposure of PPIs, there was an association with an increased risk of cancer after the age of two (aHR 3.68, 95% CI 2.24–6.06). For early life antibiotics exposure, the association was not significant (aHR 1.04, 95% CI 0.90–1.21) (Table 3).

Neither prenatal exposure to PPIs and antibiotics (p = 0.92) nor early life exposure to both drugs (p = 0.15) showed an interaction.

When looking at interaction between prenatal and early life exposure to PPIs (p = 0.89) and prenatal and early life exposure to antibiotics (p = 0.25) showed no interaction.

No association was found with a specific cancer location (Supplementary Table S4).

Cumulative incidence curve can be seen in Supplementary Material (Supplementary Fig. 3).

Estimation of the proportional hazard assumption showed that all predictors met that assumption (p = 0.65).