Does Keratoconus Follow Rundle’s Curve?

Abstract

Background: Rundle’s curve describes the natural progression of disease as gradually worsening before reaching a peak and stabilizing. This study aimed to investigate whether Rundle’s curve could be applied to keratoconus over a five-year follow-up period. Methods: Longitudinal study. Patients with keratoconus who underwent Pentacam tomography imaging from the Australian Study of Keratoconus were included in this study. Patients who received surgical treatment for keratoconus were excluded. Latent class analysis was performed for five parameters: Kmean front, Kmean back, pachymetry pupil, pachymetry minimum and pachymetry apex. A total of 522 patients and 1041 eyes were included for analysis. Most parameters were stable. However, worsening keratoconus in a minority of patients (less than 5% of the population) was observed across the last year of follow-up. The patients that showed progression in the final year were younger in age and had higher baseline parameters. Results: This study suggests keratoconus does not conform to the classic Rundle’s curve of disease progression. Instead, keratoconus exhibits a distinct course characterized by an increased risk of progression among younger individuals and eyes with higher baseline parameter values. Conclusions: These findings underscore the importance of considering treatments that halt disease progression, such as corneal collagen crosslinking, particularly in this specific subgroup of patients.

1. Introduction

Keratoconus is a progressive corneal ectasia with a complex and still not completely understood etiopathogenesis [1]. Both genetic factors, as evidenced by a high association with positive family history, and environmental factors, such as the habit of intense eye rubbing and nocturnal ocular compression, seem to play a role [1]. This condition apparently involves pro-inflammatory cytokines and proteases and is characterized by irregular corneal steepening which usually progresses from childhood to adolescence [2,3]. Ultimately, it may lead to irreversible vision loss due to high irregular astigmatism and sometimes corneal scars. In the last couple of decades, the management of keratoconus has transitioned from observation to intervention, focusing on halting disease progression and preserving sight. Procedures such as corneal collagen crosslinking (CXL), synthetic intrastromal ring segments [2,4] and more recently, the utilization of allogeneic lamellar corneal transplants (including the Bowman layer or stroma), corneal allogenic intrastromal ring segments, and Donut-shaped stromal tissue intrastromal transplantation, have become increasingly utilized [5,6,7,8]. In much more advanced cases, with corneal scars and extreme irregularity, more invasive corneal transplantation techniques (like deep anterior lamellar keratoplasty or penetrating keratoplasty) need to be performed [9]. CXL minimizes the need for corneal transplants; however, additional CXL may be required if keratoconus continues to progress [10]. Hence, understanding the natural course of progression for keratoconus, and when treatment should be provided is essential as it can maximize treatment outcomes by guiding treatment timelines according to the predicted course of the disease.

Unfortunately, the long-term natural history of keratoconus is not well described [11], and therefore it can be difficult to ascertain the average age group that certain interventions like CXL should be considered for patients. Various epidemiological studies suggest keratoconus advances at varying rates according to age [12,13]. Generally, the younger age of onset, the more rapid progression of disease with more adverse sight and vision changes [11], while older age patient onset tends to not progress as much and stabilizes [14]. Moreover, progressive deterioration of Kmax and parenchymal measurements have been observed in up to 20% of patients who underwent CXL, implying the choice or timing to undergo such procedures may be unsuitable for some [15,16].

Comparatively, other ocular conditions, such as Grave’s ophthalmopathy, have been shown to follow a relatively standard disease progression course over time, and therefore treatments can be better targeted at certain stages of disease. In Grave’s ophthalmopathy, this has been depicted by Rundle’s curve (Figure 1).

Rundle’s curve was first coined by F.F. Rundle in 1957, who used the phenomenon to describe the natural course of Grave’s ophthalmology [18]. It attempts to describe the natural history of disease through the severity of signs and symptoms which are said to rapidly worsen and reach a maximum, before falling and eventually stabilizing without complete resolution. This phenomenon has allowed for a consensus in the clinical management of Grave’s ophthalmopathy, where it has been agreed that immunosuppressants should target the active signs and symptoms phase, while surgery is held off until the curve stabilizes [11]. The curve has been validated across a range of Grave’s ophthalmopathy cohorts including Rundle’s own, but Rundle never specified that his observations were reserved solely for describing Grave’s ophthalmopathy itself [17].

To understand whether the natural course of keratoconus coincides with Rundle’s curve, the current study retrospectively followed a cohort of 522 participants from the Australian Study of Keratoconus, who had keratoconus but had not yet undergone any procedural treatment. This study hypothesized a characteristic disease course of keratoconus would be found and would resemble Rundle’s curve. If the progression of keratoconus can be mapped as per Rundle’s curve, then treatment for keratoconus, like CXL, can be timed most accurately along the disease time course, leading to the most impactive treatment outcomes, and fewer treatment failures.

2. Materials and Methods

Patients were recruited from Australian Study of Keratoconus with keratoconus. Keratoconus patients were required to complete a study questionnaire, together with clinical and physical examination, which included the anthropometric measurements of height and weight. Patients with keratoconus who underwent Pentacam software version 1.20r127 tomography imaging (Oculus, Wetzlar, Germany) and received at least three follow-up visits (and greater than 3 months between visits) across five years were included in this study. The duration of the disease before being recorded at the hospital was unknown. If the patient had undergone any procedural treatment for keratoconus, their data were included up to the time of surgery. Patients who did not have a recorded age and/or gender were excluded. In brief, keratoconus was diagnosed based on the presence of one or more of the following: an irregular cornea (as determined by distortion of keratometric mires and/or corneal topography/tomography imaging), scissoring of the retinoscopic reflex, and demonstration of at least one biomicroscopic sign, including Vogt’s striae, Fleischer’s ring, or corneal thinning and/or scarring typical of keratoconus. To assess the association of risk factors on disease, we obtained corneal curvature data from the four-map selectable display of the Pentacam tomography system (Oculus, Wetzlar, Germany). The study protocol was approved by the Royal Victorian Eye and Ear Hospital Human Research and Ethics Committee (Project # 10/954H). This protocol followed the Declaration of Helsinki and all privacy requirements were met. A patient information sheet, consent form, privacy statement and patient rights were provided to all individuals who wished to participate in the study.

Statistical Analysis

This was an observational retrospective study; therefore, no sample size was calculated. To enable generalizability of the results, we have included all keratoconus patients that were reviewed and had Pentacam measurements at corneal clinics from May 2018 to December 2020. To evaluate the natural progression of keratoconus and whether it followed Rundle’s curve, latent class analysis was performed separately for each ocular parameter of interest. The five parameters chosen are commonly used in clinical practice to track keratoconus progression: Kmean front (mean curvature at the corneal front surface), Kmean back (mean curvature at the corneal back surface), Pachymetry apex (thickness of the cornea at the apex), Pachymetry pupil (thickness of the cornea over the pupil), and Pachymetry minimum (thickness of the cornea at its thinnest point). Latent class analysis is a form of cluster analysis which was used to group keratoconus patients into 5 groups that have similar values of each parameter over the 5-year period. These groups are called ‘classes’. As each ocular parameter was evaluated independently, classes across ocular parameters are independent of each other (for example, participants in class 1 for Kmean front are not necessarily the same participants as those in class 1 for Kmean back). Models were estimated for up to 5 latent classes and all parameters were included as continuous variables. The best model for each class was chosen based on Akaike and Bayesian information criteria (AIC and BIC). Models were compared, and the model with lowest AIC and BIC was chosen as the model of best fit. For all parameters, best model included either 4 or 3 classes. Summary of the distribution of the outcomes among classes is presented graphically. Baseline characteristics among classes were compared using chi-square test and Kruskal–Wallis test with Dunn’s post hoc pairwise comparisons (using Benjamini-Hochberg procedure to adjust for family wise error rate). All analyses were performed using Stata 16.1.

3. Results

3.1. Main Results

A total of 904 keratoconus patients were initially identified from Australian Study of Keratoconus. A total of 382 patients were excluded in accordance with the inclusion and exclusion criteria described earlier. Therefore, a total of 1041 eyes from 522 patients with keratoconus were included in this study. The median age was 26.1 years with an interquartile range (IQR) of 20.2, 34.1 years. Males comprised 61.9% of the cohort. Cohort characteristics are listed in

Table 1. Cohort characteristics.

Descriptions	Values
N	522
Age at baseline, median (IQR)	26.1 (20.2, 34.1)
Age categories, total (percentage of N)
- Children (<18 years)	84 (16.1%)
- Young adults (19–30 years)	251 (48.1%)
- Middle age (31–50 years)	146 (28.0%)
- Seniors (>50 years)	41 (7.9%)
Gender, total (percentage of N)
- Female	199 (38.1%)
- Male	323 (61.9%)

. The cohort underwent a median follow up time of 3.1 years (IQR 1.7, 42 years), with median time between visits being 270 days (IQR 175, 449 days).

For the Kmean front, 4 groups were identified (Figure 2). The Kmean front value (in dioptres) was calculated for the anterior corneal surface using the keratometric index (1.3375) for the anterior corneal surface. This value represents the average curvature of the cornea and is expressed in diopters. The Pentacam software uses various measurements, including the anterior and posterior corneal curvatures, to calculate this value. The majority of participants (class A3 and A4, total 84% of participants) followed a stable trajectory over the 5-year follow up. A small proportion (class A2, 11.8% of participants) showed stable values at a Kmean front of 56 dioptres at baseline and remained similar for 4 years but showed some worsening in the fifth year of follow up (Kmean increase to 58 dioptres). Similarly, class A1 (4.2% of participants) had a stable Kmean front for 4 years at baseline with a value of 70 dioptres but worsened in the last year up to a Kmean of 75 dioptres.

The Kmean back had similar outcomes. The Kmean back was calculated using the keratometric index (1.3375) for the posterior corneal surface. This value represents the average curvature of the posterior corneal surface and is expressed in diopters. Similar to the K mean front value, the Pentacam software uses measurements of the posterior corneal curvature to calculate this value. The majority of participants (class B3 and B4, total 83.7% of participants) had stable values overtime. Those with worse Kmean back baseline values (class B1, 4.0% of participants; and class B2, 12.2% of participants) had worsening trajectory particularly in the last year of follow up (class B2 stable for 4 years at ap-proximately −8.7 dioptres K mean back, worsened to −9.2 dioptres; and B1 was more erratic throughout the 5 year follow up, but worsened particularly in year 5 from −10.9 dioptres to −11.8 dioptres).

The majority of Pachymetry apex, pupil and minimum parameter classes showed stable values over the follow up period. In Pachymetry apex classes, a large proportion of participants (class C2, total 51.3% of participants) had stable values compared to baseline (remained stable at approximately 450). Class C3 (total 37.7%) had some minor change in value slowly over time, with a baseline value of 419, which progressed to 527 in the final year. Progression was more notable in classes C1, where 8.7% of total participants had baseline values of 376, and over the five-year follow up, this valued decreased to 340. Class C4 showed erratic changes over time, with 2.3% of total participants having Pachymetry apex values that were 579 at baseline, and 648 at the fifth year of follow up.

Pachymetry pupil classes had similar outcomes. Class E2 and E3, 86.9% of total participants were largely stable, ranging from 469 to 465, and 525 to 535, respectively. Class E1 (11.0% of total participants) showed a larger change across the five years, with a baseline value of 403 to 381 in year 5. Class E4 (2.2% of total participants) was erratic but showed an overall increase in values from 576 at baseline to 647 in the last year of follow up.

Latent class analysis for Pachymetry minimum identified 3 classes rather than 4. Class D2 and D3 (65.8% of total participants) were largely stable and ranged from baseline value of 442 to 442, and 509 to 512, respectively. Class D1, which represents 10.5% of total participants, had a steep decrease in Pachymetry minimum, from baseline of 356 to 317 in the last year of follow up.

The overall analyzed progression for each parameter did not follow Rundle’s curve.

3.2. Overlap with Classes

Table 2. Overlap of the 5 classes across all five parameters. Classes are groups of participants that have similar values of each parameter over time. Classes are specific to each parameter and are independent of classes in other parameters.

Kmean Front Class	Kmean Back Class	Pachymetry Apex Class	Pachymetry Minimum Class	Pachymetry Pupil Class	N (%) of Participants
3	4	3	3	3	228 (22%)
3	4	2	2	2	172 (17%)
4	3	2	2	2	172 (17%)
2	2	2	2	2	57 (6%)
3	4	3	3	2	28 (3%)
4	3	3	3	3	27 (3%)
3	4	3	2	2	24 (2%)
1	1	1	1	1	24 (2%)
3	3	2	2	2	23 (2%)
4	4	2	2	2	23 (2%)

shows the overlap of the classes across all five parameters. Most participants (76%) show no progression in any parameter (belonging to class A3/A4, B3/B4, C2/C3, D2/D3 and E2/E3). A few participants (6%) showed no progression in pachymetry parameters (C2/C3, D2/D3 and E2/E3), but worsening in Kmean front and Kmean back in the last year (A2 and B2). A total of 2% of participants that had more severe baseline values showed worsening in all parameters (A1, B1, C1, D1 and E1). A total of 1.6% of participants showed stable Kmean front and Kmean back (A3/A4, B3/B4), but worsening of pachymetry values (C1, D1 and E1). Another 1.6% of participants showed stable Kmean front and Kmean back (A3/A4, B3/B4), but variable pachymetry values (C4, D2 and E4). All other participants had a variable mix of classes.

3.3. Comparison of Baseline Characteristics Between Classes for Each Parameter

For the parameter Kmean front, patients that show worsening in the last year were younger in age (24 years old versus 26 or 27 years old), although this was not significant (p = 0.15). There were no associations found with gender. All ocular parameters are predictive of worsening (those with worse parameters at baseline are more likely to progress).

In analyzing Kmean back, it was found that patients that showed worsening in the last year were younger compared to those who remained stable or worsened throughout (23 years versus 27 years versus 26 years, respectively), p = 0.031. There were no associations found with gender. All ocular parameters at baseline were predictive of class (those with least severe parameters showed no progression, while those with most severe parameters showed progression throughout).

Pachymetry apex and pupil parameters shared a common outcome, where older patients had variable pachymetry apex and pupil values over time (49 years for class C4 or E4 compared to 25/26 years for C1–C3 or E1–E3), p < 0.001. There were similar age groups in those that progressed compared to those that remained stable. There were worse other ocular parameters in those that progressed in these two parameters.

Pachymetry minimum analysis showed no association with age or gender. Those with higher baseline ocular parameters are the ones that progress.

4. Discussion

This study investigated the five-year natural course of keratoconus to ascertain if it adheres to a distinct trajectory akin to Rundle’s curve for Grave’s ophthalmopathy. Our analysis predominantly revealed a stable disease course among most keratoconus patients. Notably, we identified Kmean front, Kmean back, and pachymetry minimum as potential valuable proxies for predicting disease progression, particularly evidenced by worsening keratoconus in the final year of follow-up.

Furthermore, our findings underscore the significance of age-related variables, contributing valuable insights to the existing body of evidence concerning their influence on keratoconus severity and disease progression. Notably, no gender-related associations were observed. These novel insights hold promising clinical implications, offering clinicians the potential to better predict the trajectory of keratoconus and make informed decisions regarding patient management strategies.

A major finding of this study was the generally stable natural history of keratoconus across time. This was a characteristic trend observed in most classes across every ocular parameter. Most patients were oscillating around baseline measurements for the 5-year follow-up period. It was observed that the few that did have progressive disease (particularly in Kmean front, Kmean back), had higher baseline ocular parameters and were younger in age. In some pachymetry parameters (apex and pupil), older age groups had erratic and variable progression before stabilizing and returning to baseline. This was an interesting result, as with increasing age, there is an expected biomechanical change in the cornea where stiffening and therefore reduced thinning occurs [19]. Although the IQR of this study was up to 34 years of age, where such biomechanical changes may not be significant [19], it will be fruitful to identify in future studies whether a link between older age and worsening pachymetry values exists.

While keratoconus can be stable, for it to make up such a large proportion in this cohort is surprising. In a meta-analysis of 11,000 eyes across 41 studies, Ferdi et al. concluded Kmax significantly increased by 0.7 D per 12 months although the study had considerable heterogeneity [20]. Indeed, inconsistent findings have led to the perception that keratoconus randomly deteriorates [21]; however, the current study of 1041 eyes suggested keratoconus only progressed in a small proportion of cases across five years, and consistent monitoring of topography and pachymetry provides robust information about the outlook of disease course.

While stability was found in most classes across parameters, a small proportion of groups tended to behave erratically and worsen. An important consideration is our analysis only stratified for age and gender, while the literature has for a long time suggested atopy, eye rubbing, and environmental/genetics are independent factors which increase the odds of keratoconus progression and severity [22,23,24]. Although unavailable, the stratification by these factors in our analysis would have clearly shown whether they could explain these rapidly and erratically deteriorating populations. It is our opinion that such longitudinal data would assist in the construction of a prediction score, and future studies should seek to investigate these factors to incorporate their involvement for making a clinically useful risk calculator.

It cannot be overlooked that keratoconus disease history in older ages were protected from deteriorating within five years and most unexpectedly, was associated with significant improvement in some ocular parameters. This was especially noticeable within pachymetry apex and pupil parameters. These results are supported by numerous other studies which report minimal active disease in older age [25,26], a lack of keratoconus onset in older populations [25], and low spontaneous progression in older people [27,28]. This may be attributed to biochemical changes in the cornea related to age as described earlier, which alters collagen matrix deposition, resulting in a stiffer phenotype [19]. Such changes have an exponential relationship with age and are protective for thinning and wear of the cornea, which ultimately protect against keratoconus worsening. That being said, the higher IQR range of this study was only 34.1, at which age a stiffening effect may be negligible [19].

While these results suggest older patients with keratoconus are generally better off clinically, it does not provide statistical evidence that younger ages in our included age range were the contrary. Currently, literature on the impact of younger ages on keratoconus progression is mixed; however, ages between cohort studies vary vastly and the mean age of our cohort cannot account for studies with extremely young cohorts. For example, while Ferdi et al. found that ages <17 are likely to experience a 0.7D increase in Kmax per 12 months [20], in this study, ages rarely dipped below 18 and our mean age was 26.1. While it is clear pediatric keratoconus deteriorates suddenly, the exclusion of such populations likely biases the effect of younger age from being a significant factor for worsening keratoconus in this analysis. It does, however, infer most younger aged people in this study are not prone to spontaneous deterioration in topographic measurements, which is in support of other studied non-pediatric keratoconus in young cohorts. Considering this study’s mean age of 26.1, the minority that undergo rapid corneal changes are likely to be explained by other factors that are associated with keratoconus but not investigated by the current study such as eye rubbing, genetic predisposition, and environmental influences.

As the results of this large-scale study have suggested keratoconus is largely stable across long term follow-up, it does present some discussion about Rundle’s curve and the management of keratoconus. Firstly, the trends we observed were not aligned with Rundle’s curve because when latent class analysis determined separate classes across every ocular parameter, a majority were not maximizing to one point; rather the disease history oscillated around baseline. Most classes exhibited stable disease throughout the five-year follow up, and any deterioration that was detected in the last year were related to baseline ocular parameters described above, and younger age. However, the only classes that did exhibit similarity to Rundle’s curve were pachymetry apex and pupil parameters, and still their disease trajectories were worsening upon five years follow up rather than stabilizing. This could be an artifact of follow-up duration, which might need to be longer to establish a Rundle’s curve in keratoconus. These results are intriguing because currently clinicians prescribe CXL surgery early in the disease course as preventative for Kmean deterioration. Because prescription of CXL is on the basis that Kmean will eventually deteriorate, this study presents the point that for most eyes across every severity, course of disease is unlikely to sporadically deteriorate. As an expensive and invasive surgery, the benefits of CXL need to be weighed alongside the likelihood of keratoconus progression without such interventions; however, more studies of a similar study design are needed for this information to be clinically literate.

Limitations of the Study

Despite the novel outcomes of this study some limitations should be acknowledged. Firstly, the relatively smaller cohorts of moderate/severe disease compared with mild disease led to not being able to analyze results according to severity, for a high-powered analysis. This prevented them from being separately analyzed and understood, and it is possible the merging of disease courses would impact the generalizability of these results. Secondly, only age and gender were considered in our analysis. To deepen our understanding of keratoconus progression, future studies should consider other factors such as eye rubbing, genetic predisposition, and environmental influences, known to influence keratoconus progression. Thirdly, this study excluded eyes that underwent procedural treatments for keratoconus which could infer some selection bias towards stable disease which did not warrant intervention and may question the representativeness of this cohort. While this is a valid concern, the inclusion of all severity groups inferred populations severely affected by keratoconus were assessed equally as those with mild disease. Fourthly, the association between keratoconus and Rundle’s curve cannot be excluded, because this study only followed participants for five years and actual disease history might be much longer than this time.

Overall, our findings suggest that a longer-term follow-up study could provide more definitive insights into the natural history of keratoconus and the optimal timing of interventions like CXL. This could potentially lead to a re-evaluation of current clinical guidelines and the development of more personalized treatment approaches based on individual risk factors and disease progression patterns. Additionally, understanding the natural history of keratoconus and the factors influencing disease progression could also contribute to the development of novel therapies aimed at slowing or halting the progression of the disease, ultimately improving the long-term outcomes for patients. While our study has limitations due to its relatively short follow-up period and exclusion criteria, we believe that it lays the groundwork for future research to explore the potential benefits of early intervention in keratoconus and the development of more targeted treatment strategies.

5. Conclusions

In conclusion, this pioneering study represents the first comprehensive examination of keratoconus over a five-year duration, shedding new light on its natural history. Our analysis revealed that keratoconus does not conform to the classical Rundle’s curve of progression during this time frame. Instead, it exhibits a distinct pattern characterized by an increased risk of progression among younger individuals and those with higher baseline parameters. These findings not only challenge existing paradigms, but also raise intriguing questions about the potential evolution of keratoconus over more extended follow-up periods.

The study underscores the need for further investigation to determine if keratoconus might eventually align with Rundle’s curve, particularly as it pertains to the timing of crucial interventions such as CXL. Such investigations should involve larger cohorts and extended follow-up durations to fortify these findings and potentially refine the timing of treatments like CXL, especially for younger age groups and individuals with elevated baseline values.