purpose. To test whether the regularity and short-term changes of retinal vessel diameter are related to the history of cold hands and feet and to nailfold circulatory response to cooling.

methods. In 13 vasospastic and 13 nonvasospastic young healthy women (based on their history of cold extremities and nailfold capillaroscopy) 20- to 30-second recordings of the ocular fundus was obtained with a retinal vessel analyzer. The spatial regularity of arterioles and venules was analyzed by means of the coefficient of variation of vessel diameter and by exploratory Fourier analysis of spatial frequencies. Temporal variability was analyzed as excursion amplitude of the vessel diameter, as a correlation of means and standard deviations of vessel diameter within a defined time period, and by Fourier analysis of temporal diameter change in the heartbeat frequency range.

results. Mean diameters of selected segments of arterioles (129.9 ± 15.3 and 124.4 ± 24.4 μm) and venules (150.8 ± 14.6 and 149.3 ± 19.6 μm) were not different between the vasospastic and nonvasospastic groups, respectively. Spatial variability: The coefficient of variation in arterioles was 8.8% ± 2.8% and 6.1% ± 1.7%, in venules 3.8% ± 1.4%, and 3.6% ± 0.9% in the vasospastic and nonvasospastic groups, respectively (difference by ANOVA, *P* = 0.017). Fourier analysis revealed differences between arterioles in the two groups, with relative Fourier power spectrum amplitudes of spatial frequencies higher in vasospastic eyes (Mann-Whitney *P* = 0.029). Temporal variability: The excursion amplitudes of vessel diameters were comparable in the two groups. Individual coefficients of correlation of successive means and standard deviations of the vessel diameter were 0.11 ± 0.23 and 0.09 ± 0.23 in the nonvasospastic group, and 0.25 ± 0.40 and 0.24 ± 0.22 in the vasospastic group, in the arterioles and venules, respectively (ANOVA: vasospastic versus nonvasospastic; *P* = 0.038). Fourier analysis in the heartbeat frequency range revealed differences in relative power spectrum amplitudes of temporal frequencies between arterioles in the two groups (higher in the vasospastic group, Mann-Whitney *P* = 0.029).

conclusions. Retinal arterioles in healthy vasospastic women show higher spatial irregularity and an increased vessel diameter variation within the temporal frequency of the heartbeat than do arterioles in nonvasospastic women.

^{ 1 }

^{ 2 }Vascular dysregulation has been associated with several ocular diseases, including glaucoma,

^{ 3 }central vein thrombosis,

^{ 4 }nonarteritic anterior ischemic neuropathy,

^{ 5 }and central serous chorioretinopathy.

^{ 6 }Changes in retinal vessels have also been observed in patients with vascular dysregulation in other organs, such as the heart

^{ 7 }or the brain.

^{ 8 }Evaluating morphology of retinal vessels has been mostly limited to funduscopic assessment, a method with poor reproducibility of measurements,

^{ 9 }

^{ 10 }or to static fundus photography.

^{ 11 }

^{ 12 }The retinal vessel analyzer offers high spatial vessel width resolution,

^{ 13 }and a high reproducibility of measurements.

^{ 14 }

^{ 15 }In contrast, temporal resolution down to 40 milliseconds enables dynamic analysis of the vessel diameter changes.

^{ 13 }

^{ 16 }Mean retinal vessel diameter was statistically comparable between vasospastic and control subjects.

^{ 17 }However, dynamic recordings may offer relevant clinical details. To investigate a potential yield of information with such an approach, spatial properties and short-term changes in the vessel diameter of retinal arterioles and venules in vasospastic eyes were analyzed with a retinal vessel analyzer (Retinal Vessel Analyzer [RVA]; IMEDOS GmbH, Weimar, Germany).

^{ 18 }subjects not taking contraceptive pills had to be in the postovulation phase of the cycle, which was verified by a subsequent phone interview ascertaining that menstrual bleeding had occurred less than 2 weeks after the study examination day.

^{ 19 }

^{ 13 }

^{ 16 }An essential part of the RVA device is the fundus camera (FF450; Carl Zeiss Meditec GmbH, Jena, Germany), which allows the examination and recordings of the ocular fundus. It incorporates the illumination and the observation optical pathway. After being reflected from the retina, the light is delivered through the observation pathway to the observation ocular and the charge-coupled device (CCD) chip of the video camera simultaneously. The standard video signal from the CCD then goes to the RVA control computer and to the SVHS recorder, which enables subsequent offline measurements of the recorded session later on. The measuring principle of the RVA is as follows: Inside the walls of the retinal blood vessels, there is a column of red blood cells, separated from the walls by the plasma edge stream. Red blood cells absorb one part of the light. The RVA measures the diameter of the column of the red blood cells. For measurements, the fundus camera is adjusted to the dilated pupil and a clear fundus image with good contrast and no reflection is obtained on the monitor. For the current experiment, temporal resolution was set at 40 ms, which translates to 25 captured video frames per second. RVA produces one vessel width measurement, expressed in units of measurement (UM), for each segment length of 12.5 UM. In an emmetropic person, 1 UM = 1 μm.

*coefficient of variation*. In each of the 250 video frames containing 40 individual data points as described, the mean, the standard deviation, and the coefficient of variation (CV of diameter values along the analyzed segment for a single frame = SD/mean · 100%) were calculated. Thus, CV is expressed as a percentage of the mean diameter of that frame. The obtained 250 CVs were then averaged, producing one mean CV per vessel. Each video frame was captured within 40 ms and contained the whole vessel segment. As each frame provided its own CV value, this method does not artificially decrease spatial variability by averaging. Overall mean CVs, one for each arteriole and a venule, were obtained.

*spatial frequencies*in the retinal vesselsof the two groups with an oscillating function, we performed a Fourier analysis of spatial frequencies in a selected vessel segment. As mentioned, one video frame produces a spatial series of 40 measurements: one diameter measurement every 12.5 UM of segment length. There were 250 such video frames. The following method was applied: Two groups of five consecutive video frames were sampled from data for each vessel and averaged across time, providing two spatial series of vessel diameters. In an attempt to sample data from different phases of the heart cycle, these two video frame groups were separated from each other with six video frames that were left out of averaging. The two averaged spatial series underwent the Fourier analysis separately. Not knowing the frequency range of interest in advance, we inspected each Fourier power spectrum curve and identified the spatial frequency with the highest relative amplitude (in other words, the spatial frequency with the highest statistical likelihood to be found in the spectrum of frequencies). Means of the two amplitudes and two frequencies are reported.

*excursion amplitude*(EA) was calculated. EA is a quantitative measure of the excursion of the vessel diameter during the defined time of 10 seconds. The mean diameter in each of the 250 frames in the same time–space tables from the previous two analyses was calculated, producing a data series of 250 mean diameters for 10 seconds. These were sorted in ascending order, and the mean of the lower and upper 25 values (corresponding approximately to the 5 th and 95 th percentiles) were used to calculate the EA of each vessel, according to the formula: EA = (mean of highest 25 values − mean of lowest 25 values)/(mean of highest 25 values + mean of lowest 25 values)/2) · 100%. Thus, EA was expressed as a percentage of the mean diameter. EAs, one for each arteriole and venule, were tested in a two-way ANOVA model, with vasospastic propensity as one factor (difference between the two groups), and the difference between vessels (arteriole/venule) as the other.

*pattern of temporal vessel diameter change*was calculated as a correlation between the means and standard deviations of the vessel diameter. Each video frame provided one mean and one SD of the vessel diameters along the segment. Pearson’s linear correlation coefficient between this mean and the corresponding SD was obtained in 250 successive frames for each vessel. In this model, a positive correlation would mean an increased vessel “waviness” (higher spatial irregularity) in the moment of largest average vessel diameter, no correlation would indicate a uniform diameter increase in the whole vessel segment, and “smoothening” of the vessel segment in the moment of largest average vessel diameter would result in a negative correlation.

*Fourier analysis of the temporal course of mean vessel diameter*was performed on the temporal series of 250 successive mean vessel diameters. Unlike in the Fourier analysis of spatial frequencies, which was more exploratory in character, in this case, an analysis of the power spectrum for the frequency range of 50 to 90 beats per minute was performed for each vessel. A mean relative Fourier power spectrum amplitude (the statistical likelihood of the given frequency to be found in the spectrum of frequencies) was calculated for the 50- to 90-bpm frequency range. This frequency range was based on the measured pulse rate (see the Results section).

*t*-test for independent samples,

*P*= 0.35), IOP was 13.2 ± 1.9 and 12.3 ± 2.2 mm Hg (

*P*= 0.30), systolic BP 113.1 ± 9.8 and 113.8 ± 13.2 mm Hg (

*P*= 0.87), diastolic BP 72.3 ± 9.5 and 72.7 ± 8.9 mm Hg (

*P*= 0.92), pulse rate 67.7 ± 8.6 (range, 56–83) and 69.3 ± 9.1 (range, 55–88) beats per minute (

*P*= 0.65), ocular perfusion pressure 44.1 ± 5.2 and 45.3 ± 5.6 mm Hg (

*P*= 0.58), spherical equivalent in the experimental eye −2.4 ± 1.5 (0 to −3.0) and −1.3 ± 1.2 D (+0.5 to −3.0;

*P*= 0.14) in the nonvasospastic and vasospastic groups, respectively. Both groups were also comparable in terms of contraceptive pill use (9 positive/4 negative vs. 10 positive/3 negative in the nonvasospastic and vasospastic groups, respectively; χ

^{2}

*P*= 0.66). Mean diameters of selected segments of arterioles (129.9 ± 15.3 and 124.4 ± 24.4 μm; unpaired

*t*-test

*P*= 0.50) and venules (150.8 ± 14.6 and 149.3 ± 19.6 μm; unpaired

*t*-test,

*P*= 0.82) were not different between the vasospastic and nonvasospastic groups, respectively.

*P*= 0.017) in general. However, arterioles and venules seem to behave differently in this respect (significant interaction,

*P*= 0.0097). Indeed, planned comparison testing revealed that the difference in coefficient of variation between the vasospastic and nonvasospastic groups is present only in arterioles (

*P*= 0.007).

*P*= 0.029) but not in venules (

*P*= 0.90). The corresponding spatial frequencies were 0.042 ± 0.013 spacial interval

^{−1}and 0.049 ± 0.017 spacial interval

^{−1}in arterioles and 0.052 ± 0.015 and 0.054 ± 0.024 in venules (1 spacial interval corresponds to 12.5 UM = 12.5 μm in an emmetropic eye), in the vasospastic and nonvasospastic groups, respectively. Being normally distributed, this parameter was subjected to two-way ANOVA: vasospastic versus nonvasospastic

*P*= 0.36, arterioles versus venules

*P*= 0.14, interaction

*P*= 0.61.

*P*= 0.0002), but it was not different between the vasospastic and nonvasospastic groups. In the analysis of the pattern of short-term vessel diameter change, the correlation of coefficients between successive means and standard deviations of the vessel diameters (Table 1 ; Fig. 4 ) were significantly different between nonvasospastic and vasospastic subjects (

*P*= 0.038). The observed positive correlation and testing in the ANOVA model indicated an increased vessel waviness at the average vessel diameter peak in vasospastic subjects, both in arterioles and venules. (Fig. 5) .

*P*= 0.029) but not in venules (

*P*= 0.90).

^{ 20 }

^{ 21 }

^{ 22 }To eliminate an effect of gender, we recruited only female subjects. One negative aspect of such an approach is that the relevance of the study findings for men is unclear.

^{ 23 }There was no difference in BP or IOP between our two cohorts, still we observed an increased “waviness” (spatial irregularity) in arterioles of vasospastic eyes. Digital image analysis has been used in the past to analyze the tortuosity of retinal vessels in retinopathy of prematurity.

^{ 24 }

^{ 25 }

^{ 26 }

^{ 27 }Venous beading in diabetic retinopathy has been analyzed by means of Fourier transformation of the data from digitalized fundus images.

^{ 28 }

^{ 29 }However, the origin of increased vessel waviness in patients with diabetic retinopathy and retinopathy of prematurity is probably different from our group of otherwise healthy subjects with vascular dysregulation. A suggested underlying cause of vascular dysregulation is endotheliopathy,

^{ 1 }

^{ 2 }and the observed phenomenon may be an indicator of endotheliopathy.

^{−1}, which in the present setting translates to a repetitive oscillating spatial sequence between 231.5 μm in an emmetropic eye (1/0.054 × 12.5 μm) and 297.6 μm (1/0.042 × 12.5 μm). In a vessel segment of 500 μm there are 10 to 20 endothelial cells, which were 25 to 50 μm long, 10 to 15 μm wide, and elongated in the direction of blood flow. This segment is covered by 5 to 20 smooth muscle cells 25 to 80 μm long.

^{ 30 }

^{ 31 }The RVA length resolution of 12.5 UM (equivalent to micrometers in an emmetropic eye) would theoretically be able to detect a contribution of a single endothelial and/or smooth muscle cell to the vessel diameter variability. In this case, however, an expected length of repetitive spatial sequences would correspond to the length of a single cell (minimum 25 to maximum 80 μm), which is 3 to 10 times shorter than observed in our data. A putative unit of vessel spatial variability (graphically presented in Fig. 5 ) seems to be longer than a single endothelial or smooth muscle cell in both arterioles and venules. One possible explanation is that endothelin-1, which seems to be produced more in vasospastic subjects,

^{ 32 }diffuses abluminally and simultaneously affects several smooth muscle cells. Another possible explanation is that extravascular factors, such as the retinal nerve fiber layer in which the vessels are embedded, the state of vitreous attachment to the vessels, or small retinal vessels that closely cross the measured vessel may have contributed to the spatial vessel diameter variability.

^{ 33 }

^{ 34 }

^{ 35 }

^{ 36 }An intraluminal–extraluminal pressure gradient is much lower in venules than in arterioles. Moreover, composition of a vessel wall is different in arterioles and venules, the former having a stronger smooth muscle layer.

^{ 30 }

^{ 31 }Still, this pulsation pattern was similar between arterioles and venules and significantly different between the vasospastic and control group. Endothelial cells and the disturbance thereof could be one explanation for the common behavior of arterioles and venules. Extravascular factors, as discussed earlier, could also be implicated. Elucidating a cause of different spatial and temporal behavior of retinal vessel diameter between vasospastic and nonvasospastic subjects requires further studies.

^{ 37 }and in the choroid,

^{ 20 }and only after a hemodynamic challenge does altered response become obvious. This may be clinically relevant, because systemic vascular dysregulation can contribute to ocular disease, such as in central serous chorioretinopathy,

^{ 6 }central vein thrombosis,

^{ 4 }nonarteritic anterior ischemic neuropathy,

^{ 5 }and glaucoma.

^{ 3 }Present data lend further support to previous observations that there is a strong association between peripheral and ocular circulation. Vascular dysregulation in the eye itself has been elusive and insufficiently characterized. A proper definition of vascular dysregulation in the eye would in the clinical setting translate to the possibility of its objective assessment and treatment. The results of the present study contribute to our understanding of the manifestations of vascular dysregulation in retinal vessels.

^{ 38 }so it is unlikely that a pulse amplitude (IOP pulsation) could have had any influence on the observed difference between the vasospastic and nonvasospastic subjects.

^{ 39 }

^{ 40 }

^{ 41 }However, none of our subjects reported classic symptoms of Raynaud (tricolor phenomenon).

**Figure 1.**

**Figure 1.**

Vessel Parameter | Vasospastic | Nonvasospastic | Two-Way ANOVA |
---|---|---|---|

Coefficient of variation arterioles (%) | 8.8 ± 2.8 | 6.1 ± 1.7 | Factor 1: P = 0.017 |

Coefficient of variation venules (%) | 3.8 ± 1.4 | 3.6 ± 0.9 | Factor 2: P < 0.0001 |

Interaction: P = 0.0097 | |||

Excursion amplitude arterioles (%) | 11.4 ± 5.9 | 9.3 ± 2.1 | Factor 1: P = 0.29 |

Excursion amplitude venules (%) | 6.1 ± 2.3 | 6.1 ± 1.7 | Factor 2: P = 0.0002 |

Interaction: P = 0.31 | |||

Mean/SD correlation coefficients arterioles | 0.25 ± 0.40 | 0.11 ± 0.23 | Factor 1: P = 0.038 |

Mean/SD correlation coefficients venules | 0.24 ± 0.22 | 0.09 ± 0.23 | Factor 2: P = 0.77 |

Interaction: P = 0.96 |

**Figure 2.**

**Figure 2.**

**Figure 3.**

**Figure 3.**

**Figure 4.**

**Figure 4.**

**Figure 5.**

**Figure 5.**