The interblink interval in normal and dry eye subjects

Figures & data

Figure 1 Illustration of the transformation of blinks over seconds to IBIs over blink number to the distribution of IBI based on the first 30 seconds of recorded time for a normal subject. (A) Blink sequence (red dots) over time in seconds (blue timeline): 12 blinks (beyond the initial ten blinks) over 30 seconds. (B) IBI sequence (blue line) over blink number: 12 IBIs over 12 blinks. (C) IBI histogram (yellow bars) overlaid with a log normal distribution (blue curve).

Notes: The number and percent of IBIs in the intervals (0–1 seconds, 1–2 seconds, 2–3 seconds, 3–4 seconds, 4–5 seconds) were (0, 5, 4, 2, 1) and (0%, 42%, 33%, 16%, and 8%), respectively.
Abbreviations: IBI, interblink interval; N, number; SD, standard deviation.

Table 1 Baseline signs and symptoms of the ten dry eye subjects included in the study

Parameter	Mean ± standard deviation
Visual acuity	0.0778 ± 0.118
TFBUT*	4.46 ± 2.12
Discomfort**	1.30 ± 0.919
Burning†	1.30 ± 1.42
Dryness†	2.20 ± 1.14
Grittiness†	1.90 ± 1.29
Stinging†	1.60 ± 1.35
Inferior corneal staining**	2.81 ± 0.855
Superior corneal staining**	2.06 ± 0.818
Central corneal staining**	1.92 ± 0.952
Temporal conjunctival staining**	1.94 ± 1.04
Nasal conjunctival staining**	2.25 ± 0.696

Notes:

* Average of two readings,

** 0-4 Scale, each eye graded separately,

† 0-5 Scale, one grade given for both eyes.

Figure 2 IBI time series for a normal and dry eye subject. IBI time series for a normal and dry eye subject plotted on log-linear axes: observed IBI (solid red or blue lines) and predicted IBI for a given blink number based on the previous blink number (dashed green line). (A) Blue = observed; green = predicted for normals; (B) red = observed; green = predicted for dry eye.

Abbreviation: IBI, interblink interval.

Figure 3 Log IBI standard deviations versus log IBI means for ten normal and ten dry eye subjects.

Abbreviation: IBI, interblink interval.

Table 2 Group means, SD from the mean, and AC for IBI and log IBI in ten normal and ten dry eye subjects

IBI	Normal subjects	Dry eye subjects	Ratio	P-value
Mean	5.93	2.65	2.24	0.003
CV	0.77	0.66	1.17	0.359
AC	0.14	0.17	0.85	0.711
Log IBI	Normal subjects	Dry eye subjects	Ratio	P-value
Mean	1.32	0.58	0.75	<0.001
SD	0.84	0.61	0.24	0.024
AC	0.26	0.21	0.06	0.415

Notes: Group differences were tested using the Student’s t-test, with P < 0.05 demonstrating significance.

Abbreviations: SD, standard deviation; AC, autocorrelation; IBI, interblink interval; CV, coefficient of variation.

Table 3 Random effects model: IBI and log IBI in ten normal and ten dry eye subjects

IBI	Normal subjects	Dry eye subjects	Ratio	P-value
Mean	5.97	2.56	2.33	0.004
CV	1.10	0.70	1.56	<0.001
AC	0.28	0.16	1.79	0.044
Log IBI	Normal subjects	Dry eye subjects	Difference	P-value
Mean	1.32	0.58	0.74	0.004
SD	0.89	0.64	0.25	<0.001
AC	0.28	0.16	0.12	0.044

Notes: Group differences were tested using the Student’s t-test, with P < 0.05 demonstrating significance.

Abbreviations: IBI, interblink interval; AC, autocorrelation; CV, coefficient of variation; SD, standard deviation.

Figure 4 Predicted probability of a subject having dry eye based on their mean log IBI.

Notes: The fit of a logistic model using the mean log IBI as the single predictor. Twenty binary observations are illustrated: “1” for the ten dry eye subjects (red dots), and “0” for the ten normal subjects (blue triangles). The green dashed fitted curve from the logistic model shows the predicted probability of a subject having dry eye based on their mean IBI. Predictions for the 20 subjects are represented by dots and triangles on this curve. Note that as the mean IBI increases, the probability of dry eye decreases; red dots on the lower half of the fitted curve would be false negatives and blue triangles on the upper part of the fitted curve would be false positives.
Abbreviation: IBI, interblink interval.

Figure 5 (A) The ROC curve based on one predictor (mean IBI) AUC = 0.85 and (B) on three predictors representing, IBI level (mean), variability (standard deviation), and relationship to past IBIs (autocorrelation), AUC = 0.89.

Notes: The cutoff points of ≤1.12 for the log mean IBI, ≤0.65 for the log standard deviation, and ≤0.21 for the log autocorrelation give a true positive probability of 0.8 and a false positive probability of 0.2 (shown as a blue square on the ROC curve). On the raw scale, this corresponds to a median IBI of ≤3.05 seconds and a coefficient of variation of ≤0.73.
Abbreviations: ROC, receiver operating characteristics; IBI, interblink interval; AUC, area under the curve.