Study design

A population-based study with a case–control design was undertaken to investigate crude all-cause mortality in ONH/SOD.

Setting

Manitoba, a Canadian province with universal health care, has a population of 1.34 million residents (Statistics Canada Census, 2021). Patients with ONH/SOD are assessed and managed in the only tertiary pediatric hospital (Winnipeg Children’s Hospital) in the province.

Case ascertainment

Case ascertainment has been described in detail previously. ^{Reference Salman, Ruth, Yogendran and Lix7,Reference Salman, Ruth, Yogendran, Rozovsky and Lix13} Briefly, since 1986, a database of patients established by the section of pediatric endocrinology at Winnipeg Children’s Hospital was searched for cases with ONH/SOD up to August 2019. Additional cases were found through searching clinic letters from the sections of pediatric ophthalmology and pediatric neurology between 1990 and August 2019, the study end date for identifying cases. This search method ensured the systematic inclusion of patients with ONH/SOD. A diagnosis of ONH/SOD at any time defined the cases for this study.

Case definition

Septo-optic-pituitary dysplasia is diagnosed when patients with ONH have any of hypopituitarism, absent/small septum pellucidum or abnormalities in the corpus callosum. ONH was diagnosed when a small optic disc was reported by a pediatric ophthalmologist.

Manitoba residents only were included. Their diagnosis was confirmed by the first author (MSS), who has expertise in this disorder, by carefully reviewing all their medical hospital charts. Patients living outside Manitoba were not included, since their clinical charts were not linkable to provincial administrative health data.

Study population database

We used population-based administrative data from the Manitoba Population Research Data Repository at the Manitoba Centre for Health Policy (MCHP). ^{Reference Salman, Ruth, Yogendran, Rozovsky and Lix13,Reference Jutte, Roos and Brownell14} The repository contains de-identified and linkable databases encompassing all publicly funded healthcare services of residents of Manitoba. The validity and use of this repository have been documented. ^{Reference Singal, Brownell, Chateau, Ruth and Katz11,Reference Jutte, Roos and Brownell14}

One hundred and twenty-four independent and unrelated patients with ONH/SOD, who were born between April 1984 and March 2018, were available for matching and could be linked to MCHP Repository databases using a de-identified personal health identification number as described in prior studies. ^{Reference Salman, Ruth, Yogendran and Lix7,Reference Salman, Ruth, Yogendran, Rozovsky and Lix13} Tables 1 and 2 show their characteristics.

Table 1.

Pituitary gland functional status, frequency of hormonal replacement therapy, other morbidities and neuroimaging features in patients with optic nerve hypoplasia/septo-optic-pituitary dysplasia^†

S = suppressed per policy of data provider (Manitoba Centre for Health Policy, MCHP).

^†Adapted with permission from Salman MS et al. ^{Reference Salman, Ruth, Yogendran and Lix7}

^‡Some patients had more than one type.

Table 2.

Characteristics of the cases and their matched unrelated controls^†

^{Q = quintile.}

^†Adapted with permission from Salman MS et al. ^{Reference Salman, Ruth, Yogendran and Lix7}

^aAge was calculated in those alive on March 31, 2022.

^bFollow-up period is birth to March 31, 2022, date of leaving the province (if any) or date of death, whichever came first.

Controls

The unrelated control group was created from the Manitoba Population Research Data Repository databases as detailed previously. ^{Reference Salman, Ruth, Yogendran and Lix7,Reference Salman, Ruth, Yogendran, Rozovsky and Lix13} None had a diagnosis of ONH/SOD.

The control group was matched one case to five controls (N = 620) to achieve efficiency in the study design by using more controls to detect the association of group membership (i.e., case/control) with all-cause mortality. ^{Reference Coggon, Rose and Barker15} Matching variables included sex, birth date (within one year) and Health Region Zone at birth. There are five large health regions in Manitoba with 17 smaller divisions – known as Health Region Zones. These zones were used to achieve high accuracy when matching on geographic location.

Study dates

The study start date for assessing the outcome in cases and controls was the date of birth, when all health records of Manitoba residents are entered in the population health repository. The study end date was one of the following: March 31, 2022, the date of moving out of the province prior to March 31, 2022 (if any), or the date of death.

Ethics approval was granted by the Health Research Ethics Board, University of Manitoba, and approval for data access was provided by the Provincial Health Research Privacy Committee and provincial data custodians.

Study variables

The cohort demographics included age at the end of the study period (March 31, 2022 – the date of the latest available repository data coverage), sex, follow-up duration, income quintile (IQ) and neuroimaging abnormalities. More details regarding neuroimaging features have been published previously. ^{Reference Salman, Hossain and Rozovsky2,Reference Salman, Ruth, Yogendran, Rozovsky and Lix13} The date of death was identified for cases and controls who died prior to the study end date of March 31, 2022, from the population registry.

Supplementary Tables S1 and S2 provide detailed information on the study data sources and definitions of variables.

Data sources

The data on cases and controls were obtained from the following databases at MCHP: population registry and Statistics Canada Census (see Supplementary Table S2 for details).

Statistical analysis

Study data were descriptively analyzed using means and standard deviations (SD) (for normally distributed data), medians with interquartile ranges (for data that did not have a normal distribution), frequencies and percentages as appropriate. Follow-up duration for cases and controls was the difference between the date of birth and study end date (March 31, 2022), date of moving out of the province prior to March 31, 2022 (if any), or date of death, whichever came first. We calculated the median duration and the interquartile range.

Crude all-cause mortality rates were calculated for cases and controls per 1000 person-years. Cox proportional hazards semi-parametric regression models were used to test for a difference in the time to all-cause death for cases and controls; censoring was at the date of death, end of the study period (i.e., March 31, 2022) or date of moving out of province prior to March 31, 2022. The assumption of proportional hazards was assessed via visual assessment of the plots of the survival function for the groups. The hazard ratio (HR) and 95% confidence interval (CI) were estimated. A nominal α = 0.05 was used to assess statistical significance. All statistical tests were two sided.