In vivo assessment of the SNP morphology using CCM has emerged as a valuable clinical modality to improve our understanding of the relationship between this rich nerve plexus and various ocular and systemic conditions and diseases. As reviewed in more detail elsewhere,
35,36 morphometric evaluation of the SNP has been used to diagnose, assess, and follow-up ocular surface conditions, including ocular allergy, dry eye, infectious keratitis, and glaucoma and after keratorefractive surgery and contact lens wear. Currently, considerable evidence exists that advocates the utility of CCM for assessment of small nerve fiber pathology induced by systemic and neurological conditions, in particular DPN. This study examined the longitudinal aspect of the utility of CCM to serve as an acceptable measure of DPN in clinical research and practice.
We report data from a cohort of individuals with type 1 diabetes (
n = 147) and healthy controls (
n = 60) collected from baseline to a median duration of 3.7 years. Although the stability of corneal nerve morphology has been previously demonstrated in a 3-year longitudinal study in healthy individuals,
37 to our knowledge no previous study has examined the dynamic natural course of SNP microstructures in relation to DPN. With reference to the lack of previous investigation concerning the natural history of corneal nerves in diabetes, the present study is a positive response and attempts to fill this research gap.
At the baseline visit, age was matched between participants with diabetes and controls. Diabetic participants showed moderate glycemic control and excellent control of cardiovascular risk factors, including the BP and lipid profile in accordance with the current treatment guidelines.
38 The lower level of total cholesterol and LDL cholesterol in our diabetic patients as compared with controls is attributed to the fact that 35% were receiving lipid-lowering medications.
Comparison of the clinical parameters at baseline and final visit showed that there were no clinically significant changes to health or metabolic and anthropometric measurements, indicating stable glucose control and desirable maintenance of cardiovascular risk factors. Although the Hawthorne effect
39 may have been involved, the finding of lower alcohol consumption in the diabetic patients at baseline, which is maintained at follow-up, reflects good diabetes education. And the significant reduction in tobacco consumption over time in both diabetic patients and control subjects presumably reflects overall population level of education to stop smoking.
Except for peroneal nerve conduction velocity, with a statistically significant but clinically trivial decline (−0.9 m/s), the remaining established measures of neuropathy remained unchanged or improved slightly from baseline to the final visit. However, LMM5 showed that changes in peroneal nerve conduction velocity in DPN+ve and DPN–ve patients did not differ significantly from controls, indicating a similar effect of time for groups. The low rate of change over time in these measures may be attributed to the following: the maintenance of a healthy lifestyle and compliance with medical advice among our diabetic cohort, the inclusion of participants with only mild neuropathy, and the relatively short duration of study. Negligible worsening or no progression of the traditional measures of DPN also has been observed in the placebo arm of a recent interventional study
40 of 227 patients with predominantly type 2 diabetes, but with substantially worse glycemic control at baseline (8.8% ± 1.9%) and a reduction of 0.67% ± 1.41% over 4 years. Our findings are further supported by a 3-year longitudinal study of 62 subjects with predominantly type 2 diabetes and good glycemic control (HbA
1c 7.23% ± 1.03%), which interestingly demonstrated stability in a range of neurological examinations, symptom scores, autonomic testing, QST, and NCS with worsening only in the sural nerve amplitude and the axon-reflex vasodilation test, a measure of small fiber neuropathy.
41
All three SNP parameters were significantly reduced in diabetic participants without and with neuropathy at the baseline visit. This finding is consistent with other studies, which also show a depletion of SNP tissue in diabetic patients without and with DPN, reflecting early subclinical small fiber damage.
22–24,42 Based on the reported association of SNP parameters and DPN severity,
23,24 we hypothesized that participants with diabetes and DPN would demonstrate quicker deterioration of SNP tissue than those without DPN. To examine this hypothesis, we built several LMMs. Such models afford robust methods of analyzing longitudinal data with repeated measurements, in particular when the data are incomplete or unbalanced due to missing data, dropouts, or differences in observation time points.
34
According to the three basic mixed models developed here and regardless of group, there was no significant effect of time for any of the three SNP parameters. A group*time interaction term was not significant for CNBD or CNFL (P = 0.24 and P = 0.20), indicating that the presence or absence of DPN at baseline did not appear to impact CNBD and CNFL changes over time. Mean CNBD (23.7 ± 20.9 vs. 22.7 ± 16.9, no./mm2) and CNFL (15.0 ± 4.3 vs. 14.4 ± 4.1 mm/mm2) declined slightly over 4 years in the neuropathy group, but to an extent that is neither clinically nor statistically significant.
However, the Type III F-test for the interaction between time and group was statistically significant for CNFD (
P = 0.02), suggesting that the relationship of time with CNFD change varies depending on the group. The LMM4 (
Table 4) demonstrated that whereas CNFD trajectories were not statistically different between DPN–ve and controls, the mean CNFD decreased significantly in the DPN+ve group during follow-up, with a loss of approximately 1 nerve/mm
2 per year. This observed CNFD change was best predicted by participant age and duration of diabetes (both
P < 0.05). One may anticipate that such a change would be influenced by glycemic control; however, HbA
1c did not reach statistical significance (
P = 0.10) in LMM4, where CNFD was considered as a dependent variable, possibly because of the relative stability of this factor during the study period. Although the outcome of CNFD decline indicates dynamic structural small nerve fiber damage at the SNP, the relevance of CNFD change in the neuropathy group and the relative stability of CNBD and CNFL are not clear. Disparate changes to these three corneal nerve parameters also have been reported in diabetic individuals after improvement in risk factors for DPN
27 and after simultaneous pancreas and kidney transplantation,
43 suggesting a complex, dynamic, and perhaps nonlinear relationship between these parameters.
The baseline cross-sectional findings in the present study confirmed that all three SNP parameters were reduced in the neuropathy group compared with controls. The parameter that underwent the most marked reduction over time was CNFD. This suggests that branch damage (thinner branches emanating from major nerves) might represent the primary pathological change in DPN, whereas CNFD (a parameter related to the major nerve trunks) deterioration occurs later. The reduction in CNFD along with a nonsignificant decline of the other two parameters also may suggest degeneration of major nerve trunks with concomitant regeneration reflected by an increase in the CNBD and CNFL. Therefore, it is conceivable that loss and indeed repair of different SNP parameters may occur at different stages of the disease.
Limited studies are available documenting the link between corneal small nerve fiber change and risk factors of DPN.
21,27,44 In the present study, when the data were restricted to include only diabetic individuals and on removal of the effect of group in the LMMs, we found that every 1-unit increase of HbA
1c was associated with a decrease of approximately 0.6 nerve/mm
2 in CNFD. There also was a negative effect of diabetes duration on CNFD and CNBD. Each 10-year increase of diabetes duration at baseline resulted in 0.8 and 2.0 nerve/mm
2 decline of central corneal CNFD and CNBD, respectively. Corneal nerve branch density was also significantly affected by smoking. Increasing one cigarette per day had a negative effect of 0.25 nerve/mm
2.
These results demonstrate the link between risk factors of DPN and morphologic parameters of corneal nerves. We have no clear explanation why HbA
1c has an effect on CNFD, but not CNBD and CNFL. Nevertheless, this finding is consistent with the study of Tavakoli et al.
27 who reported a significant correlation between changes in HbA
1c and CNFD (
r = −0.52,
P < 0.01) but not for CNBD and CNFL. In a study of 38 type 1 diabetic patients with and
without neuropathy, Ishibashi et al.
44 reported time-dependent effects of HbA
1c on SNP parameters. Although nerve beading frequency was positively correlated with the mean HbA
1c levels at time of, or up to 3 months before CCM examination, no significant association was found between CNFD and CNFL with HbA
1c up to 6 years before CCM examination.
These findings emphasize the importance of including different SNP parameters in future studies, where these parameters are to be used as measures of small nerve fiber damage and in particular repair. Additionally, in this study, only the central cornea has been investigated. Recent studies have revealed that loss of corneal nerve structure in the SNP mainly occurred at the inferior whorl, which is slightly more distal than the central cornea and may therefore be expected to show more marked pathology.
45,46 Further longitudinal work assessing the inferior whorl as opposed to the central cornea may provide additional insights and ability to discriminate change in relation to DPN.
In previous cross-sectional studies, SNP parameters have been shown to correlate with functional and structural measures of neuropathy.
19,23,25 Quattrini et al.
19 reported a significant correlation between CNFD versus NDS (
r = −0.30,
P = 0.03) and cold sensation threshold (
r = −0.40,
P < 0.01). In a subsequent study, moderate correlations were found between NDS and corneal nerve parameters (
r, −0.48 to −0.58;
P < 0.001).
23 In a recent study by Sivaskandarajah et al.,
25 CNFD, CNBD, and CNFL were related to cold sensation threshold (
r, 0.32 to 0.37;
P ≤ 0.01). In this longitudinal study, we examined the relationship of change in corneal nerve parameter with conventional measures of neuropathy by calculating the absolute change from baseline to final visit for participants with diabetes. We found a modest correlation between CNBD and peroneal conduction velocity (Pearson
r = 0.23,
P = 0.02). When the data were restricted to the DPN+ve group, this correlation increased to 0.38. Furthermore, CNFL also correlated to cold sensation threshold (
r = 0.40,
P = 0.03), which indicates that SNP parameters do change in a fashion comparable with some traditional measures of neuropathy.
The key strengths of this study are its longitudinal nature, inclusion of a range of traditional neuropathy measures (small and large nerve fiber dysfunction) in a relatively large number of type 1 diabetic participants, the consistency and strict adherence to technical and methodological procedures, such as capturing and selection criteria of the SNP images, and using a fully automated image analysis algorithm, which is essential to eliminate the variability associated with manual and semiautomated analysis. Thus, we used a fully automated image analysis algorithm that has been validated and compared against the manual and semiautomated analysis
33,42,47 in individuals with diabetes.
A limitation of this study is that most type 1 participants were enrolled from specialized clinics, where the glycemic and cardiovascular factors were optimally controlled, which may not represent the typical population with type 1 diabetes. Additionally, 4 years of study might be insufficient to discern changes over time, particularly in the case of patients with mild neuropathy or the limited number of apparently motivated participants with well-controlled diabetes available in the neuropathy group.
In conclusion, the findings presented herein provide evidence that CCM has the potential to track the structural alterations of the small nerve fibers in DPN. Furthermore, these findings support the notion that quantification of the SNP morphology has a substantial potential to be used as an appropriate adjunct measure to conventional measures of DPN.