Distribution of family score in case (n = 37) and control (n = 83) probands. Black bars indicate case probands; shaded bars, control probands. Affected relatives had definite or probable open-angle glaucoma (see the "Measurements and OAG Definitions" subsection of the "Methods" section for the definitions of these classifications). Family score values ranged from − 0.98 to 7.08 and were stratified in categories of 0.5 units of family score (lowest category, − 0.75 to − 0.25; highest category, 6.75 to 7.25). A high family score represents a greater likelihood for open-angle glaucoma.
Hulsman CAA, Houwing-Duistermaat JJ, van Duijn CM, Wolfs R, Borger PH, Hofman A, de Jong PTVM. Family Score as an Indicator of Genetic Risk of Primary Open-Angle Glaucoma. Arch Ophthalmol. 2002;120(12):1726-1731. doi:10.1001/archopht.120.12.1726
To assess the genetic risk of open-angle glaucoma (OAG) in individuals by calculating a family score (FS), which summarizes the information of all relatives including their disease status, age, sex, and degree of kinship and to examine the genetic contribution to OAG with and without an increased intraocular pressure.
Case and control probands, derived from the Rotterdam Study, underwent the same ophthalmologic examination as their relatives. The FS of each proband was the sum of the differences between observed and expected values of OAG for all relatives. The FSs were compared between case and control probands using logistic regression analysis, adjusted for intraocular pressure.
Of 37 case probands, 44 (half-) siblings and 86 children were available, and of 83 control probands there were 97 (half-) siblings and 155 children. Family scores ranged from − 0.44 to 7.08 in case probands and from − 0.98 to 2.46 in control probands. One unit increase in FS was significantly associated with a higher risk of OAG (odds ratio, 1.59; 95% confidence interval, 1.14-2.23). Adjustments for intraocular pressure did not change the odds ratio.
These data show that the FS strongly predicts OAG, independent of the intraocular pressure. Therefore, the FS is useful to identify individuals with a high genetic risk.
PRIMARY OPEN-ANGLE glaucoma (POAG) could be described as a degenerative process of the ganglion cells in the retina. This neurodegeneration is clinically characterized by a glaucomatous optic neuropathy (GON) with a glaucomatous visual field defect (GVFD) and open chamber angles, in the absence of another(secondary) glaucoma cause. The prevalence of open-angle glaucoma (OAG) is about 1% in subjects aged 55 years and older in the western world.1- 4 A positive family history and an elevated intraocular pressure (IOP) have consistently been shown to be major risk factors.5- 7 It is generally recognized that genetic factors play an important role. So far, genetic factors usually were not considered to be a secondary glaucoma cause in the definition of OAG, with the exception of the autosomal recessive disorder buphthalmos. Linkage with several genomic regions has been found in different selected families with POAG, where the disease segregates as an autosomal dominant trait.8- 13 Two genes have been identified. The myocilin gene accounts for 3% to 5% of adult-onset OAG14; mutations in the optineurin gene are found in 16.7% of families with hereditary OAG.15 Apart from this subset of families, most families with POAG do not follow a clear-cut mendelian inheritance pattern. Instead, POAG is considered a heterogeneous, multifactorial disease, in which probably multiple genetic and nongenetic factors are involved.16- 19 To quantify the effects of these factors and interactions between them on the origin of POAG, family studies are carried out.
Recently, a familial aggregation study had been performed in (half-) siblings and children of case and control subjects with OAG, drawn from the population-based Rotterdam Study.20 Because pseudoexfoliation was not ruled out at baseline, we spoke about OAG. No pseudoexfoliation was seen at the follow-up examination and, thus, we consider these data similar to POAG. In this family study, the lifetime risk of OAG in relatives of cases was 9.2 compared with relatives of controls.
Herein we use a new approach that integrates all information from a family to calculate a family score (FS). Houwing-Duistermaat and Van Houwelingen21 derived that a FS defined as the sum of the weighted difference between the observed and expected value of each relative for a certain disease performs well under various genetic mechanisms. The expected value for each relative is based on the age- and sex-adjusted prevalence in the population from which the families are drawn. The weights are proportional to the degree of kinship between the relative and the proband. This method summarizes the information about all relatives including their disease status, age, sex, and degree of kinship.
The first aim of this study was to apply this novel method to our family data of OAG in the Rotterdam Study and to compare the family-derived FS between probands with OAG and control probands. The second aim of the study was to address the issue of clinical heterogeneity in OAG. Therefore, we examined whether there was evidence for a difference in the contribution of genetic factors to the prevalence of OAG with and without an increased IOP.
Design and methods of the familial aggregation study have been described elsewhere.20 Case and control individuals, herein to be called "case" and "control" probands to discern them from their relatives, were recruited from the baseline phase of the Rotterdam Study. This population-based cohort study aims to assess the occurrence of and the risk factors for chronic diseases in subjects aged 55 years and older.4,22 The study was performed according to the Declaration of Helsinki and was approved by the Medical Ethics Committee of Erasmus University, Rotterdam. Written informed consent was obtained from all participants.
In the present analysis, the group of case probands differed from the group in the original aggregation study because we recently changed our definitions of OAG.4 Of the 48 case probands from the original familial aggregation study,20 37 probands fulfilled the new OAG criteria. Together with 13 new case probands, a total of 50 case probands had definite OAG. Three of them were deceased at the time of the family study, resulting in 47 eligible case probands for the present study. Of the 135 control probands (87%) who consented to participate in the original familial aggregation study, 23 were excluded because after resetting the definition of OAG, they met the criteria for possible or probable OAG. Thus, 112 control probands were eligible for the present study.
Relatives of case and control probands underwent a standardized ophthalmologic examination, performed by 3 investigators.4,7 In short, the ophthalmologic examination included measurement of IOP, direct and indirect ophthalmoscopy, stereofundus photography, and simultaneous stereo disc photography. Optic disc characteristics were measured with a digital image analyzer (Imagenet System; Topcon Optical Co, Tokyo, Japan) and an ophthalmoscope.23 Visual fields were tested with the full threshold 24-2 application of the Humphrey perimeter (Humphrey Visual Field Analyzer II, model 750; Carl Zeiss, Dublin, Calif) and repeated when unreliable.
Cases of OAG were classified into 3 categories according to our new definition4: definite, probable, or possible OAG. Definite OAG was defined as a GVFD in combination with either a probable GON (vertical cup-disc ratio ≥0.8, asymmetry in vertical cup-disc ratio≥0.3, or neuroretinal rim <0.05) or a possible GON (vertical cup-disc ratio 0.7, asymmetry in vertical cup-disc ratio 0.2, and neuroretinal rim 0.1, respectively). Probable OAG was defined as a probable GON without a GVFD or a GVFD without any GON. Cases of possible OAG had a possible GON without a GVFD.
Only cases of definite OAG were included as case probands. Control probands were included when they had neither GON nor GVFD, nor age-related maculopathy.24 The IOP was defined as elevated if it was higher than 21 mm Hg during our examination or if treatment to lower the IOP had been used. To investigate whether hypertension or diabetes mellitus could account for the familial risk, either directly, or through their relation with IOP, these disorders were included in the analyses. Hypertension was defined as a systolic blood pressure of 160 mm Hg or higher, or a diastolic blood pressure of 95 mm Hg or higher, or the use of antihypertensive treatment. Diabetes mellitus was considered present if the nonfasting serum glucose level was 200 mg/dL (11.1 mmol/L) or higher or if antidiabetic medication was used.
Baseline characteristics of case and control probands were compared using the t test. For each family a FS was computed, based on the disease status, number of relatives affected, age, and sex of all relatives available for analyses. As a few half-siblings were included, a correction was made for the degree of kinship. The FS was calculated according to the formula described by Houwing-Duistermaat and Van Houwelingen21:
where the sum (Σ) is taken over the values of all relatives j of a proband; Ψj is the kinship coefficient between each relative j and the proband (0.25 for siblings and children of the proband, and 0.125 for half-siblings); Oj is the observed disease status(1, affected, 0, otherwise); and Ej is the expected value for each relative j. Expected values were age- and sex-specific prevalences of OAG in the Rotterdam Study, based on the digital image analyzer data combined with ophthalmoscopic data (Table 1); for relatives younger than 55 years a prevalence of 0% was assumed. For the purpose of presentation, we multiplied the sum of all individual scores by 10. As an example, for a family that consists of relatives a, b, and c, the formula can be written as:
Logistic regression analysis was performed to estimate the odds ratio(OR) of OAG in probands given the FS. First, analyses were performed after calculating FSs assigning both relatives with definite [d]OAG and relatives with probable [p]OAG as affected (FSdp): definitely and probably affected relatives were included as observed cases; expected values were derived from prevalences of definite and probable OAG (Table 1). Next, the analyses were repeated after calculation of the FSs assigning only definitely affected (FSd) relatives as observed cases and excluding relatives with probable OAG(Fp). Expected values were derived from prevalences of definite OAG. Relatives with possible OAG and without OAG were assigned as unaffected.
Analyses were adjusted for age, sex, diabetes mellitus, and hypertension status of the proband. To examine whether an elevated IOP had effect on the relation between FS and OAG, additional adjustments were made for continuous IOP and IOP-lowering treatment. We tested whether the relation between FS and OAG was different for OAG with and without an IOP of more than 21 mm Hg. Therefore, an interaction term between FS and IOP (1, elevated; 0, normal) was added to the model including all variables except the continuous IOP and the IOP-lowering treatment.
Of the 47 eligible case probands, 43 (91%) consented to participate. For 6 case probands, no relatives were available for analyses. Among the relatives of case probands, 48 (half-) siblings (80%) and 87 children (95%) responded, of whom 42 siblings, 2 (half-) siblings, and 86 children had complete data to diagnose or to exclude the presence of OAG. Of 112 control probands, 101(90%) participated, and 110 (half-) siblings (81%) and 158 children (81%) responded. Complete data on OAG status were available for 95 siblings, 2 (half-) siblings, and 155 children. For 18 control probands, no relatives were available for analyses. This resulted in 37 case and 83 control probands with families available for analyses (Table 2).
The mean age and proportion of females were not significantly different between case and control probands. Of the 37 case probands, 19 (51%) were previously known to have an elevated IOP and be receiving IOP-lowering treatment. Their maximum IOP before or during treatment often was not known. In 3 of the remaining 18 cases an IOP exceeding 21 mm Hg was measured during our examinations. Of control probands, 5 (6%) had IOP-lowering treatment and in 7 of the remaining controls, we measured an IOP exceeding 21 mm Hg.
The number of examined relatives varied from 1 to 9 among case probands and from 1 to 11 among control probands and did not differ significantly (P = .29). The number, mean age, and proportion of women were similar in (half-) siblings of cases and controls. These characteristics were also similar in their children, except for age: children of cases were on average 2.4 years younger than children of controls (46.1 vs 48.5 years).
Relatives of cases had a higher mean IOP than relatives of controls, especially siblings (17.7 vs 13.8 mm Hg). (Half-) siblings of cases also had received IOP-lowering treatment more often than those of controls (11% vs 2%). The distribution of case and control families with different numbers of affected relatives is given in Table 3.
First, FSdps were calculated (Figure 1). In OAG case probands, the FSdp ranged from–0.44 to 7.08, with a mean FSdp of 0.98. In control probands, FSdps were lower, ranging from –0.98 to 2.46, with a mean FSdp of 0.24.
Logistic regression analysis showed that an increase of 1 unit in FSdp was associated with a significantly increased risk of OAG in the probands (OR, 1.59; 95% confidence interval [CI], 1.14-2.23), adjusted for age and sex (Table 4). With the inclusion of IOP as a continuous variable, and IOP-lowering treatment, diabetes mellitus, and hypertension of the proband as covariates, the risk of OAG was even more increased per unit of FSdp and also significant (OR, 1.76; 95% CI, 1.15-2.70). After inclusion of the interaction term between FSdp and an elevated IOP, this term was not associated with OAG(OR, 1.10; 95% CI, 0.41-2.94), while FSdp alone was still significantly associated with OAG (OR, 1.67; 95% CI, 1.05-2.65).
The FSd in case probands ranged from − 0.14 to 4.89, with a mean FSd of 0.48, and in control probands ranged from − 0.29 to 2.41, with a mean FSd of − 0.01. An increase of 1 unit of FSd was associated with a higher risk of OAG in the proband(OR, 3.32; 95% CI, 1.25-8.82) than 1 unit increase of FSdp. Adjusted for all risk factors mentioned earlier, and after the additional adjustments of IOP and IOP-lowering treatment, the OR did not change and was still statistically significant. Finally, the interaction term between FSd and an elevated IOP was not significantly associated with OAG in the proband (P = .67), while FSd alone was still significantly associated with OAG.
To illustrate the FS method, we describe 3 probands with different FSdps. The proband with the highest FSdp (7.08) had 8 siblings, of whom 3 were affected. Two sisters, aged 71 and 72 years, had definite OAG and 1 sister, aged 73 years, had probable OAG. The age- and sex-adjusted prevalence for each of them was 0.029 (Table 1). Subtracting the expected (0.029) from the observed value (1) and weighting with Ψj (0.25) resulted in an individual score of (1 − 0.029) × 0.25 = 0.24275 each. Three unaffected sisters aged 63, 65, and 74 years had expected values of 0.018, 0.035, and 0.029, respectively, resulting in individual scores of − 0.0045, − 0.00875, and − 0.00725, respectively, after weighting. Two unaffected sons, aged 44 and 41 years, had an expected value of 0, resulting in an individual score of 0. The sum of these individual scores of all relatives was an FSdp of 0.24275+ 0.24275 + 0.24275–0.0045–0.00875–0.00725 + 0 + 0 = 0.708. Multiplying by 10 resulted in an FSdp of 7.08.
In relatives of a second proband, no OAG was observed among all 8 siblings, aged between 67 and 80 years, and 2 children, aged 48 and 57 years. Subtracting the expected values from these observed values of 0 resulted in negative individual scores. After weighting with j, these individual scores were added up to an FSdp of − 0.098 and multiplied by 10 to arrive at− 0.98.
A third proband had 7 unaffected children. Since they were relatively young (age range, 35-52 years), expected values were 0, resulting in individual scores of 0, and, subsequently, an FSdp of 0. This example demonstrates that a proband with an FSdp equaling 0 does not yield much information about the genetic risk of his or her family.
In this study, we applied a new statistical method to estimate the risk of OAG for an individual, given the disease distribution in his or her relatives, based on the results of ophthalmologic examinations. The presence of OAG in(half-) siblings and children was quantified by an FS that considered the disease status, number of relatives affected, age, sex, and degree of kinship coefficient of the relatives. The FS, which can be interpreted as the genetic risk for that particular individual, showed considerable variability among case probands. We found that an increase in FS was significantly associated with an increased risk of OAG. This association was independent of the presence of an elevated IOP.
Family scores are distributed around 0. A positive FS indicates that in a family, more cases are observed than expected and, therefore, points to genetic risk in that family and vice versa. The magnitude of the FS depends on the size and age distribution in the family. Small and/or young families do not contain much information about the genetic risk of the proband. This is reflected in the model by the fact that small families produce an FS of 0 or close to 0. As long as candidate genes are not identified, the risk estimation of OAG for the individuals with FSs around 0 should not be based on the occurrence of OAG in their family, but on other risk factors. Since the sizes and age distributions of the case families were similar to those of the control families, confounding by family size was unlikely.
A strength of the study is the fact that all case and control probands and their relatives underwent a standardized ophthalmologic examination and that we did not use the family history. Features of OAG were assessed separately in a masked fashion to ensure an unbiased diagnosis. Examining relatives instead of taking a family history from the probands reduces misclassification of disease status that otherwise might occur owing to the insidious course of the disease, lack of knowledge of the disease status in the proband or the relative, and different and changing definitions of OAG. By recruiting probands from the population-based Rotterdam Study, ascertainment bias due to family size and selection bias toward a specific type of OAG was reduced. By using strict criteria for the definition of definite OAG, we excluded a group of cases with questionable OAG. Therefore, the resultant FS was associated with true moderate to advanced OAG.
The absence of significance of systemic hypertension and diabetes mellitus may be explained by the limited number of OAG cases together with the inclusion of several factors in one model. A second explanation may be the stronger association of OAG with familial factors than with hypertension and diabetes mellitus. The latter explanation would be in accord with findings from the literature,7,25,26 where familial factors are more consistently and more strongly found to be associated with OAG than hypertension and diabetes mellitus.
Our findings indicate that the contribution of genetic factors to the occurrence of OAG is independent of IOP. Both low and high FSs were observed in case probands with and without an elevated IOP (data not shown). Adjustment for IOP did not change the effect of FS on the risk of OAG; the OR of OAG remained significantly increased. Finally, the interaction term between FS and IOP was not significant and did not change the effect of FS. Thus, FS had an effect on OAG with and without an elevated IOP, and the effect of FS on OAG was not statistically different within these 2 subgroups (data not shown). Our findings match the observation in literature where families with multiple cases of OAG and a high IOP10,11,13 low IOP,12,27 and both9,28 are reported, as well as the implication of the myocilin gene in OAG with an elevated IOP,14 and the optineurin gene in OAG without an elevated IOP.15 It is hypothesized that the optineurin gene plays a role in neuroprotection. However, the design of many studies complicates distinguishing between the genetic influence leading directly to OAG and that leading to OAG through IOP, because OAG through IOP is often included in the definition of OAG. Genetic factors are involved in IOP,29- 31 while IOP is considered an important risk factor for OAG. In our study, we were able to investigate the effect of IOP apart from OAG because the level of IOP was not included in our definition of OAG and because an elevated IOP was defined independently from the OAG diagnosis.
It is generally accepted that OAG is a multifactorial disease16- 19 caused by the actions of many genes and effects from the environment. Still, it is difficult to discriminate whether familial occurrence of a multifactorial disease is the result of shared genetic or environmental factors.32 It seems unlikely that shared environmental factors early in life contribute to the occurrence of OAG at middle age. Indeed, no environmental factor has consistently been associated with the risk of OAG, neither in clinic-based case-control studies19 nor in population-based studies.33- 35 Also, a recent twin study showed concordance of OAG in twins but not in spouses, suggesting the involvement of genetic rather than environmental factors.31 To identify possible environmental risk factors, further research in cohort studies is necessary. To separate the effects of genetic and environmental factors, and the interactions between these, genetic epidemiological studies in families with OAG should be carried out, preferably encompassing more generations.36 In such studies, the FS can be included as one of several risk factors. By doing so, the FS is used to represent the genetic risk of an individual. Furthermore, the FS is valuable to select individuals for clinical follow-up and for further genetic analysis. Research into genotype-phenotype relations in families will remain necessary, also, after identification of disease-causing genetic defects, to study possible modifier genes or environmental factors influencing their expression.
Submitted for publication November 30, 2001; final revision received August 2, 2002; accepted September 3, 2002.
This study was supported by grants from Zorg Onderzoek Nederland (28-2903), The Hague (Dr de Jong); Topcon Europe BV, Capelle and IJssel, the Netherlands.(Dr de Jong); Optimix Foundation, Amsterdam (Dr de Jong); the Netherlands Society for Prevention of Blindness, Doorn (Dr de Jong); Ondersteuning Oogheelkunde's Gravenhage Foundation, The Hague (Dr de Jong); Landelijke Stichting voor Blinden en Slechtzienden, Rotterdam (Dr de Jong); Blindenhulp Foundation, The Hague(Dr de Jong); Rotterdamse Vereniging voor Blindenbelangen (Dr de Jong); Rotterdamse Oogheelkundige Onderzoek Stichting Foundation (Dr de Jong); and Blindenpenning Foundation, Amsterdam (Dr de Jong).
We thank Ada Hooghart and Corina Brussee for assistance in data collection and optic disc grading, Tom van den Berg, PhD, Arthur Bergen, PhD, Willem Kamphuis, PhD, Caroline Klaver, MD, PhD, and Jacqueline Willemse-Assink, MD, PhD for their stimulating comments.
Corresponding author: Paulus T. V. M. de Jong, MD, PhD, FRCOphth, The Netherlands Ophthalmic Research Institute, KNAW, Meibergdreef 47, 1105 BA Amsterdam, the Netherlands (e-mail: email@example.com).