Abstract
purpose. To determine whether intraocular lens (IOL) power calculations for cataract surgery as measured by postoperative refractive error using partial coherence interferometry (PCI) are more accurate in improving postoperative outcomes than applanation ultrasound biometry (AUS).
methods. A double-blind randomized controlled trial consisting of 205 patients was undertaken by the Southern Health Ophthalmology Unit, Victoria, Australia. Mean absolute postoperative refractive error (MAE) represented the dependent variable; the biometric technique (PCI; AUS) used to determine the IOL power to be implanted in the surgical eye represented the independent variable. An intention-to-treat analysis was used to prevent loss of randomization caused by the effects of crossover and drop-out.
results. The MAE in patients with implanted PCI-calculated IOLs was 0.40 ± 0.37 D (SD; 95% confidence interval [CI], 0.32–0.48 D) compared with 0.45 ± 0.41 D (SD; 95% CI, 0.36–0.54 D) for patients with implanted AUS-calculated IOLs. There was no statistically significant difference between MAE in patients with implanted PCI-calculated IOLs and that in patients with AUS-calculated IOLs in analysis of best possible outcomes (t 167 = 1.0, P = 0.315).
conclusions. The results of this trial demonstrated that the calculation of IOL power based on ocular axial length measurement with PCI technology provided no clinical advantage over conventional applanation ultrasound, as measured by postoperative refractive outcome (anzctr.org.au number, ACTRN12608000077369).
Cataract extraction with the implantation of an artificial intraocular lens (IOL) is the most frequently performed ophthalmic surgical procedure in Australia. However, accurate calculation of the IOL power necessary for attaining the desired postoperative refraction remains an issue.
1 The refractive outcome after cataract surgery is dependent on several factors, including axial length, keratometry, anterior chamber depth, and lens formulas. Of these factors the preoperative measurement of axial length (AL) is considered to be a key determinant in calculating the IOL power to be implanted.
2 Historically, applanation ultrasound (AUS) biometry has been the most commonly used technique for AL measurement among biometrists.
3 More recently, partial coherence laser interferometry (PCI) has gained preference for calculating AL measurements for IOL implantation.
1 3
It is generally believed that PCI offers superior accuracy of the AL measurement and the IOL implant power calculation compared with AUS.
4 5 6 7 However, a review of published studies related to PCI, AUS, and cataract surgery failed to identify any randomized controlled trials (RCTs) that support this belief.
1 The one published RCT of small sample size reported no statistically significant difference in the postoperative refractive error in patients with PCI-calculated IOLs implanted, compared with those with implanted AUS-calculated IOLs.
8 It is also acknowledged that PCI is unable to measure IOL power in up to 22% of patients because of dense cataracts, subcapsular opacities, or a signal/noise ratio less than 2:1.
4 8 9 10 11 12 13 The failure of IOLMaster (Carl Zeiss Meditec, Inc., Dublin, CA) to predict IOL power in some eyes, all of which were able to be measured by AUS, has resulted in exclusion of patients adverse to PCI from comparison investigations, resulting in inconclusive clinical results and outcomes that are biased toward PCI.
8 12 13 In the present study an intention-to-treat (ITT) analysis was used to prevent loss of randomization caused by the effects of patient crossover and drop-out and to provide an unbiased assessment of the efficacy of the trial.
The objective of this prospective double-blind RCT is to determine whether PCI technology increases the accuracy of postoperative refractive outcomes compared with the current-generation AUS after cataract surgery, as determined by the mean absolute error (MAE) defined as the mean of the absolute difference between the measured and the predicted postoperative spherical equivalent.
13
An RCT with blinding of both patients and outcome observers to biometric group allocation (double-blinding) formed the fundamental study design. The MAE represented the dependent variable and the biometric technique (PCI or AUS), which determines the IOL power to be implanted in the surgical eye represented the independent variable. MAE was considered the primary outcome and was defined as the mean of the absolute difference between the measured and the predicted postoperative spherical equivalent.
The target population for the RCT was defined as the general population of patients with cataract, with the accessible population being those patients referred to, approved, and booked for cataract surgery at the Southern Health day surgery center. Before the trial commenced, a sample size of 158 patients was calculated as the minimum required to detect a 0.24 D difference (power 90%, α = 0.05) in the MAE between patients with PCI-calculated IOLs and those with AUS-calculated IOLs. After factoring for sample attrition and the reported failure rate of PCI to obtain AL measurement,
1 the predetermined sample size was increased to 205 patients. Eligibility criteria were kept simple to maximize generalization of the results to the target population.
14
During the trial period of 142 days (April 6, 2006 to August 24, 2006) all patients attending the preadmission clinic were initially examined by ophthalmologists and informed consent was obtained for all patients approved for cataract surgery. Before surgery, the AL and IOL power calculation for the designated surgical eye of all patients was measured using both PCI (IOLMaster; Carl Zeiss Meditec) and AUS (Microscan Model 100A+; Sonomed Inc., Lake Success, NY) corneal contact biometry and the IOL power was kept blind with regard to treatment group allocation. All PCI AL measurements were performed by the primary researcher, and all AUS AL measurements were conducted by the senior orthoptist who was blind to the PCI results. PCI AL scans were conducted before AUS AL scans for all patients. To eliminate the confounding variable introduced by keratometry performed with different techniques on treatment groups, autokeratometry with the IOLMaster was performed on all patients.
On completion of the preadmission clinic process, patients were given either a date for surgery, or were discharged as not requiring surgery. Using a lottery system, a simple random sample of 205 patients was selected as the trial participants from the 410 patients consecutively booked for cataract surgery during the specified trial period. Each patient had an equal probability of being assigned as a trial participant, and each trial participant was assigned independently of all others
(Fig. 1) .
PCI was unable to measure the AL for 36 (17.6%) of the 205 trial participants compared with AUS which successfully obtained AL measurements for all trial participants. Of the 169 patients successfully measured by both PCI and AUS, 84 were randomly allocated to receive a PCI-calculated IOL and 85 an AUS-calculated IOL. The comparison between the two groups was termed “analysis of best possible outcomes”
(Fig. 1) . The process of random treatment group allocation involved drawing an opaque envelope that contained a card stating either PCI or AUS, for each trial participant before surgery. Each participant had an equal probability of being allocated to either treatment group and each allocation was independent of all others. All 36 (17.6%) PCI failures received an AUS-calculated IOL by default; however, before implantation, the 36 patients underwent a standard random allocation process to enable an ITT analysis to be performed. Before default AUS IOL implantation, 18 PCI failures were allocated to receive a PCI-calculated IOL. The PCI and AUS ITT groups were termed PCI-ITT and AUS-ITT, respectively
(Fig. 1) .
After trial assignment and treatment group allocation, the primary researcher obtained demographic and baseline ocular information for each patient from the standard hospital surgical admission forms and preadmission ophthalmic history and examination notes. Selection and randomization of trial participants, data collection, and analysis were all centrally controlled and concealed by the primary researcher. The trial protocol was approved by the Southern Health Human Research Ethics Committee (March 2, 2006). The protocol adhered to the tenets of the Declaration of Helsinki.
IOLMaster autokeratometry was performed before AUS on all patients, to avoid corneal contact, which may affect the readings.
1 4 The IOLMaster auto-keratometry protocol was used, with the median of three measurements within 0.3 D in each meridian being selected. PCI AL measurement was conducted in accordance with the IOLMaster AL scan protocol, with measurements repeated until four scans were consistent within ±0.02 mm of ideal waveform and acceptable signal-to-noise ratio (SNR > 2.0). The average of the measurements was accepted as the final result. Based on the keratometry and PCI AL measurements, the PCI IOL implant power was calculated by the IOLMaster with the SRK/T formula,
15 with the A constant set at 118.9, as recommended by the IOL manufacturer. The SRK/T formula was chosen based on empiric evidence that demonstrated it to perform comparably to all other lens formulas over the entire distribution of ALs.
15 16 Patients for whom the criteria could not be achieved after 20 scans were termed PCI failures.
AUS, when measuring AL (SMM100A+), emits a wavelength of 10 ± 1 MHz with an estimated AL measurement accuracy of ±0.10 mm. Into each eye one drop was instilled of both oxybuprocaine hydrochloride 0.4% and liquid gel eye drops (Thera Tears; Advanced Vision Research, Woburn, MA) before AUS biometry commenced. AUS measurements were repeated until four high-quality scans were consistent within ±0.10 mm. The highest quality scan was then chosen as the final result. The SMM100A+ system was limited to the SRK-II formula and the AUS IOL implant power calculation was based on the IOLMaster autokeratometry and AUS AL measurements. The A constant was set at 118.7, as recommended by the IOL manufacturer.
Phacoemulsification was performed through a superior corneoscleral incision (3.2 mm). An aspheric acrylic posterior chamber IOL (SN60WF; Alcon, Fort Worth, TX) was implanted in the capsular bag of 201 patients. In four patients posterior capsule rupture excluded placement of the IOL within the capsular bag and each patient received a ciliary sulcus fixation IOL (MA60AC; Alcon). The surgical team comprised eight consultants and four senior ophthalmology registrars.
All patients were examined by an ophthalmologist 7 to 12 days after surgery at the cataract day surgery center. In the fifth postoperative week, patients returned for refraction to their community ophthalmologists or optometrists, who were blind to both trial assignment and treatment group allocation of patients. The community ophthalmologists and optometrists followed their own standard methods for measuring refraction—namely, subjective (59%) or autorefractor (41%). The final refraction for each patient was forwarded to the primary researcher, converted to its spherical equivalent, and compared with the preoperative prediction made by PCI and AUS.
A database was used for entry, validation, and preparation of data for analysis (Excel 2003; Microsoft, Redmond, WA), and another program was used for statistical analysis (SPSS for Windows, ver. 14.0; SPSS, Chicago, IL). Before significance testing, MAE data underwent cube root transformation to satisfy the assumptions of normality and homoscedasticity. P < 0.05 was considered to be statistically significant. Comparison of the postoperative MAE in eyes allocated to PCI versus AUS was performed with Student’s t-test (two-tailed). The χ2 statistic was used to assess the proportional variation of patients achieving an MAE within the dioptric ranges of 0.5, 1.0, 1.5, and 2.0 D, on the basis of the biometric technique. To test the validity of the postoperative refraction, we used Student’s two-tailed t-test to compare the postoperative spherical equivalent refraction in eyes refracted by subjective refraction versus autorefractor.
Since 36 PCI failures required default crossover to implantation with AUS-calculated IOLs, the MAE was undefined based on initial group allocation. It is generally accepted that patients unable to adhere to a given intervention differ with respect to outcomes from patients who are able to adhere to the given intervention.
17 18 In addition, excluding the 36 PCI failures from the analysis leaves those who may have a better outcome and destroys the unbiased comparison afforded by randomization.
19 20 ITT analysis is a method that includes noncompliant patients in the groups into which they were originally randomized and provides a conservative estimate of the treatment effect compared with what would be expected if there were full compliance.
Before treatment, the 95% confidence interval [CI] of each of the variables age, sex, best corrected visual acuity (BCVA), VF-14 score, keratometry, and AL
(Table 1)and coexisting ocular disease in the four patient groups
(Table 2)all overlapped, which confirmed the random selection process.
In the 205 patients, the proportion of cataract types within each RCT group was similar
(Table 3) . The 36 eyes in which the PCI AL measurement failed had dense cataracts: 3 (8.3%) nuclear, 2 (5.6%) cortical, 4 (11.1%) posterior subcapsular, 3 (8.3%) mature, and 24 (66.7%) mixed, of which 20 (83.3%) contained a posterior subcapsular component.
Excluding PCI failures (
n = 36) an analysis of best possible outcomes between patients with implanted PCI-calculated IOLs compared with patients with AUS-calculated IOLs found no statistically significant difference between the MAE and transformed MAE for each group,
t 167 = 1.0,
P = 0.315
(Table 4) . The mean numerical error (MNE) was near to 0 for both methods
(Table 4) .
Similarly, there was no statistically significant difference between the proportion of eyes with implanted PCI-calculated IOLs and the AUS-calculated IOLs achieving a postoperative refraction within 0.5, 1.0, 1.5, and 2.0 D (χ
2 4 = 5.59,
P = 0.133;
Table 5 ).
A group analysis that excludes PCI failures may cause serious bias and overestimate the clinical effectiveness of the outcome measure if an ITT analysis is not applied. An ITT analysis is based on the initial treatment intent, not on the treatment eventually administered, and all patients who begin the treatment are considered to be part of the trial whether they finish or not. If some participants have a refractory or serious problem such as PCI failure and are excluded from the trial, the remaining participants may have a better outcome, and the unbiased comparison afforded by randomization is destroyed.
19 The result of including the PCI failures in the ITT model
(Table 6)was a statistically significant difference between the proportion of eyes able to achieve a postoperative refraction within 0.5, 1.0, 1.5, and 2.0 D of the predicted spherical equivalent between PCI-ITT compared with AUS-ITT (χ
2 4 = 18,
P < 0.01).
Analysis of postoperative spherical equivalent refraction between patients measured by subjective refraction compared with those measured by autorefractor found no statistically significant difference between the two groups. The mean spherical equivalent for the subjective refraction group was −0.07 D (95% CI, −0.20 to 0.05), the mean for the autorefractor group was −0.08 (95% CI, −0.23 to 0.06; t 203 = 0.08, P = 0.938).
To our knowledge, there has been no previous RCT comparing PCI to conventional AUS on an ITT basis. The participants of the present trial comprised a simple random sample of cataract surgical patients referred to a large cataract day surgery clinic. The baseline characteristics of the study sample were consistent with the typical cataract patient population reported in the literature providing a degree of confidence in the external validity of the results.
21 22 23 Similarly, the presence of ocular comorbidities among subjects in the present study
(Table 2)was consistent with that reported in recent local and international studies.
24 25 The sample frame of the present trial was constructed in a manner that allowed the MAE of the best possible outcomes for PCI to be directly compared to the MAE achieved by AUS on a comparable population of patients, in addition to allowing results to be analyzed on the basis of ITT.
Almost all published studies comparing intraocular lens power calculations determined by PCI and AUS are not randomized, have small sample sizes, and generally exclude from the analysis patients with PCI failures, dense cataract, and retinal disorders. The result of poor-quality studies, particularly those that exclude patients, is a consistent overestimation of the treatment effect by up to 40%.
26 27 Inadequate concealment of the treatment group may also result in significant bias,
26 with treatment effects of 30% on average seemingly more beneficial than treatment effects reported in studies in which adequate concealment was implemented.
27 In the present study the postoperative refraction was measured by the referring ophthalmologists or optometrists who were blind to both trial assignment and treatment group allocation of patients and used their preferred methods for refraction. Consistent with previous reports,
28 29 a comparison of the subjective and autorefractor methods of postoperative refraction showed no significant difference in measurement, confirming the validity of this methodology.
Many published studies that we reviewed excluded up to 20% of patients who were not measurable by PCI because of inattention, corneal scarring, and dense cataracts, but all were measurable by AUS.
1 10 30 For interventions to be directly compared, all randomized patients must be included in the analysis and the final results analyzed with respect to the original groups.
31 32 33 34 When results are not compared on an ITT basis, selection bias renders the results invalid.
17 26 27 34 This problem is highlighted by a recent RCT in which four PCI failures were excluded from the PCI sample. The results showed that PCI improved the predictive value of the postoperative refraction compared with AUS.
8 However, a recalculation of the data
8 in an ITT analysis that included the four PCI failures found no significant difference in the postoperative refraction between the two study groups.
It is generally agreed that accurate biometry is the most important factor in achieving a successful refractive outcome after lens implantation,
35 and currently AUS is the most widely used technique for biometry.
3 12 It is generally accepted that the IOLMaster offers superior reproducibility of AL measurement in comparison with AUS biometry.
1 13 36 It is also apparent that the most significant limitation of PCI is poor laser penetration in eyes with dense media opacities, especially posterior subcapsular cataracts.
12 In this study, there was a PCI measurement failure rate of approximately 17.6% or one in six of all surgically treated eyes because of dense media opacities, all of which could be measured by AUS, indicating that at present PCI cannot supersede AUS for all routine biometry. This situation may change if cataracts are removed before they become too dense for the PCI measurement.
10
Systematic measurement errors translate into equivalent errors in IOL power prediction and may be corrected by optimization.
36 In this RCT we used the A constants provided by the lens manufacturer for PCI and AUS rather than customize the IOLMaster- and AUS-derived A constants. This decision was supported by the evidence that a comparison of the SRK/T formula to that of the SRK-II, Binkhorst II, Hoffer, and Holladay formulas in 1050 eyes, reported that the SRK/T formula was not significantly better than the SRK-II, Binkhorst, or Hoffer formulas for all eyes.
16 Since the MNE of predictions counting both negative and positive deviations averaged near zero for the study dataset, it is unlikely that the use of customized A constants would have significantly altered the outcomes.
It is accepted that PCI is less time consuming and more patient friendly than is AUS. However, PCI has a significant failure rate, particularly in the presence of dense cataracts, which does not occur with AUS. Although AUS entails topical anesthesia and corneal applanation, it is able to measure AL in all eyes except those with intravitreous silicone oil. The most revealing finding of this RCT is that although many publications claim the IOLMaster to be superior to AUS in calculating IOL power, it is only by excluding those eyes that cannot be measured by the IOLMaster. The most significant finding of this prospective double-blind RCT is that the calculation of IOL power based on ocular AL measurement using PCI technology provided no clinical advantage over conventional AUS, as measured by postoperative refractive outcome. It is likely that the apparent improvement of PCI over AUS reported in present studies will not be reinforced in future randomized clinical trials.