To evaluate cost-effectiveness of penetrating keratoplasty (PK), femtosecond laser-assisted Descemet stripping endothelial keratoplasty (FS-DSEK), and Descemet stripping automated endothelial keratoplasty (DSAEK).
Cost-effectiveness analysis based on data from a randomized multicenter clinical trial and a noncomparative prospective study.
Data of 118 patients with corneal endothelial dysfunction were analyzed in the economic evaluation. Forty patients were included in the PK group, 36 in the FS-DSEK group, and 42 in the DSAEK group. The primary incremental cost-effectiveness ratio (ICER) was the incremental costs per clinically improved patient, defined as a patient with a combined effectiveness of both a clinically improved BSCVA (defined as an improvement of at least 2 lines) and a clinically acceptable refractive astigmatism (defined as less than or equal to 3.0 diopters). Analysis was based on a 1-year follow-up period after transplantation.
The percentage of treated patients who met the combined effectiveness measures was 52% for DSAEK, 44% for PK, and 43% for FS-DSEK. Mean total costs per patient were €6674 (US$7942), €12 443 (US$14 807), and €7072 (US$8416) in the PK group, FS-DSEK group, and DSAEK group, respectively. FS-DSEK was less effective and more costly compared to both DSAEK and PK. DSAEK was more costly but also more effective compared to PK, resulting in incremental costs of €4975 (US$5920) per additional clinically improved patient.
The results of this study show that FS-DSEK was not cost-effective compared to PK and DSAEK. DSAEK, on the other hand, was more costly but also more effective compared to PK. Including societal costs, a longer follow-up period and preparation of the lamellar transplant buttons in a national cornea bank could improve the cost-effectiveness of DSAEK.
For many years, penetrating keratoplasty (PK) has been the gold-standard technique for corneal transplantation resulting from endothelial disease or endothelial dysfunction. Several studies have shown that the technique is safe and effective. However, PK has several drawbacks, including slow visual recovery, suture-related events, graft failure, and wound healing problems.
In recent years, interest has grown in endothelial keratoplasty (EK), in which only the diseased endothelium is transplanted and the healthy anterior stromal tissue is preserved. As compared to PK, EK shows faster visual recovery, less change in astigmatism, better stability and predictability of the postoperative refraction, and fewer suture-related problems. However, several EK studies found an increased risk of graft dislocation and higher endothelial cell loss.
In EK, different techniques have been developed for the preparation of the donor cornea, including Descemet stripping automated endothelial keratoplasty (DSAEK), where the donor cornea is prepared with a microkeratome, and femtosecond laser-assisted Descemet stripping endothelial keratoplasty (FS-DSEK), where the donor tissue is prepared with a femtosecond laser.
The efficacy and visual outcomes of EK techniques vs PK have been well studied. However, the efficiency of the techniques is less elucidated. In an economic evaluation, outcomes and costs of interventions are compared with the aim to improve resource allocation decisions by policymakers and insurers. One type of economic evaluation is the cost-effectiveness analysis, in which the costs of an intervention are related to a single outcome measure, for example “clinical success.”
In this study, we performed an evaluation of the cost-effectiveness of FS-DSEK, DSAEK, and PK, for which data from a randomized multicenter clinical trial and a noncomparative prospective study were used.
Patients and Methods
In this study, costs and effects of PK and 2 different techniques of EK (ie, FS-DSEK and DSAEK) were compared. Empirical data were obtained from 2 sources. First, PK and FS-DSEK were compared in a randomized clinical trial, the Dutch Lamellar Corneal Transplantation Study (DLCTS), which was performed between April 2005 and April 2007. Patients were included in 5 ophthalmic centers in The Netherlands. The study was approved by the institutional review boards of all participating centers. Before inclusion, informed consent was obtained from all patients. More details about the safety, efficacy and quality of vision of PK vs FS-DSEK are described elsewhere.
Second, a third group of patients who received a DSAEK transplantation in Maastricht University Medical Center between April 17, 2008 and March 11, 2010 was included in the study. Inclusion in the economic evaluation was based on the completion of quality-of-life questionnaires on at least 2 out of 3 measurements.
For all 3 groups, inclusion criteria were patients with Fuchs endothelial dystrophy, aphakic or pseudophakic bullous keratopathy (ABK/PBK), posterior polymorphous dystrophy, or iridocorneal endothelial syndrome; a minimum patient age of 18 years; and best spectacle-corrected visual acuity (BSCVA) lower than 20/50. Patients with mental retardation, previous PK, and human leukocyte antigen–typed keratoplasty were excluded.
All patients underwent a comprehensive ophthalmic examination and completed quality-of-life questionnaires. Data were obtained at study inclusion (T0) and 6 (T1) and 12 (T2) months after the transplantation, which was 9 and 15 months after baseline measurement, respectively.
In the FS-DSEK group, all donor posterior lamellar discs (PLDs) were prepared in Maastricht University Medical Center (MUMC) with a 30-kHz femtosecond laser (AMO-Intralase Corp, Irvine, California, USA), as previously described. In the DSAEK group, the PLDs were created with the use of a microkeratome (Moria Inc, Doylestown, Pennsylvania, USA). In both groups, an 8.0-mm donor corneal disc was trephined from the corneoscleral button with a disposable Barron trephine. In the recipient, a 5.0-mm corneoscleral incision and 2 limbal paracenteses were made. The Descemet membrane was scored and a circle of 7.5 mm Descemet membrane and endothelium was stripped from the posterior stroma. Four transcorneal incisions were made in the midperipheral recipient cornea to drain fluid between the recipient cornea and PLD. After coating the endothelial surface with viscoelastic material, the PLD was gently folded into a taco configuration and inserted. The corneoscleral incision was closed with 4 10-0 nylon sutures. An air bubble was injected to unfold the PLD and to press the PLD against the recipient cornea.
In the PK group, the recipient cornea was trephined using a 7.75-mm or 8.0-mm Hessburg-Barron vacuum trephine (Barron Precision Instruments, Grand Blanc, Michigan, USA), whereas the donor cornea was trephined with an 8.0-mm or 8.25-mm disposable trephine (Medical Workshop, De Meern, The Netherlands). In all cases, a combined suturing technique of a running 11-0 nylon suture with 8 interrupted 10-0 nylon sutures was performed. Selective suture removal was based on topographic astigmatism pattern.
In all treatment groups, patients with cataract either underwent primary cataract extraction with posterior chamber intraocular lens (IOL) implantation either before the corneal transplantation or combined with EK or PK in 1 surgical session.
The economic evaluation was performed from a health care perspective with a time horizon of 15 months, from study inclusion (about 3 months before surgery) up to 12 months after surgery. All relevant resources consumed within the health care sector were taken into account.
Effectiveness was based on 3 outcome measures. Two clinical outcome measures were used, namely the BSCVA and the refractive astigmatism. Furthermore, the National Eye Institute Visual Functioning Questionnaire (NEI VFQ-25) was used as an outcome measure, reflecting vision-specific health-related quality of life.
To evaluate cost-effectiveness, incremental cost-effectiveness ratios (ICERs) were calculated by dividing the difference in costs by the difference in effectiveness between 2 treatments. The primary ICER was the incremental costs per clinically improved patient, defined as a patient with a combined effectiveness of both a clinically improved BSCVA and a clinically acceptable refractive astigmatism. In the secondary analyses, other outcome measures were used, namely costs per improved patient on the NEI VFQ-25 and costs per improved patient on all 3 outcome measures.
Best spectacle-corrected visual acuity
The BSCVA was determined using the Early Treatment of Diabetic Retinopathy Study (ETDRS) letter charts and was converted to logarithm of the minimal angle of resolution (logMAR) measurements. Vision levels of counting fingers, hand movements, light perception, and no light perception were substituted by logMAR values of 1.7, 2.0, 2.5, and 3.0, respectively. A patient with an improvement of at least 2 lines on the ETDRS letter chart between T0 and T2 was considered to be clinically improved.
Refractive astigmatism was used as an outcome measure because a lower level of postoperative astigmatism is a major advantage of EK techniques compared to PK and refractive astigmatism is considered to be more important for the patient compared to topographic astigmatism. In this study, a postoperative refractive astigmatism less than or equal to 3.0 diopters (D) at T2 was considered to be clinically acceptable.
National Eye Institute Visual Functioning Questionnaire
The NEI VFQ-25 is a vision-specific health-related quality-of-life questionnaire that measures domains related to daily visual functioning that are most important for patients with vision deficits. The NEI VFQ-25 consists of 25 questions that can be expanded with 13 additional items. All questions comprise 12 subscales. Eleven of them can eventually be converted into 1 single score. The remaining subscale (general health) is included in the questionnaire to provide robust information about an individual’s general health status. After recoding, the scores of the subscales and the single composite score range from 0 (worst possible value) to 100 (best possible value). As several studies have suggested that a 10-point change in the composite score is clinically important, a patient with a minimum gain of 10 points in the composite score at T2 as compared to T0 was considered to be clinically improved in this study.
All relevant health care costs of the 3 interventions were assessed according to the Dutch guidelines for cost calculations. Costs outside the health care sector, such as productivity losses, were excluded from analyses. Costs were calculated by multiplying the volumes of resource use by the cost price per resource unit. All costs were converted to 2010 Euros. Costs occurring after 12 months were discounted at an annual rate of 4%. All costs are reported in Euros (€) and United States dollars (US$) (€1 = US$1.19).
Resources included preparation, preservation, allocation, and transportation of the donor corneas, which were provided by the Bio Implant Services (BISLIFE) Foundation (Leiden, The Netherlands); outpatient visits; diagnostic procedures; preparation of the donor PLD; surgical procedures; hospitalization; and postoperative drugs use. Data of outpatient visits, surgical and diagnostic procedures, and hospitalization were obtained from the registries of the participating hospitals.
Services of the BISLIFE Foundation were valued using Dutch reimbursement rates, as no unit prices were available. Costs of outpatient visits, surgical procedures, hospitalization, and diagnostic procedures were valued using standardized integral unit prices (consisting of personnel, material, capacity, and overhead costs) as calculated by MUMC, in which 71% of the patients were included. Integral cost prices of surgical procedures depended on the operation time per patient and were based on 2 cost drivers, namely general operating room costs and specific ophthalmology costs. Costs of preparation of the donor PLD were based on the costs of the femtosecond laser and microkeratome, costs of disposables, and personnel costs. Costs of postoperative drugs were obtained from the Dutch Pharmacotherapeutic Compass.
Outcomes were analyzed according to the intention-to-treat principle. Incomplete data from the NEI VFQ-25, incomplete cost data, and missing data concerning the BSCVA and refractive astigmatism were imputed, using SPSS Multiple Imputation (version 17.0 for Windows, SPSS Inc, Chicago, Illinois, USA), under the assumption that the data were missing at random. A linear regression model was used with a total run length of 100 iterations. Covariates included in the imputation model were visual outcome measures, which were obtained at T0, T1, and T2. Furthermore, age, sex, study group, and results on additional questions to measure patient satisfaction after a corneal transplantation were included in the model. In total, 5 imputed data sets were obtained. Data analyses were performed with each of these sets and the results were pooled.
For continuous data, the differences between groups were analyzed using 1-way ANOVA or a linear regression model with 2 dummy variables representing the groups with PK as the reference. The Pearson χ 2 test was used to compare categorical data. To analyze the change between the postoperative measurements and the preoperative measurement, a paired sample t test was used. A P value of less than .05 was considered to be statistically significant.
As cost data regularly are highly skewed, traditional parametric and nonparametric statistical methods are not appropriate to analyze the difference in mean costs between groups. Therefore, we performed nonparametric bootstrapping with 1000 replications to estimate the uncertainty in the incremental costs and effects, using Microsoft Excel for Windows (Microsoft Corp, Redmond, Washington, USA).
To investigate cost-effectiveness, ICERs were calculated. For this purpose, the treatments were ranked from the most effective to the least effective. A treatment that was less effective and more costly than at least 1 alternative was said to be dominated and ruled out from the calculation of ICERs. Then, each treatment was compared to the next most effective treatment by calculating the ICER. Each treatment with a higher ICER than that of a more effective intervention was ruled out based on so-called extended dominance.
In order to show the probability of each treatment being the optimal choice for a range of ceiling ratios representing the maximum amount of money that the decision maker is willing to pay for an additional health effect, cost-effectiveness acceptability curves (CEACs) were created. Based on these curves, the cost-effectiveness frontier can be determined, which indicates which strategy is to be preferred for a range of ceiling ratios.
Sensitivity analyses were performed to test the robustness of the results. First, the cut-off point for a patient with an improved BSCVA was at least 2 lines on the ETDRS letter chart between T0 and T2. In the sensitivity analyses, we tested the impact on the primary analysis of changing this to at least 1 line and at least 3 lines.
Second, postoperative refractive astigmatism less than or equal to 3.0 D at T2 was considered to be clinically acceptable. In the sensitivity analyses, we tested the impact on the primary analysis of changing this to less than or equal to 2.0 D and less than or equal to 4.0 D.
Third, it has been stated that a 5-point change in the composite score of the NEI VFQ-25 is clinically relevant. Therefore, we considered a patient with a gain of 5 points in the composite score at 12 months postoperatively to be clinically improved.
Fourth, we replaced the cost prices as determined in the MUMC by standardized Dutch unit prices—if available—determined by Oostenbrink and associates.
Finally, the cost price of the femtosecond laser was based on the actual use at the time of the study. Currently, however, the laser is also used for the implantation of intracorneal ring segments in about 30 keratoconus patients per year and in the future will be used for about 300 LASIK procedures per year.
In all treatment groups, a number of patients with cataract underwent primary cataract extraction with posterior chamber IOL implantation combined with EK or PK in 1 surgical session. As this resulted in an increased operation time for these patients, this could bias the costs per patient. In addition, these patients may have a lower preoperative BSCVA compared to patients who do not have cataract. Therefore, we performed a subgroup analysis on the patients who had only EK or PK surgery without cataract surgery.
Data of 118 patients were analyzed in the economic evaluation. Forty patients were included in the PK group, 36 in the FS-DSEK group, and 42 in the DSAEK group.
In Table 1 , baseline patient characteristics of the 3 treatment groups are shown. The mean age was 71.4 ± 11.3, 68.9 ± 8.8, and 70.8 ± 11 years in the PK group, FS-DSEK group, and DSAEK group, respectively ( P = .59). In all groups, the main reason for keratoplasty was Fuchs endothelial dystrophy and pseudophakic bullous keratopathy. In the DSAEK group, more patients were pseudophakic (27/42; 64.3%) as compared to the PK patients (18/40; 45.0%) and FS-DSEK patients (17/36; 47.2%), but this was not significantly different ( P = .08).
|Age in years (mean ± SD)||71.4 ± 11.3||68.9 ± 8.8||70.8 ± 11||.59 b|
|Women, n (%)||27 (67.5%)||21 (58.3%)||27 (64.3%)||.70 c|
|Diagnosis, n (%)||.66 c|
|Fuchs endothelial dystrophy||20 (50.0%)||21 (58.3%)||22 (52.4%)|
|Pseudophakic bullous keratopathy||19 (47.5%)||15 (41.7%)||19 (45.2%)|
|Posterior polymorphous dystrophy||1 (2.5%)||0 (0%)||0 (0%)|
|Iridocorneal endothelial syndrome||0 (0%)||0 (0%)||1 (2.4%)|
|Recipient lens status, n (%)||.08 c|
|Aphakic||1 (2.5%) a||0 (0%)||3 (7.1%) a|
|Phakic||21 (52.5%)||19 (52.8%)||12 (28.6%)|
|Pseudophakic||18 (45.0%)||17 (47.2%)||27 (64.3%)|
a Aphakic eyes with iris-fixated anterior chamber intraocular lens.
At 12 months follow-up, only 1 eye in the PK group had all sutures out, whereas the remaining 39 eyes still had most or all of their sutures in.
In Table 2 , preoperative and postoperative BSCVA and refractive astigmatism are displayed. Preoperatively, mean logMAR BSCVA was 0.72, 0.82, and 0.76 in the PK group, FS-DSEK group, and DSAEK group, respectively ( P = not significant [NS]). At 6 and 12 months postoperatively, mean BSCVA was significantly better in the PK and DSAEK groups as compared to the FS-DSEK group. However, mean BSCVA gain was not significantly different. Preoperatively, mean refractive astigmatism was −1.27 ± 0.19 D, −0.99 ± 0.17 D, and −1.27 ± 0.19 D in the PK, FS-DSEK, and DSAEK groups, respectively ( P = NS). At 6 and 12 months postoperatively, mean refractive astigmatism was significantly lower in the FS-DSEK group and DSAEK group as compared to the PK group.
|BSCVA (logMAR), mean ± SD|
|Preoperative||0.72 ± 0.38||0.82 ± 0.42||0.76 ± 0.45|
|6 months||0.35 ± 0.25 a||0.62 ± 0.30 a , b||0.34 ± 0.26 a|
|12 months||0.36 ± 0.25 a||0.50 ± 0.18 a , b||0.31 ± 0.19 a|
|BSCVA gain (logMAR), mean ± SD|
|6 months||0.37 ± 0.38||0.20 ± 0.48||0.42 ± 0.39|
|12 months||0.36 ± 0.38||0.31 ± 0.42||0.45 ± 0.39|
|Refractive astigmatism (D), mean ± SD|
|Preoperative||−1.27 ± 1.2||−0.99 ± 1.02||−1.27 ± 1.23|
|6 months||−3.17 ± 1.9 a||−1.53 ± 1.44 a , b||−1.92 ± 1.1 a , c|
|12 months||−2.95 ± 1.96 a||−1.47 ± 1.32 a , b||−2.01 ± 2.01 a , c|
|BSCVA gain 12 months, n (%)|
|<0.0 logMAR||7 (17.5)||7 (19.4)||4 (9.5)|
|0.0–0.2 logMAR||9 (22.5)||11 (30.6)||10 (23.8)|
|0.2–0.4 logMAR||10 (25.0)||7 (19.4)||10 (23.8)|
|0.4–0.6 logMAR||4 (10.0)||3 (8.3)||6 (14.3)|
|0.6–0.8 logMAR||3 (7.5)||2 (5.6)||3 (7.1)|
|>0.8 logMAR||7 (17.5)||6 (16.7)||9 (21.4)|
|Refractive astigmatism at 12 months, cumulative n (%)|
|<1 D||5 (12.5)||11 (30.6)||8 (19.0)|
|<2 D||17 (42.5)||27 (75.0)||24 (57.1)|
|<3 D||21 (52.5)||30 (83.3)||32 (76.2)|
|<4 D||32 (80.0)||36 (100.0)||39 (92.9)|
a P < .05 between preoperative measurement and postoperative measurement.
b P < .05 between FS-DSEK and PK.
Table 3 shows the preoperative and postoperative mean composite score on the NEI-VFQ 25. Preoperatively, mean scores were 60.9, 58.2, and 56.1 in the PK, FS-DSEK, and DSAEK groups, respectively ( P = NS). At 6 and 12 months postoperatively, mean scores were found to be significantly higher in all groups as compared to the scores preoperatively. However, the change scores showed no significant differences between the groups.
|Mean ± SE||Mean ± SE||Mean ± SE|
|Preoperative||60.9 ± 2.6||58.2 ± 3.0||56.1 ± 2.3|
|6 months||69.8 ± 2.7 a||67.2 ± 3.4 a||69.3 ± 1.7 a|
|12 months||73.3 ± 2.4 a||69.5 ± 2.8 a||71.2 ± 2.2 a|
|Change scores T1 – T0 b||8.8 ± 2.2||9.0 ± 2.4||13.2 ± 2.5|
|Change scores T2 – T0 b||12.4 ± 2.2||11.3 ± 2.1||15.2 ± 2.6|
a P < .05 between preoperative measurement and postoperative measurement.
b T0 = preoperatively; T1 = 6 months postoperatively; T2 = 12 months postoperatively.
At 12 months postoperatively, pooled data showed that 44% of the patients in the PK group were clinically improved (defined as both a clinically improved BSCVA and a clinically acceptable refractive astigmatism), 43% in the FS-DSEK group and 52% in the DSAEK group.
Table 4 shows the mean resource use and mean costs per patient. Mean total costs per patient were €6674 (US$7942), €12 443 (US$14 807), and €7072 (US$8416) in the PK group, FS-DSEK group, and DSAEK group, respectively. Differences in costs were mainly caused by the costs of the preparation of the posterior lamellar disc.
|Costs Per Unit (€)||Mean Resource Use||Mean Costs ± SD (€)|
|PK (n = 40)||FS-DSEK (n = 36)||DSAEK (n = 42)||PK (n = 40)||FS-DSEK (n = 36)||DSAEK (n = 42)|
|BISLIFE Foundation services||3747/cornea||1||1||1||3747||3747||3747|
|Subtotal preoperative costs||3779||3790||3783|
|Costs femtosecond laser||4821/lamel||—||1||—||—||4821||—|
|Costs microkeratome transplantation||433/lamel||—||—||1||—||—||433|
|Operating room costs||9.68/minute||103||139||126||999||1343||1217|
|Operating room costs||9.68/minute||25||38||2||246||371||22|
|Subtotal surgical procedures||1802||7277||2252|
|Subtotal follow-up visits||425||471||265|
|Total costs (€)||6674 ± 82||12443 ± 97||7072 ± 30|