**Purpose.**:
To analyze the effect of using one reading, the mean of two readings (from the same eye), or the mean of four readings (two from each eye) on the heritability estimates of intraocular pressure (IOP). This was a cohort study in which 344 pairs of twins, 163 monozygotic (MZ) and 181 dizygotic (DZ), were enrolled.

**Methods.**:
IOP was measured using three tonometers: the gold standard Goldmann applanation tonometer (GAT), the Ocular Response Analyzer (ORA; Reichert Buffalo, NY), and the Dynamic Contour Tonometer (DCT, Pascal; Swiss Microtechnology AG, Port, Switzerland). The main outcome measure was the heritability of IOP correlated with the number of measurements.

**Results.**:
The mean IOPs of all four readings with the three tonometers were: 14.1 ± 2.9 mm Hg for GAT, 15.9 ± 3.2 mm Hg for ORA, and 16.9 ± 2.7 mm Hg for DCT. As the number of readings increased, the calculated heritability (*h* ^{2}) of IOP measured using the GAT readings increased from 0.56 for one reading (95% confidence interval [CI], 0.44–0.65) to 0.58 for the mean of two readings (95% CI, 0.46–0.67) to 0.64 for the mean of all four readings (two right and two left; 95% CI, 0.55–0.72). Similar results were seen with the other two instruments.

**Conclusions.**:
The results demonstrated that the use of the mean of several readings from both eyes reduced measurement error, yielding a higher heritability estimate. The higher heritability would increase the power to detect linkage in a genome-wide analysis.

^{ 1 }Currently identified glaucoma-susceptibility genes contribute to pathogenesis in only a minority of cases.

^{ 2–4 }Given the relative rarity of glaucoma in the population, researchers have focused on the intermediate traits of glaucoma such as intraocular pressure (IOP). The heritability of IOP, defined as the proportion of variance attributed to genetic factors, has been estimated in family and twin studies to range between 0.30 and 0.64. In all these studies, the IOP from only one eye was used.

^{ 5–10 }In one of them, the higher IOP measurement from either eye was used for quantitative analyses.

^{ 6 }Of the published linkage analyses for IOP, data from one eye was used in two studies,

^{ 11,12 }whereas in another study, measurements were taken in duplicate on each eye, and the average was used.

^{ 13 }

^{ 14 }it is not possible to use the data from both eyes. Where there are measures from both eyes such as IOP, the use of only one eye for statistical analysis is valid, but inefficient.

^{ 15 }Bias may be introduced if data from one eye are actively selected (such as highest IOP) from data available from both eyes. The use of data from one eye per individual avoids the statistical analysis complication of high correlation between eyes, but there may be a considerable waste of available data.

^{ 15 }Newcombe and Duff

^{ 16 }concluded that averaging the data from both eyes gives greater precision and greater power to detect a difference of a given size.

*h*

^{2}) and the broad-sense heritability (

*H*

^{2}), respectively.

^{ 17 }Repeated measurements of an individual can be taken for some traits such as IOP. If it is assumed that these repeated measures are expressions of the same genotype, then the variation within individuals is caused by measurement error and other random environmental factors.

^{ 17 }

^{ 18 }although they were recruited for studies other than eye studies and subsequently asked to attend for an eye examination. Historically, the TwinsUK registry was established to investigate predominantly female disorders, such as osteoporosis. Since then, the number of males recruited has increased, but the majority of twins who volunteer are female, consistent with other volunteer twin registries. All subjects provided informed consent, in accordance with the Declaration of Helsinki, and the study was reviewed by the Local Research Ethics Committee. Six hundred ninety-two individuals had two readings performed on both eyes with all three instruments by a single investigator (FC). Twins of each pair were tested in immediate succession.

^{ 19,20 }Goldmann correlated IOP (IOPg) readings were used in the study. The DCT is a contact tonometer with a concave surface with a radius of curvature of 10.5 mm, which creates a distribution of forces between the central contour of the tip and the cornea that equals the forces generated by the internal pressure of the eye.

^{ 21 }It gives a measure of IOP that is independent of CCT.

^{2}. The submodel with the lowest Akaike information criterion is the best fitting. Heritability was calculated for first readings from the right eye, the mean of the first and second readings from the right eye, and the mean of all four readings (two right and two left). Results for right and left eyes demonstrated no significant differences, and so right eye measures were arbitrarily used for analysis, as in the previous heritability studies.

^{ 22 }which provides automated power analysis for linkage and association tests. In power calculations, assumptions are made, such as the proportion of variance explained by the trait locus, gene action, and marker heterozygosity.

^{ 22 }Other parameters that must be specified for a nonparametric variance components linkage analysis are the quantitative trait locus (QTL) heritability (i.e., the proportion of variance explained by the trait locus), the mode of inheritance at the locus, and the marker heterozygosity. When making these calculations, we did not test for dominance, and therefore it cannot be specified in the model. The module used was the QTL linkage for sibships; the number of sibships considered was two. The power to detect linkage with α = 0.05 was taken, with the sample size required for 80% power. In our calculations, we also assumed that the recombination fraction was 0. In reality, although possible, this is rarely the case; however, the assumption was the same for all calculations. It therefore would not alter the results in terms of comparison of numbers of readings.

Monozygotic | Dizygotic | |
---|---|---|

Number of twin pairs | 163 | 181 |

Mean age (range), y | 50.2 (16.1–81.0) | 55.1 (19.2–77.0) |

Female twins, % | 95.6 | 89.8 |

GAT-IOP (range), mm Hg | 13.7 ± 2.7 (7.5–22) | 14.3 ± 2.9 (8.3–24.3) |

ORA-IOP (range), mm Hg | 15.6 ± 3.1 (8.2–26.6) | 16.1 ± 3.2 (8.7–25.8) |

DCT-IOP (range), mm Hg | 16.9 ± 2.6 (11.3–26.9) | 17.0 ± 2.8 (11.3–27.2) |

No. of Readings | ORA | DCT | ||
---|---|---|---|---|

rMZ | rDZ | rMZ | rDZ | |

1 | 0.56 | 0.28 | 0.59 | 0.25 |

2 | 0.61 | 0.26 | 0.60 | 0.24 |

4 | 0.69 | 0.35 | 0.66 | 0.32 |

*h*

^{2}) of IOP from the GAT readings increased with increasing the number of readings from 0.56 (95% confidence interval [CI], 0.44–0.65) for one reading, to 0.58 (95% CI, 0.46–0.67) for the mean of two readings, to 0.64 (95% CI, 0.55–0.72) for the mean of all four readings (two right and two left). The remaining proportion of variance was due to individual environmental effects of 0.44 (95% CI, 0.33–0.53), 0.42 (95% CI, 0.35–0.56), and 0.36 (95%CI, 0.28–0.45) for one, two, and four readings, respectively. The results of the calculated

*h*

^{2}with the three instruments are tabulated for comparison in Table 3. There was a significant difference in standard deviation for GAT, when comparing one reading and the mean of four (

*P*= 0.01). The differences comparing one and the mean of two readings (

*P*= 0.15) and comparing the mean of two to the mean of four readings (

*P*= 0.11) were not statistically significant.

No. of Readings | GAT | ORA | DCT | |||
---|---|---|---|---|---|---|

h ^{2} | Environment | h ^{2} | Environment | h ^{2} | Environment | |

1 | 0.56 (0.44–0.65) | 0.44 (0.35–0.56) | 0.55 (0.44–0.64) | 0.45 (0.36–0.56) | 0.57 (0.47–0.66) | 0.43 (0.34–0.53) |

2 | 0.58 (0.46–0.67) | 0.42 (0.33–0.53) | 0.61 (0.50–0.69) | 0.39 (0.31–0.50) | 0.59 (0.48–0.68) | 0.41 (0.32–0.52) |

4 | 0.64 (0.55–0.72) | 0.36 (0.28–0.45) | 0.69 (0.60–0.76) | 0.31 (0.24–0.40) | 0.67 (0.57–0.74) | 0.33 (0.26–0.43) |

No. of Readings | GAT | ORA | DCT |
---|---|---|---|

1 | 7825 | 8127 | 7527 |

2 | 7231 | 6366 | 6939 |

4 | 5538 | 4256 | 4753 |

^{ 9,23 }

^{ 17 }Heritability ranges from 0 (all variation environmental) to 1 (all variation due to genetic factors). One contributor to the individual environmental component is measurement error.

^{ 9 }In a previously published study by our group, the estimated heritability using the mean of four readings with GAT was 0.62,

^{ 23 }whereas in this present study it was 0.64. The reason for this slight difference in heritability estimates probably reflects a smaller group of twins studied: data were available on 422 pairs of twins for our first study, whereas for this present study only 344 pairs, who had measurements from all three instruments, were analyzed. The main scope of this current work was not to determine the heritability of IOP, but to examine the effect of increasing the number of readings and therefore reducing the measurement error on the heritability estimate. Heritability is a population-specific factor, and our study applies to this population of largely Caucasian British females; the results could be different for other populations. However, the principles of repeated measures resulting in reduced measurement error and higher heritability are likely to apply to all populations.

^{ 5–8,10 }The lower heritability values may reflect the fact that those studies were not twin-based. Twin studies tend to show a higher heritability, in part due to the shared common environment that twins have, such as age, intrauterine environment, and family background. From this study, it was demonstrated that by increasing the number of readings used to calculate the heritability, the value was increased.

^{ 24 }Trait variation within and between populations results from genetic differences between people as well as environmental effects. These within- and between-population differences reflect underlying genetic and environmental variation. If errors have occurred in the measurement of the trait, due to either instrument or human measurement errors, they can compromise many statistical methods and may falsify the partitioning into genetic and environmental components.

^{ 25 }Macgregor et al.

^{ 26 }examined the effect of measurement error on heritability using height as the variable and comparing the heritability determined from clinical measures of height with that determined from self-reported height. They found that there was a bias toward overestimating one's height in the reported group. They concluded that moderate reduction in measurement error (through the use of accurate clinical measurement or multiple self-report measures) increases the effective sample size by 22% and that eliminating the measurement error completely leads to increases in effective sample size of up to 41%. This finding is in agreement with our use of the mean of several measurements in the present study.

^{ 5,7,9,10 }The eye was selected randomly or the one with the highest IOP was used. In all these studies, except the Beaver Dam Eye Study

^{ 7 }for which only one reading was taken, the mean of two or three readings was used for analysis. In an earlier study by Levene et al.,

^{ 8 }there is no clear reference to the number of readings used for the analysis. It is fair to assume that if the other studies had used the mean of more readings for their analyses, or the readings from both eyes, they too may have found higher heritability estimates for IOP.