The relation between preterm birth and self-reported spinal pain in pre-adolescence—a study of 47,063 subjects from the Danish National Birth Cohort

DNBC is a cohort consisting of approximately 100,000 children born from 1996 to 2003 and their mothers, followed by several data collections from early pregnancy through the children’s life [17]. When the children turned 11 years old, they got invited to participate in the 11-year follow-up (DNBC-11), which included questions about spinal pain [17, 18]. The source population consisted of all pre-adolescents included in DNBC. In DNBC-11, a total of 47,063 pre-adolescents contributed with full information on all explanatory measures and spinal pain (Fig. 1); thus, all missing data were excluded. Data from DNBC were linked with nationwide register data on Statistics Denmark through the unique personal identification number assigned to all persons with a permanent residency in Denmark. All data were stored and processed in the secure remote server environment at Statistics Denmark.

Gestational age at birth was obtained from the Medical Birth Register containing information on all births in Denmark [19]. Gestational age was operationalized into full weeks and further into the following three groups: term (≥ 37 full weeks of gestation), moderate preterm (34–36 full weeks of gestation), and very preterm (< 34 full weeks of gestation).

Spinal pain was assessed in DNBC-11 in which a subdivision of the Young Spine Questionnaire (YSQ) was included. YSQ is a validated questionnaire used to measure the intensity and prevalence of pain in the neck, mid back, and lower back [20]. The original questionnaire consists of 19 questions, of which nine were included in DNBC-11. Frequency was measured by the following options: (1) often, (2) once in a while, (3) once or twice, and (4) never. Intensity was measured by the Revised Faces Pain Scale and ranged from 1 = no pain to 6 = very much pain [20]. There is no verified classification of spinal pain among children but in this study, we used an overall combined measure, which was divided into no pain, moderate pain, or severe pain in the low back, mid back, or neck. This measure was directly adopted from previous work based on the same population [7, 18, 21]. For further detailed information regarding the operationalization of the variable, see the previous study by Joergensen et al. [18]. Thus, severe pain was defined as having pain often or once in a while with an intensity of four or more on the Revised Faces Pain Scale [18, 22]. If the child reported pain in more than one region, the region in which the child reported the most severe pain was included in the outcome. In the main analysis, the outcome was divided into two groups: no pain (including moderate pain) versus severe spinal pain. Secondly, we operationalized a variable in which having experienced spinal pain covered both moderate and severe spinal pain. This variable included having experienced pain once or twice or once in a while or often with an intensity of 3 or more or having pain often with an intensity of 1 or more [18]. The three spinal regions were also investigated separately.

With the use of a causal diagram, potential confounders were selected a priori [23] (Supplementary File 1). The selected confounders were parity, maternal age at childbirth, major congenital anomalies, and parental educational level. Information on sex was obtained from DNBC. Age corresponded to the child’s age when responding to DNBC-11. Parity and maternal age at birth were obtained from the Danish Medical Birth Register. Parity was divided into nulliparous or parous. Maternal age at birth was divided into the following groups: < 25, 25–29, 30–34, and > 34 years old. Major congenital anomalies originated from the Medical Birth Register and were defined according to the EuroCAT guidelines [24]. Parental educational level was obtained from The Danish Population’s Education Register [25] and described the highest ongoing or completed educational level of the parents. It was categorized into low (0–2 ISCED), medium (3–4 ISCED), and high (5–8 ISCED) educational level according to the International Standard Classification of Education (ISCED) 2011 [26].

All statistical analyses were made using STATA V.16.1. Initially, we made descriptive statistics between the exposure of interest and the selected covariables. Categorical variables were reported as proportions and tested for heterogeneity using a chi-squared test. Applying multiple logistic regression models, we examined crude and adjusted associations between preterm birth and severe spinal pain in pre-adolescents aged 11 to 14 years. We estimated odds ratios (OR) and an α-level of 0.05 in the statistical testing was used, corresponding to a 95% confidence interval (95% CI). In all analyses, no spinal pain was considered as the reference outcome. We tested the interaction between preterm birth and sex (p = 0.160) and based on this, we could not definitively reject a possible interaction with sex, and therefore we chose to divide the analyses by sex. The analyses further accounted for the dependency between siblings in the study sample including a robust standard error estimator [27]. Subsequently, we analyzed the association by using the secondary outcomes: “moderate to severe” as well as the spinal regions separately. We performed a loss-to-follow-up analysis to investigate potential selection bias in our analysis population. To this, we used chi-squared tests of heterogeneity to compare study participants with individuals lost to follow-up (Supplementary File 2). In a sensitivity analysis, we further accounted for sample selection (into the cohort and from attrition) by applying inverse probability weighting (IPW) using a reference population consisting of all children born in Denmark between 1996 and 2003 [28, 29]. These analyses were conducted in line with previous spinal pain studies performed in the same study population [7, 18, 30].

Approval of the study was obtained from the Danish Data Protection Agency through the joint notification of The Faculty of Health and Medical Sciences at The University of Copenhagen (SUND-2017–09) and the DNBC Steering Committee (2017–23). All data were stored and processed at Statistics Denmark and no personally identifiable data were accessible.

A strength of this study was the large study population, which provided a high degree of statistical power. Despite the large study population, only a small number of the children were born preterm which may not provide enough statistical power to assess the association between preterm birth and spinal pain. Additionally, it is worth noticing that the small numbers in some of the exposure categories and the many performed tests might have led to chance findings and that some of the statistically significant findings could be type 1 errors [47]. A further strength was the prospective study design ensuring temporality and further minimal risk of recall bias, since exposure data were based on register data and outcome data were collected as point prevalence in DNBC-11. Spinal pain was self-reported and pain experience is by default subjective, which may affect the accuracy of the variable [48]. Self-report has, however, been demonstrated as the best option to measure pain in children [49]. Additionally, data on spinal pain originated from YSQ, included in DNBC-11, which is validated to be used to measure the neck, mid back, and low back pain in 9–11-year-olds [20]. Finally, access to Statistic Denmark registers allowing the application of high-quality Danish population registers to obtain information on health and social conditions to adjust for potential confounders strengthened the validity of the study [19, 24, 25]. However, we can never be sure that we completely cover all variables that may confound the relationship.

As always there is selection into the cohort and from attrition, also demonstrated in the loss-to-follow-up analysis, which may have affected the study findings. However, the fact that children born preterm in this study figure in the same level as in the total Danish population is a strength [50]. In addition, a study investigating whether low participation in cohort studies is inducing bias suggested that non-participation did not affect the risk estimates in the DNBC cohort [51]. Finally, to account for potential selection bias, we performed IPW analyses, and as in previous studies in the same population, IPW did not make any essential changes to the estimates; thus, we believe that selection bias had no or a minimum influence on our results.

In contrast to our hypothesis, this study indicates that girls born very preterm were less likely to have severe spinal pain in pre-adolescence than girls born to term, whereas there is seemingly no association between gestational age at birth and spinal pain in boys aged 11–14 years.