Predictive biomarkers of intra-ocular pressure decrease after cataract surgery associated with trabecular washout in patients with pseudo exfoliative glaucoma

To investigate biomarkers of intra-ocular pressure (IOP) decrease after cataract surgery with trabecular washout in pseudo-exfoliative (PEX) glaucoma. A single-center observational prospective study in PEX glaucoma patients undergoing cataract surgery with trabecular washout (Goniowash) was performed from 2018 to 2021. Age, gender, visual acuity, IOP, endothelial cell count, central corneal thickness, medications, were collected over 16-month follow-up. Multivariable binomial regression models were implemented. 54 eyes (35 subjects) were included. Mean preoperative IOP (IOPBL) was 15.9 ± 3.5 mmHg. Postoperative IOP reduction was significant at 1-month and throughout follow-up (p < 0.01, respectively). IOPBL was a predictive biomarker inversely correlated to IOP decrease throughout follow-up (p < 0.001). At 1 and 12 months of follow-up, IOP decrease concerned 31 (57.4%) and 34 (63.0%) eyes with an average IOP decrease of 17.5% (from 17.6 ± 3.1 to 14.3 ± 2.2 mmHg) and 23.0% (from 17.7 ± 2.8 to 13.5 ± 2.6 mmHg), respectively. Performance (AUC) of IOPBL was 0.85 and 0.94 (p < 0.0001, respectively), with IOPBL threshold ≥ 15 mmHg for 82.1% and 96.8% sensitivity, 84.2% and 75.0% specificity, 1.84 and 3.91 IOP decrease odds-ratio, respectively. All PEX glaucoma patients with IOPBL greater than or equal to the average general population IOP were likely to achieve a significant sustainable postoperative IOP decrease.


Study population
The study population of our study was extracted from a prospective large-scale glaucoma register (NCT number: NCT04381611) which includes any patient suffering or suspected of glaucoma and does not affect their care, treatment choice or follow-up at Montchoisi Clinic, Lausanne, Switzerland.

Inclusion criteria
All patients with PEX glaucoma who were scheduled for routine cataract surgery associated with trabecular washout using the Goniowash technique (Fabrinal, Switzerland) 7 ; Patients who were able and willing to provide informed written or verbal consent.

Exclusion criteria
Patients who were unable to understand the implications of their inclusion in the study or who were unable or unwilling to give informed consent; missing data, loss to follow-up.

Statistical analysis
Continuous variables such as age, IOP, IOP decrease from baseline, EC, CCT, OHM number were expressed as means and standard deviations (± SD).BCVA values were recorded using the Snellen scale, and Snellen values were converted to logMAR units for statistical analyzes.LogMAR values corresponding to count fingers, hand motion, light perception, or no light perception were substituted with 2.10, 2.40, 2.70 and 3.00 logMAR, respectively 9 .Categorical variables such as gender, eye, IOP decrease from baseline (yes/no) were expressed as frequencies and percentages (%).Calculation using the Z statistic for sample size for paired tests required a sample size of 44 subjects to achieve a power of 90% (β = 10%), a level of two-sided significance α of 5%, for detecting an effect size of 0.5 (maximum variability) between pairs.After adding 10% of the calculated sample size to compensate any study-related problems, the final sample size of the study was 49.Data analyses were performed on a full analysis set using multivariate regression imputation of missing data.The IOP, BCVA and OHM variables were analyzed on day D1, week W1, and months M1, 3, 6, 9, 12 and 16 of the follow up, and compared to those of preoperative baseline ( BL ).Wilcoxon's matched pairs signed ranks test was used for pre-and post-operative comparisons of IOP, BCVA, and OHM at each relevant time of follow-up.
Multivariable binomial logistic regression analyses with mixed stepwise selection, adjusted for central corneal thickness covariate, were carried out using a logit function, Ln( p 1−p ) = logit (p), to identify predictive independent biomarkers of the decrease in IOP from the baseline (yes/no).The logit of the probability (p) of the occurrence of the decrease in IOP (Y) is expressed as a function of an intercept β0, of the biomarkers (Xi) attached to their regression coefficients βi and to a noise term ε.It consists of modeling the probability that a decrease in IOP will occur, as follows: Ln( p ; OR(X i ): the unit odds ratio of the biomarkers (Xi).
During the forward step of mixed variable selection, the maximal p-value was 0.25 for an effect to enter the model.During the backward step the minimal p-value was 0.10 for an effect to be removed from the model.
The adequacy of the model to predict the decrease in IOP from baseline was assessed by the calculation of R-square metric (R 2 ), which represents the percentage of variance of IOP decrease that can be explained by the biomarkers (Xi) included in the model.Prediction performances of biomarkers were quantified by the area (AUC) under the ROC (receiver operating characteristic) curve.Decision cut-offs for prediction were calculated according to the Youden maximal index, J = maxc [Se(c) + Sp(c)-1], with c: all possible biomarker cut-off values; Sp: specificity; and Se: sensitivity.
For the biomarker Xi, the covariables being constant, the unit odds ratio was: Lack-of-fit tests for regression models were carried out to assess whether the variables selected in the models provided sufficient information to predict IOP decrease.Likelihood Ratio Chi-Squared Tests were carried out to assess the level of significance of the effect of the whole model to predict IOP decrease.Effect Likelihood Ratio Tests were carried out to assess each effect's significant contribution to the model fit to predict IOP decrease.Data analyses were performed on a full analysis set after multivariate regression imputation of missing data.(SAS 9.1, SAS Institute Inc., Cary, NC).

Ethical approval
The study involving human participants was approved by the Swissethics (Cantonal Ethics Commission for Human Research of the Canton of Vaud (CER-VD) in Switzerland #2018-01403).The study was carried out according to the declaration of Helsinki guidelines.

Results
54 eyes of 35 PEX glaucoma patients who underwent routine cataract surgery were consecutively included in the full analysis set based on the inclusion and exclusion criteria.All these patients underwent cataract surgery associated with trabecular washout, and received an anhydrase inhibitor (acetazolamide 250 mg, single or double tablet) approximately 3 and 6 h after surgery to prevent postoperative IOP spikes.Washout was performed using the Goniowash technique in all cases.5 eyes of 5 PEX glaucoma patients had some missing data, and no patients were lost to follow-up.The demographic synopsis of the study subjects (male: n = 17; female: n = 18; overall mean age: 76.8 ± 6.9 years old) is summarized in Table 1.

Intraocular pressure
Mean preoperative IOP (IOP BL ) was 15.9 ± 3.5 mmHg.No significant change in IOP was observed at 1 day and 1 week postoperatively.Early transient IOP spikes of 30 mmHg or greater within the first 24 h following cataract surgery was observed in 4 out of 54 eyes (7.4%).

Best-corrected visual acuity
Improvement in BCVA was significant at 1 week postoperatively and throughout follow-up, (p < 0.0001, at all visits).Table 1, Fig. 1.Average BCVA improved from 0.5 ± 0.3 to 0.8 ± 0.3 at 1-week follow-up and remained stable throughout follow-up.

Mean number of OHMs
The mean number of OHMs decreased throughout the follow-up, significantly at 1-day and up to 6-months of follow-up (p = < 0.0001, respectively) then with a trend towards significance at M9 and M12 of the follow up, (p = 0.0822 and 0.0586, respectively) Table 1.

Predictive biomarkers of decrease in IOP
To predict the decrease in IOP from the baseline during the follow-up, multivariable binomial logistic regression models, adjusted for the CCT covariate, tested the effects of the variables age, sex, eye, EC, IOP BL , and variations in number of OHM at each relevant time of follow-up.IOP BL was a predictive biomarker inversely correlated to the decrease in IOP on day D1, week W1, and month M1, 3, 6, 9, 12 and 16 of the follow up, (p < 0.05, respectively).Table 2. Age, sex, eye, endothelial cell count, changes in OHM from baseline were not independent predictive factors of the IOP decrease during follow-up.AUCs obtained from ROCs, optimal decision thresholds and unit odds ratios for the preoperative classifier IOP BL to predict the decrease in IOP (yes vs no) for each relevant time of follow-up are summarized in Table 2.
According to the cut-off value of 15 mmHg of the predictive biomarker IOP BL , the decrease in IOP and improvement of BCVA were significant at 1-day and 1-week follow-up, (p = 0.0187 and 0.0056) respectively, then throughout follow-up (p < 0.001 and < 0.05, respectively) Table 1, Fig. 3.
At 1-and 12-months follow-up, conclusions for AUC of the IOP BL and IOP BL cut-off value were relatively similar in both the full analysis set with missing data imputation and the set of complete-case analysis in which AUC was 0.85 and 0.94 (p < 0.0001, respectively), with an IOP BL cut-off value ≥ 15 mmHg, for 82.1% and 93.3% sensitivity, 83.3% and 78.6% specificity, and an odds ratio of IOP decrease of 1.80 and 3.72, respectively.

Discussion
The identification of predictive biomarkers is helpful to maximize patients' chances of reaching the expected outcome of the treatment predicted by the biomarkers.Predictive models of multivariable binomial stepwise logistic regression were implemented to establish the patients' profile with PEX glaucoma likely to achieve the best expected outcome, in terms of decrease in IOP, after cataract surgery.Logistic regression consists in modeling the  probabilities of having a decrease in IOP using logit functions.The stepwise selection process was used to select a subset of effects for a regression model, knowing that there is little insight in the literature to guide the selection of terms of a model for the prediction of the decrease in IOP.This process allows to interactively explore which predictors seem to provide a good fit for the model.The control of effects entering and leaving the models was ensured by a mixed direction of stepwise selection that alternates the forward and backward steps.The stepwise model incorporates the most significant term relative to the maximum p-value threshold and removes the least significant term relative to the minimum p-value threshold.The term removing process carries on until the next terms are significant to change to the forward direction.The multivariate regression models were adjusted for the central corneal thickness covariate, a confounding variable affecting the accuracy of applanation-measured IOP 10 .The modeling results showed that preoperative IOP (IOP BL ) reliably and sustainably predicted the decrease in postoperative IOP with a high predictive performance, sensitivity and specificity.
Model convergence was reached.The whole model effect in predicting the decrease in IOP was significant (p < 0.0001) and effective (R-square = 0.612), which demonstrated that the model significantly explained the variance of the IOP decrease, and reflected the variables of age, sex, eye, endothelial cell count, baseline IOP and medications.The absence of statistical significance (p = 0.617) of the Chi-Squared Test of Lack-of-fit for the whole regression model signified that there was no lack-of-fit of the model.Therefore, the variables selected in the models provided sufficient information to predict IOP decrease.Introducing supplementary terms into the model, such as using polynomials or cross terms, would not significantly improve the model fit.This was consistent with the significant contribution of the effect of preoperative IOP (IOP BL ) to the model fit (p = 0.0016).This demonstrated that the parameter IOP BL contributed significantly and usefully to improve the model's performance of predicting the decrease in IOP.
The regressor for IOP BL was positive which indicated that the decrease in IOP increased when the value of IOP BL increased by 1 unit.At 1-and 12-months follow-up, the values of 1.84 and 3.91 of unit odds ratios of IOP decrease from baseline indicated that the probabilities of IOP reduction were 1.84 and 3.91 times greater, respectively, for an increment of + 1 mm in IOP BL .
ROC curve area (AUC), or accuracy index, is a metric of performance to quantify the IOP BL classifier's accuracy of discrimination between two patients' postoperative states, namely "IOP decrease" and "no IOP decrease ".This represented the probability that a subject with a decrease in IOP, randomly selected, is rated as more likely to have a decrease in IOP than a randomly chosen subject with no decrease in IOP.This was based on non-parametric Wilcoxon-Mann-Whitney U statistics performed in the calculation of the AUC.
This prospective study showed that cataract surgery associated with trabecular washout provided sustainable IOP decrease with a postoperative IOP of less than 15 mmHg over a 16-month follow-up period, indicating a postoperative recovery of normal physiological flow of the aqueous humor throughout the follow-up in these patients with PEX glaucoma.Goniowah technique using the Tran Canula® 7,8 is designed to remove macroscopic and microscopic pseudo exfoliative deposits and pigmented cells located in the entire anterior chamber of the eye, and to restore the physiological pathway of aqueous humor through the trabecular meshwork and the iridocorneal angle.The design of the Tran Canula® is characterized by two distinct water jets separated from each other by an angle of 30°.Thanks to the technological specificity of the cannula and to micro-hydrodynamic laws, these two independent fluid streams merge into one more powerful since they encounter an obstacle in front of them.So, the water flow goes straight ahead, really targets the trabecular meshwork and the iridocorneal angle, and focuses precisely on the area to be cleaned.Our results were consistent with previous studies 7,8,[11][12][13][14][15][16][17][18][19][20][21] , and more especially with a recent review 22 of six randomized controlled trials, including a total of 639 eyes with Figure Prediction performance (AUC) of the IOP BL biomarker (preoperatory intraocular pressure) for predicting the decrease in IOP (< 0 vs ≥ 0) from baseline at M 1 and M 12 months follow-up after cataract surgery associated with trabecular washout using the Goniowash technique.
gonioscopically confirmed primary and secondary open-angle glaucoma 23,24 , that highlighted a statistical difference in IOP and need for glaucoma medication between before and after cataract surgery.The mechanisms responsible for IOP reduction after surgery remain to be elucidated.They can be molecular, physiological and biomechanical 25,26 .Among the biomechanical mechanisms, fluidics might play a role.Per-operatively washing and rinsing with high flow rates the trabecular meshwork, the irido-corneal angle and the entire anterior segment, such as during the Goniowash procedure, facilitates the elimination of most macro-and microscopic pseudo exfoliation deposits and/or pigmented cells 25,[27][28][29] .In the setting of PEX glaucoma, previous studies reported greater IOP decrease in patients in which greater irrigation volumes were used intraoperatively 15,30,31 .The magnitude of IOP reduction correlated with the increased aspiration time and the irrigation fluid volume used during the intervention 30,32,33 .Additionally, preoperative IOP (IOP BL ) was a predictive biomarker of postoperative IOP decrease with a high prediction performance, sensitivity and specificity in PEX glaucoma patients with preoperative IOP ≥ 15 mmHg.This threshold of 15 mmHg represents the mean IOP of the normal physiological flow of the aqueous humor.Thus, over a 16-month follow-up, all PEX glaucoma patients with a preoperative IOP greater than or equal to the average general population IOP were likely to achieve, according to a prediction statistical relationship, a significant sustainable decrease in postoperative IOP; the higher the preoperative IOP, the greater the decrease in postoperative IOP.Age and gender were not predictive factors associated with IOP decrease.The results of our study were consistent with the few studies that investigated the predictors of lowering IOP after cataract surgery [34][35][36][37][38] and more specifically with the rare predictive studies in the setting of the non-glaucoma PEX syndrome, in 33 patients 32 , and PEX glaucoma, in 29 patients 30 .However, these studies did not provide decision threshold of the intraocular pressure, with a prediction performance, sensitivity, and specificity, to establish the patients' profile likely to achieve the best expected outcome, in terms of decrease in intraocular pressure, after cataract surgery, for a personalized decision-making in care of patients, especially of patients with pseudo exfoliation glaucoma.The threshold of about 15 mmHg for IOP-reducing effect has been shown for both topical medication with prostaglandin analogues (latanoprost) 39 in newly detected open-angle glaucoma patients and laser trabeculoplasty 40 in newly diagnosed multi-treated glaucoma patients.Both treatments affect the outflow of aqueous humor.The threshold of about 15 mmHg seems to corroborate the restitution of the physiological flow of aqueous humor.
In our study, the incidence of early transient IOP spike of 30 mmHg or greater within the first 24 h following cataract surgery was observed in 7.4% of patients.Previous studies showed that these IOP spikes were more common in PEX eyes than in non-PEX eyes 14,15,41 .This incidence has been reported between 11 and 35% in ocular PEX patients and between 3 and 27% in primary open-angle glaucoma patients 15,16 .IOP spike may have a harmful impact on the remaining nerve fibers, leading to further loss of visual field 42 .
One limitation of the study was the population size (54 eyes), although similar to previous studies 30,32 which included 33 and 29 eyes, respectively.Another limitation was the absence of discontinuation of hypotensive drugs before and after surgery, precluding the assessment of the real baseline IOP and underestimating IOP reduction following surgery 43 .Anterior chamber biometric factors were not included in the models.However, they did not demonstrate significant associations with changes in postoperative IOP in previous studies 32 , which could bias the models' adjustment.The strength of the study was its prospective design; the assessment of the optimal decision threshold of the predictive biomarker, with its prediction performance, sensitivity, specificity and odds ratio, for the identification of a patients' profile likely to achieve the expected outcome, in terms of decrease in postoperative IOP, in a population with PEX glaucoma.

Conclusion
All PEX glaucoma patients with a preoperative IOP greater than or equal to the average general population IOP (≥ 15 mmHg) were likely to achieve a significant sustainable decrease in IOP after cataract surgery associated with trabecular washout using the Goniowash technique.This biomarker may be useful for a personalized approach in care of those patients with particularly aggressive glaucoma, and eventually delaying the progression of ocular PEX disease. https://doi.org/10.1038/s41598-024-53893-5

Figure 1 .
Figure 1.Time effect of cataract surgery associated with trabecular washout using the Goniowash technique on intraocular pressure (IOP mmHg) and best corrected visual acuity (BCVA logMAR) for each relevant time of follow-up.

Figure 3 .
Figure 3. Decrease in intraocular pressure (IOP) and best corrected improvement of visual acuity (BCVA logMAR) according to the decision cut-off value of 15 mmHg of the predictive biomarker of IOP decrease (IOP BL , preoperative IOP) for each relevant time of follow up after cataract surgery associated with trabecular washout using the Goniowash technique.

Table 1 .
(A) Demographic synopsis of the study subjects.(B) Mean value ± standard deviation of intraocular pressure (IOP), best corrected visual acuity (BCVA logMAR) and Medication (medication number) throughout follow-up.(C,D) Decrease in intraocular pressure (IOP) and improvement of visual acuity (BCVA) according to the decision threshold of 15 mmHg of the predictive biomarker of IOP decrease (IOP BL , preoperative IOP) for each relevant time of follow-up after cataract surgery associated with trabecular washout using the Goniowash technique.*Wilcoxon's two-tailed signed rank test (matched pairs from baseline).**Wilcoxon rank sum test (Mann Whitney test).

Table 2 .
Predictive biomarkers and their predictive performance (AUC) obtained from ROCs to predict the decrease in IOP (yes vs no) from baseline for each relevant time of follow-up after cataract surgery associated with trabecular washout using the Goniowash technique.*Multivariate regression analyses with mixed stepwise selection, adjusted for central corneal thickness covariate.**Logistic regression models.