Data sources

This population-based cohort study was based on a linkage of the following four nationwide registers in Denmark:

The Danish Facial Cleft Register (DFCR) encompasses individuals born with OC in Denmark in the period 1936 to 2009. Registration and treatment of individuals with OC has been centralized in Denmark since the 1930s, which ensures high ascertainment for the complete cohort, and for CL(P) the ascertainment in the period 1983 to 1987 has been found to be 99% [34]. OC discovered later in life are also included. The register includes information on cleft type and severity of the cleft. Cleft lip (CL) includes cleft lip only. Cleft lip and cleft palate (CLP) includes cleft of the lip and cleft of the palate. Cleft palate (CP) includes cleft of the palate only and can be submucous, in the soft palate only or both the soft and the hard palate. Microforms such as bifid uvula, defects in the orbicularis oris muscle etc. are not included in the DFCR. Information on any associated anomalies are also registered and classified into major or minor depending on whether the anomaly is likely to be part of a syndrome. Major anomalies include neural tube defects, severe mental retardation, syndromes (e.g. Van der Woude, Pierre Robin, 22q11 deletion syndrome), whereas polydactyly, hip dislocation, clubfoot, umbilical hernia etc. are considered minor associated anomalies. Non-syndromic OC is defined in DFCR as individuals with no major associated anomalies or individuals with 0–2 minor associated anomalies. For the birth cohorts 1936–1987 the number of syndromic OC is likely underestimated [34], but medical records have been reviewed for the later birth cohorts in order to obtain more complete information on associated anomalies. In DFCR, 11% of CL(P) and 32% of CP are registered as syndromic. The register is described in details elsewhere [1, 2].

The Danish Civil Registration System (CRS) was established in 1968, where all persons alive and living in Denmark were registered and assigned a unique personal identification number. Since then all persons with permanent residence in Denmark are registered. This unique personal identification number enables unambiguous linkage of all national registers in Denmark. The CRS includes, among other variables, information on the unique personal identification number, sex, date of birth, birth registration code, date of death, and continuously updated information on date and place of residence [35]. The CRS was established for administrative purposes, and the information is generally accepted to be highly accurate [36].

The Danish Psychiatric Central Research Register (PCRR) comprises information on every inpatient admission to Danish psychiatric hospitals since 1969 and on all outpatient clinical visits and emergency room visits to these hospitals since 1995 [37]. The register contains information on dates of admission and discharge, start and end dates of outpatient treatment, dates of emergency visits and all diagnoses. Until December 1993, the 8th revision of the International Classification of Diseases (ICD-8) was used, and in 1994 and onwards the ICD-10 Classification of Mental and Behavioral Disorders: Diagnostic Criteria for Research (ICD-10-DCR) has been used. The PCRR includes only treatment in psychiatric hospital settings. Contacts with general practitioners or psychiatrists outside hospital settings where less severe psychiatric disorders may be treated are not included. Hospitalization due to severe psychiatric disorder takes place in psychiatric hospitals, and all psychiatric hospitals in Denmark are public and provide care free of charge. Therefore, the PCRR provides a complete nationwide assessment of all mental disorders treated in psychiatric hospitals, and several of the key diagnoses have been validated with good results [38–42].

The Danish Register of Causes of Death contains computerized records on causes of death for all deaths among Danish residents dying in Denmark since 1970 [43]. The register contains the underlying cause of death and any contributory causes of death classified according to ICD-8 until December 1993 and according to ICD-10 in 1994 and onwards. Until 2007 coding was performed by specially trained coders under the supervision of the medical staff of the National Board of Health based on the medical information on the death certificates. Since 2007, the medical doctor, who has verified the death and issued the death certificate, also classifies these causes and the death certificates are submitted by an electronic form to the National Board of Health. The data quality relies on the notification done by the individual physicians and the coding in the National Board of Health. No central validation of the classification is done.

This study was approved by the Danish Data Protection Agency (Project No. 2016-41-4534) and no informed consent was required, because the study is based on secondary, de-identified data.

Study population

All individuals born January 1, 1936 to December 31, 2009 with non-syndromic OC were identified using DFCR. A 5% random sample of the entire population in Denmark in each year was extracted by Statistics Denmark. From this random sample we included only individuals without OC. In both groups we included only individuals with a Danish birth registration code and a Danish code of residence from birth, since the DFCR only includes individuals born in Denmark. Furthermore, in both groups we included only individuals who did not die or emigrate from Denmark in their first year of life. From the remaining 5% random sample, we randomly selected a comparison cohort of individuals without OC matched 10 to 1 to the group with OC by sex and birth year. Finally we restricted both groups to those alive and living in Denmark April 1, 1969 where registration of psychiatric admissions was initiated. The final study sample consisted of 8,568 individuals with OC and 85,653 in the comparison cohort. The ratio of number of individuals with OC to number of individuals in the comparison cohort is not precisely 1 to 10 in the final study sample, since individuals who died or emigrated prior to initiation of follow-up are excluded (Flow diagram is shown in Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Inclusion Criteria for Individuals with Non-Syndromic Oral Cleft and Matched Comparison Cohort.

https://doi.org/10.1371/journal.pone.0156261.g001

This study focuses on non-syndromic OC, but results for all individuals with OC and for individuals with syndromic OC can be seen in the supporting information (Flow diagram for all individuals with OC and for individuals with syndromic OC are shown in S1 and S2 Figs, respectively).

Assessment of psychiatric disorder

We included all inpatient admissions, outpatient contacts and visits to emergency care units at a Danish psychiatric hospital initiated between April 1, 1969 and December 31, 2012 based on the PCRR. For ICD8 data, we excluded observations with a modification code, i.e. “suspected but not confirmed" diagnose, and for ICD-10-DCR data, we excluded referral diagnoses and temporal diagnoses since these are not necessarily confirmed. Individuals were classified with a psychiatric disorder if they received care at a psychiatric hospital with a psychiatric diagnosis (ICD-8:290–315 and ICD-10-DCR: F00-F99). The specific mental disorders were categorized according to Table 1 following the study [44], and the date of diagnosis of a specific mental disorder was set to be the first day of hospital encounter (inpatient, outpatient, or emergency care) indicating that disorder. Individuals with more than one diagnosis were included in each specific disorder, i.e. it is possible for individuals to be in more than one diagnosis group but obviously in the overall analyses of any psychiatric diagnosis we only included the first psychiatric diagnosis.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Classification of Psychiatric Disorders According to ICD-10-DCR and equivalent ICD-8 Diagnoses.

https://doi.org/10.1371/journal.pone.0156261.t001

Assessment of suicide

It has previously been shown using the DFCR that individuals with OC with and without associated anomalies have increased risk of suicide [30] and therefore we used the Danish Register of Causes of Death to identify individuals having suicide as underlying or contributory cause of death (ICD-8: E950-E959 and ICD-10: X60-X84 and Y87.0) from January 1, 1970 to December 31, 2010, in order to address the risk of suicide in individuals with non-syndromic OC.

Study design and statistical analysis

Individuals were followed from April 1, 1969 or the earliest possible age at which the person could develop the specific mental disorder (see Table 1), whichever came later. Follow-up were terminated at date of first diagnosis of the specific disorder, death, first emigration from Denmark or December 31, 2012, whichever came first. Emigration from Denmark was defined as residence outside Denmark for more than 2 years.

Cox proportional hazard model with age as the underlying time scale was used to compute hazard ratios (HRs) of time to first diagnosis for individuals with OC compared with the matched comparison cohort for any psychiatric diagnosis and for each specific disorder. Additional models were stratified by cleft type and each cleft type was compared with their matched comparison cohort. Furthermore, models were estimated stratifying by birth cohort, sex and age. The proportional hazard assumption across age was assessed graphically.

Strengths and limitations

The major strengths of this study are the large sample size and long follow-up time which includes ages from 1 to 76 years. There is almost no selection bias as the project is register-based and information on OC and psychiatric diagnoses is collected independently. As previously mentioned, the ascertainment of individuals with OC in the population is virtually complete [34], and several of the psychiatric diagnoses are validated with good results [38–42].

Our results might be biased if individuals with OC are more likely to be referred to a psychiatric hospital, because of their greater use of healthcare services in general. This concern mainly applies to estimates for children with OC rather than adults since differences in hospitalizations and healthcare use compared to unaffected individuals are largest during childhood [9]. This would lead to overestimated relative risks for children. However, we find smaller risks for mental retardation and pervasive developmental disorders during childhood than in adolescence/adulthood, indicating that this is unlikely to be a major bias.

Another concern is differential loss to follow-up since individuals with non-syndromic OC have significantly more deaths and fewer emigrations during the follow-up period than individuals in the comparison cohort, which could potentially underestimate the association between OC and psychiatric diagnoses. In particular, the increased risk of suicide among individuals with OC would underestimate the association, but approximately half of those committing suicide receive a psychiatric diagnosis and the number of suicides is low compared to the number of individuals with a psychiatric disorder, so this is probably not affecting the estimates notably.

Furthermore, while we are likely to capture initial psychiatric diagnoses and incidence of disorders for most of the later-born cohorts, we may not capture all incidences for earlier-born cohorts and instead measure prevalence. However, this is likely to affect individuals with OC and the unaffected cohort in the same way and should thus not bias the results. It mainly involves the early birth cohorts and thereby the older age group, in which cases the increased relative risks may not be due to newly diagnosed psychiatric disorders, but could be driven by more re-admissions of prevalent cases.

Finally, it is possible that some individuals classified as non-syndromic OC in our analyses have unidentified associated anomalies that can also relate to increased risks of mental disorders. This is likely to be more of a concern for earlier birth cohorts that have greater underreporting of associated anomalies, but should be less of a concern for our estimates specific to later-born cohorts (S3 Fig illustrates the proportion of associated anomalies for individuals with OC by birth decades). Furthermore, analyses adding all syndromic and non-syndromic cases together did not substantially increase the risk estimates, indicating that any missed syndromic cases are unlikely to inflate the results for the non-syndromic group to any substantial degree. Furthermore, we expect that the major anomalies are most likely detected and, therefore, primarily the minor anomalies will be undetected; however, we cannot rule out the possibility that some of the increased risk is due to undetected syndromes among individuals with OC, such as 22q11 deletion syndrome. One limitation of register-based studies such as ours is the inability to contact individuals to conduct more detailed phenotyping and genotyping.

Conclusions

Individuals with non-syndromic OC in Denmark have a higher risk of being diagnosed with a psychiatric disorder at a psychiatric hospital compared to individuals without OC. However, the absolute risk increase is modest and the largest risk is for individuals with CP followed by CLP, whereas the risk is not elevated for individuals with CL. These increased risks are observed at both younger and older ages. Risks are mainly higher for schizophrenia-like disorders and mental retardation for individuals with CLP or CP and for personality disorders and pervasive developmental disorder for those with CP. There is no increased risk of mood disorders and anxiety-related disorders, however, which suggests that neurological differences or common genetic disposition are more plausible mechanisms for the association between OC and psychiatric disorders rather than the psychological consequences of being born with OC.