Study population and data sources

We used data from the Trøndelag Health Study (HUNT), the Medical Birth Registry of Norway (MBRN), and hospital records in the catchment area to identify fractures. Data were linked via their unique national 11-digit personal identification number.

HUNT

Data were collected from the third survey of the Trøndelag Health study, HUNT 3 (2006–2008), a large longitudinal population-based health study in central Norway. The geographic, demographic, and occupational structure of this region are considered representative of the country as a whole [19, 20]. All individuals aged 20 years or older in that year were invited to participate, and 50,821 (54.1%) responded. The participants completed comprehensive questionnaires and underwent a short clinical examination at the screening station. We included 11,099 participants (64,314 women and 4665 men) born 1967–1988 with available information on their own birth in the MBRN, see flowchart (Fig. 1).

Fig. 1

Flowchart of the included participants in the present study. *MBRN, Medical Birth Registry of Norway

Medical Birth Registry of Norway

Information on birth characteristics was collected from the Medical Birth Registry of Norway (MBRN). The MBRN is a national health register established in 1967 that collects information on all births reported by Norwegian maternity units as well as home births and births during transportation [21]. In addition to the name and personal identification number of both the child and the parents, the register contains information about the mother’s health before and during pregnancy and any complications in connection with the birth. Furthermore, it documents the newborn’s health information, such as birth length, birth weight, head circumference, and Apgar (Activity, Pulse, Grimace, Appearance, Respiration) score, among others.

Exposures: birth characteristics

Birth weight was categorized as follows:

 < 2.5 kg (low birth weight, LBW) [22]; 2.5–2.9 kg; 3.0–3.4 kg; 3.5–3.9 kg (reference group, representing the average weight of both boys and girls in Norway) [23]; 4.0–4.4 kg; ≥ 4.5 kg (high birthweight, HBW).

Additionally, birth weight was included as a continuous variable.

Gestational age at delivery was defined as the duration of a pregnancy, measured from the first day of the woman’s last menstrual period:

Preterm (< 37 weeks) [24]

Term (37–41 weeks, reference group) and

Post-term birth (≥ 42 weeks)

Birth weight for gestational age was defined as the infant’s weight relative to their gestational age, categorized by standardized birth weight (z-score) [25] into the following groups:

Small for gestational age (SGA): Birth weight below the 10th percentile (z-score < − 1.28).

Appropriate for gestational age (AGA): Birth weight between the 10th and 90th percentile (z-score between − 1.28 and 1.28 SD) (reference group).

Large for gestational age (LGA): Birth weight above the 90th percentile (z-score > 1.28 SD).

Outcome: fractures

Fractures were identified from the only two hospitals in the catchment area by using the International Classification of Disease (ICD) codes: version 9 (for fractures obtained before 2000 in the secondary analysis) or version 10. None of these fractures is treated in primary care or private hospitals. The fracture types we aimed to study, and those we had access to, included proximal humerus (812.0–812.3; S42.2–S42.31); distal forearm (813.4, 813.5; S52.5–S52.61); hip (820.0–820.3; S72.0–S72.21); and spine (805.2–805.5; 806.0–806.5; S12.0–S12.21; S22.0– S22.1; S32.0– S32.01; T08, T08.90). Where available, we also used the NOMESCO Classification of Surgical Procedures codes (NCPS) for surgery, cast, or splint.

To ensure accuracy and avoid misclassification of suspected fractures, a non-vertebral fracture was defined as one of the following: (1) two identical ICD codes within 3 months; (2) one ICD code and one relevant NCSP code registered within 2 months before and 3 months after the ICD code. For fractures of the spine, only one ICD diagnosis code was required. Fracture diagnosis codes that occurred less than a year after a defined fracture were considered to represent new registrations of the same fracture [26]. The fractures were recorded from 1988 (the beginning of electronic recording at regional hospitals) until October 21, 2021.

Covariates

Potential confounders were selected based on a Directed Acyclic Graph (Fig. 2) and included sex, birth year, maternal age, and maternal morbidity. Maternal morbidity comprised conditions recorded before or during pregnancy that may potentially affect the offspring [27], including chronic inflammatory joint disease, diabetes, and preeclampsia/eclampsia.

Fig. 2

Directed Acyclic Graph with confounders and mediators

Statistical analysis

Descriptive data are presented as means with standard deviations (SD) for continuous data and numbers and percentages for categorical data. All birth characteristics were analyzed as categorical variables, and birth weight was additionally analyzed as a continuous variable using either standard deviation scores or per 100 g increase in birth weight. Crude and adjusted hazard ratios (HRs) of fracture associated with the various birth characteristics were estimated using Cox regression. The precision of all estimated associations is given by a 95% confidence interval (CI). Sex and maternal morbidity were included as categorical variables, while maternal age and birth year were continuous. In analyses of birth weight, we also adjusted for gestational age, to isolate the specific impact of birth weight on the outcome variable. While we focus on young adults, the time at risk in the main analysis started on the date of participation in HUNT 3 when the participants were 19–41 years old, and ended on the date of first fracture, emigration, death, or end of follow-up (October 21, 2021). This approach helps address potential selection bias that could arise if we started from birth. We also present a secondary analysis, where the time at risk started on January 1, 1988, and ended on the same events as in the main analysis.

We also conducted a test for linear trend across categories of birth characteristics adjusting for relevant confounders, treating the categories as an ordinal variable in the regression model.

In the main analysis, several sensitivity analyses were conducted: excluding participants with birth weights < 2.5 kg to address potential non-linear relationships, assessing the impact of cesarean section on birth weight categories and gestational groups in relation to fracture risk, and analyzing outcomes for distal forearm fractures and excluding spine fractures to evaluate their impact on results.

The proportional hazards assumption was evaluated using Schoenfeld residuals and graphical methods, including visual inspection of log-minus-log plots. All statistical analyses were performed using Stata 18.0 (StataCorp LCC, College Station, TX, USA).

Ethics

Participants in HUNT 3 gave written, informed consent for the data to be used for research, including a linkage with other registers. The study was approved by the Regional Committee for Medical Research Ethics, Central Norway (application number 246732) [21].