Assessing the Impact of Smoke-free Legislation on Perinatal Health in the Netherlands

Primary research questions

1. Is the phased introduction of smoke-free legislation in The Netherlands associated with reductions in adverse perinatal outcomes (e.g. perinatal mortality, preterm birth, small for gestational age)?
2. How do these associations compare with those described for other European countries in comparable studies (i.e. Belgium (Cox 2013), England (Been et al under review), Scotland (Mackay 2012))?

Study design Retrospective cohort study (using prospective routinely collected health care data)

Study population All singleton births in the Netherlands between January 1st 2000 and December 31st 2011.

Intervention The intervention under study is the ban on smoking in workplaces, and in bars and restaurants implemented in the Netherlands on January 1st, 2004 and July 1st, 2008, respectively.

Inclusion and exclusion criteria We will include all registered singleton births in the Netherlands occurring between January 1st, 2000 and December 31st, 2011. This is the maximum time period surrounding the ban's introduction for which the required birth data are available through the data source. Multiple pregnancies, neonates with chromosomal anomalies, pregnancies with unknown gestational age, pregnancies that ended before 24 weeks and pregnancies resulting in the birth of a child weighing less than 500 grams will be excluded.

Outcome

The primary outcomes are:

• Perinatal mortality (stillbirth + early neonatal mortality, i.e. within the first 7 days of life)
• Preterm birth (live birth with gestational age \<37+0 weeks)
• Small for gestational age (SGA; live birth with birth weight below 10th centile; The Dutch PRN reference curves for birth weight by gestational age according to parity, sex and ethnic background will be used)

To assess whether smoke-free legislation had a selective impact on certain subgroups of outcomes we furthermore identified a number of secondary outcomes:

• Stillbirth (born dead from 24+0 weeks of gestation)
• Early neonatal mortality (live birth and death within first 7 days)
• Very preterm birth (live birth with gestational age \<32+0 weeks)
• Low birth weight (live birth with birth weight \<2500 grams)
• Very low birth weight (live birth with birth weight \<1500 grams)
• Very small for gestational age (live birth with birth weight below percentile 2.3rd centile)
• Major congenital anomalies (based on reported associations with antenatal smoke exposure (Hackshaw 2011)).

Data sources Individual level health care data will be extracted from The Netherlands Perinatal Registry (PRN). Linked midwifery, obstetric, and neonatal data are available from 2000 to 2011 (including 2011).

Data extraction and handling All relevant variables regarding our outcomes, as well as relevant potential confounders will be extracted from the database.

Sample size Power calculation for interrupted time series modelling is complicated given the complexity of the analysis. Similar previous studies have demonstrated statistically significant and clinically relevant effects of smoke-free legislation on preterm birth (Cox 2013; Mackay 2012; Page 2012), low birth weight (Mackay 2012), SGA (Mackay 2012; Kabir 2013), and perinatal mortality (Mackay et al. and Been et al. both under review). Given the larger population size of the Netherlands as opposed to the regions in which these studies were carried out (except for Been et al. under review), we expect our study to have sufficient power to detect similar effect sizes, should these be present.

Statistical analysis Incidences for each outcome will be presented graphically for each time period to facilitate visualisation of temporal fluctuations and trends in changes of incidence levels. To facilitate timing of the events, date of delivery and expected term date are required for each pregnancy. Data will be presented as outlined in different Tables. Interrupted time series analyses with adjustment for potential confounders will be performed to assess the associations between implementation of smoke-free legislation and primary and secondary outcome measures. Individual-level analysis will be performed using logistic regression analysis. The models will account for the underlying temporal trend in incidence, and will allow for a sudden change in incidence ('step change') following the introduction of the smoking bans. We will test and adjust for any non-linearity in the underlying time trends, and seasonality will be accounted as appropriate.

Sensitivity analyses (primary outcomes only) In recent years, gestational age is usually estimated based on early ultrasonography findings, which is more reliable than estimation based on the last menstrual period. Although the method of ascertainment is not recorded in PRN, there is an item indicating 'certainty' of the gestational age, which is positive in about 93% of records. For the primary outcomes preterm birth and SGA we will perform a sensitivity analysis including only cases in whom gestational age estimation is considered reliable according to this item.

Recent perinatal management changes have been implemented in The Netherlands resulting in increased active management of babies born at the edge of viability (i.e. 23-24 weeks gestation) (NVOG 2010). This has resulted in increased survival at this gestational age as well as altered management of 25-26 week infants, which likely affects the number of babies born preterm (although this effect is expected to be small given the small percentage of all preterm babies being born at this stage), as well as mortality indicators. For the primary outcomes preterm birth and perinatal mortality we will therefore perform a sensitivity analysis excluding babies born before 26 completed weeks of gestation.

Smoking during pregnancy is known to be underreported in the PRN database. Definitions between different caregivers differ; (any) smoking and heavy smoking (>20 cigarettes daily). We will consider performing subgroup analyses of the impact of smoke-free legislation on the primary outcomes according to maternal smoking status during pregnancy.

In a sensitivity analysis we will investigate whether smoke-free legislation has any differential impact on spontaneous preterm birth versus medically indicated preterm birth.

All analyses will be performed using Stata 13.0.