The current methods to remove a cataract are now very reliable with great reproducibility. A key component of determining a successful outcome is the ability to calculate the power of the lens implant used to replace the natural lens.
The refractive power of the human eye is dependent on three factors, the power of the cornea, the power of the lens (and where it will sit in the eye) and the length of the eye. During cataract surgery a replacement intraocular lens (IOL) is inserted. By knowing the power of the cornea and the axial length of the eye, it is possible to calculate the power of this replacement lens to give the desired refractive outcome.
Biometry is the process of measuring the corneal power and length of the eye. Inaccuracy in either of these measurements will lead to an unpredicted postoperative refractive error.
Corneal power accounts for about 2/3 the total power of the eye and errors in calculation will have a significant effect on the refractive outcome. Corneal power is calculated from measurements made by a keratometer or by a corneal topographer. The calculation of corneal power is based on the curvature (steepness) of the cornea. In keratometry, assumptions are made of a fixed relationship between the front and back corneal surfaces and its uniform spherical shape when making this calculation. This relationship between corneal surfaces is particularly altered during corneal refractive surgery. Some corneal topographical methods measure the anterior and posterior radii of corneal curvature as well as corneal thickness and use these to calculate corneal power.
The accuracy of axial length measurement is crucial in IOL power calculations. A 1mm error in measurement can lead to an equivalent power error of 3.00D. The axial length of the eye may be measured by ultrasound (contact or immersion) or by optical means. Ultrasonography Amplitude scan (A-scan) measures the time taken for an ultrasonic pulse to travel from the cornea to the retina and from this calculates the distance travelled between the two points. Optical methods use partial coherence laser tomography, and use the interferometry principle to calculate distance from the cornea to the retina.
Once the measurements of the eye have been made, the power of the replacement intraocular lens can be calculated. The formulas for these calculations are generally incorporated into the biometry equipment software and include one or more constants which are specific for a particular lens. They are supplied by the lens manufacturer but may be refined or optimised by a surgeon, taking into account their previous surgical results.
The accuracy and consistency of biometry is dependent on the operator, the individual equipment and the appropriateness of the formulas used, all of which contribute to accuracy and therefore to the refractive outcome of surgery
Optimal biometry is critical to the success of the cataract surgery in terms of the actual refractive outcome being congruent with the required refractive outcome. It is critical therefore that the person undertaking the biometry is competent to undertake the procedure, and a competence framework has been developed by the ophthalmic professional organisations and is available from: https://www.rcophth.ac.uk
Risk Stratification
Risk stratification is a tool for identifying or predicting which patients are at high risk of complications, in this case in cataract surgery. By analysing a large database of patients undergoing cataract surgery and the incidence of complications and their outcomes, it has been possible to determine which patient characteristics and what preoperative comorbidities are likely to be associated with per- and postoperative complications and a poor visual outcome.
Risk stratification tools can be used to alert the surgeon to potential complications and poor outcomes and therefore be able to more accurately counsel the patient and arrange for the cataract surgery to be performed by surgeons with the appropriate skills.
Risk stratification is also an important component of surgical audit, allowing more accurate assessment and benchmarking of outcomes.
7.1. Biometry techniques
7.1.1. Review question
What is the effectiveness of different techniques for undertaking biometry?
7.1.2. Introduction
The review focussed on identifying studies that fulfilled the conditions specified in Table 12. For full details of the review protocol, see Appendix A. The main outcome for this review question was the predictive accuracy of the different techniques, assessed by deviations from the predicted refractive outcome expressed as a spherical equivalent. As suggested by Gale et al. (2009), a benchmark standard of 85% of individuals achieving a final spherical equivalent within 1.00 dioptre of the predicted refraction and 55% of individuals within 0.50 dioptres was used to evaluate the clinical relevance of the review findings.
Table 12
PICO inclusion criteria for the review question on biometry techniques.
Randomised controlled trials (RCTs) comparing different biometry and keratometry techniques in adults undergoing phacoemulsification cataract surgery to predict the accuracy of postoperative refraction were included. Papers were excluded if they:
were guidelines/health technology assessment reports, narrative reviews, case studies/reports/series, reliability studies, diagnostic accuracy studies, non-comparative studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
focused on combination surgical procedures, that is cataract surgery in tandem with other surgical procedures (for example, phacotrabeculectomy, canaloplasty, Descemet’s stripping automated endothelial keratoplasty)
compared biometry techniques with no biometry only or standard care that was not specified
were not published in the English language.
For flow charts of study inclusion and exclusion, see Appendix K. For the list of excluded studies with reasons, see Appendix F.
Protocol deviation
Only one RCT published in 1995 comparing standard keratometry and corneal topography on 46 people undergoing phacoemulsification cataract surgery was identified (Antcliff et al., 1995). The Guideline committee noted that keratometry techniques are routinely used as current standard practice in the NHS, while topography which requires greater expertise (training) and time is used in specific circ*mstances, such as for individuals with a history of corneal refractive surgery that results in an increased risk of postoperative refractive errors stemming from difficulties in estimation of corneal power. Therefore, the committee agreed that it would be useful to further consider observational evidence comparing keratometry techniques and topography only within this specific subgroup. Two observational studies comparing keratometry with topography in individuals with a history of corneal refractive surgery undergoing phacoemulsification cataract operations were identified.
7.1.3. Evidence review
In total, 18,080 references were found for a combined database search for all 4 related review questions on biometry and postoperative refractive errors, with 315 articles ordered for full-text review. Five unique RCTs were identified for the comparison of ultrasound and optical biometry (Fontes et al., 2011; Kolega et al., 2015; Naicker et al., 2015; Rajan et al., 2002; Raymond et al., 2009). One RCT was identified for the comparison of keratometry and topography (Antcliff et al., 1995), while two retrospective case series were identified for this comparison in the specific subgroup of individuals undergoing phacoemulsification cataract surgery with a history of corneal refractive surgery.
No additional relevant studies were identified in the update searches undertaken at the end of the guideline development process.
7.1.3.1. Description of included studies
Details of the included studies are found in the evidence tables (see Appendix C).
7.1.3.1.1. Ultrasound (immersion and contact) and optical biometry to measure axial length
The 5 RCTs including a total of 588 participants (629 eyes; range n=40 to 200) were carried out in England (Rajan et al., 2002), Australia (Raymond et al., 2009), Croatia (Kolega et al., 2015), Brazil (Fontes et al., 2011) and Malaysia (Naicker et al., 2015). Only 1 trial included multiple eyes per participant (Fontes et al., 2011). Baseline characteristics of participants across all studies included mean ages ranging from 67 to 74 years (only age range of 60 to 84 years was reported by Kolega et al., 2015), similar distributions of male and female (57% to 60% female were reported in 4 studies; not reported by Rajan et al., 2002) and mean axial lengths ranging from 23.22mm to 23.45mm (reported in 3 studies; Naicker et al., 2015 specifically excluded people with axial lengths <20mm or >25mm while Kolega et al., 2015 provided no details of this characteristic). With the exception of the study conducted by Raymond et al. (2009), the other 3 trials specifically excluded participants with ocular pathologies that may result in poor visual prognosis. Only Naicker et al. (2015) provided information on specific diagnosis using the Lens Opacities Classification System III (LOCS III), while Raymond et al. (2009) provided details of the types of cataracts that were observed in the sample.
Four trials randomised participants to partial coherence laser interferometry (IOLMaster; Fontes et al., 2011; Kolega et al., 2015; Rajan et al., 2002; Raymond et al., 2009), while Naicker et al. (2015) examined optical low-coherence reflectometry (Lenstar) in its optical biometry group. Two studies examined immersion ultrasound biometry (Fontes et al., 2011; Naicker et al., 2015), while the other 3 trials focused on applanation or contact ultrasound biometry (Kolega et al., 2015; Rajan et al., 2002; Raymond et al., 2009). Only Kolega et al. (2015) did not provide details of the preoperative assessments/assessors. The remaining 4 RCTs highlighted that the persons undertaking the biometry were experienced, with only Naicker et al. (2015) quantifying the years of experience as a clinical technician (4 years); other studies specified experienced biometrist (Rajan et al., 2002), experienced ophthalmologist (Fontes et al., 2011) and senior orthoptist (Raymond et al., 2009). With the exception of the study conducted by Raymond et al. (2009), these other 3 trials used the same individual to assess both biometry techniques.
Keratometric measurements were standardised in 2 studies (Naicker et al., 2015; Raymond et al., 2009). Rajan et al. (2002) and Kolega et al. (2015) used the Javal keratometer and Righton Speedy-K type automated keratometer respectively for the ultrasound group only, while Fontes et al. (2011) did not provide any details of keratometric measurements. Four studies used the same formula for both biometry techniques (Hoffer Q – Naicker et al., 2015; SRK-T and same intraocular lens (IOL) constant – Rajan et al., 2002; Holladay I – Fontes et al., 2011, Holladay II – Kolega et al., 2015), while Raymond et al. (2009) used the SRK-T formula and manufacturer-recommended constant for the optical group and the SRK-II formula and IOL manufacturer-recommended constant for the ultrasound group. Four studies did not provide any details of IOL constant selection and/or optimisation (Fontes et al., 2011; Kolega et al., 2015; Naicker et al., 2015; Rajan et al., 2002).
All phacoemulsification cataract surgery was undertaken by the same surgeon for 3 studies, while 2 and 12 different surgeons performed operations in the studies conducted by Kolega et al. (2015) and Raymond et al. (2009) respectively. Postoperative refractive assessment varied from up to 2 weeks (Fontes et al., 2011), 5 weeks (Raymond et al., 2009), 6 weeks (Kolega et al., 2015) and 2 months (Naicker et al., 2015; Rajan et al., 2002). Only 2 studies provided details of the methods employed to assess postoperative refraction: autorefractor confirmed with subjective refraction (Rajan et al., 2002) and mixture of subjective refraction and autorefractor conducted by community ophthalmologists and optometrists as per standard practice (Raymond et al., 2009).
The quality of the evidence ranged from very low to low (see Appendix D for the GRADE tables and Appendix E for the forest plots).
7.1.3.1.2. Keratometry (manual and automated) and topography to measure corneal curvature
One RCT comparing standard keratometry (details not provided) and topography (3mm zone keratometric equivalent readings using the Eyesys Corneal Analysis System) in 46 participants (46 eyes) undergoing phacoemulsification cataract surgery with no specified history of corneal refractive surgery was carried out in England (Antcliff et al., 1995). Individuals who had fundal lesions sufficient to reduce postoperative acuity and accuracy of refraction or were unable to undergo the keratometry techniques were excluded. Reported baseline characteristics were limited to mean age of 74 years (range 32 to 92) and proportion of women (34; 73.9%). Biometry measurements for all patients were standardised using the A-scan biometer and the SRK-II formula was used to calculate the IOL power. No further details of the preoperative assessment were provided. Two surgeons performed uncomplicated phacoemulsification cataract surgery with implantation of the same type of 5mm posterior chamber lens in the capsular bag. Postoperative refraction was carried out 3 months after surgery by a “masked” investigator but no further details were provided. The quality of the evidence was low (see Appendix D for the GRADE tables and Appendix E for the forest plots).
Two retrospective case series conducted in the USA (Canto et al., 2013) and South Korea (Kim et al., 2013) compared automated keratometry (IOLMaster) and topography (TMS or Pentacam) in a total of 80 people (93 eyes) with a history of corneal refractive surgery who had phacoemulsification cataract surgery. Kim et al. (2013) specifically included people who had corneal refractive surgery for myopia. The mean ages were 52.4 and 60 years, with a greater proportion of men included in Canto et al. (2013)’s study (n=22/33) compared with an even distribution of men and women in the Kim et al. (2013) study (22 men and 25 women). The mean duration between refractive and cataract surgery was reported by Kim et al. (2013) to be 8.67 years (SD 5.45, range 1 to 16). The mean axial length was only reported by Kim et al. (2013) to be 27.75 mm (SD 2.19). Biometry measurements were only standardised by Canto et al. (2013) using the IOLMaster, while Kim et al. (2013) used immersion ultrasound for the keratometry group and the IOLMaster for the topography group. The SRK/T formula was used for all groups in both studies, but neither study provided details of IOL constant optimisation. Uneventful phacoemulsification cataract surgery was performed by 8 surgeons with 4 IOL models in Canto et al. (2013), while 1 surgeon and 1 IOL model were reported in the study by Kim et al. (2013). Canto et al. (2013) did not provide details of the timing of the postoperative refraction assessment, while Kim et al. (2013) noted that these measurements were undertaken 2 months following surgery. The quality of the evidence was very low (see Appendix D for the GRADE tables and Appendix E for the forest plots).
7.1.4. Health economic evidence
A literature search was conducted jointly for all review questions in this guideline by applying standard health economic filters to a clinical search for cataracts (see Appendix D). A total of 4,306 references was retrieved, of which none were retained for this review question. Health economic modelling was not prioritised for this review question.
7.1.5. Evidence statements
7.1.5.1. Ultrasound (immersion and contact) and optical biometry to measure axial length
Low-quality evidence from 5 RCTs containing 588 participants found no statistically significant between group differences in mean absolute prediction errors for ultrasound (including separate subgroup analyses for immersion and contact) compared with optical biometry in people undergoing phacoemulsification cataract surgery. Similarly, no statistically significant between group differences were observed in the proportion of individuals achieving postoperative refraction within various predicted ranges (<0.50 dioptres, <1.00 dioptre, <1.50 dioptres and <2.00 dioptres). Both the ultrasound and optical biometry groups demonstrated similar levels of achieving the standard benchmarks for individuals attaining a final spherical equivalent within 1.00 dioptre of the predicted refraction (90.7% with ultrasound biometry vs. 93.6% with optical biometry) and within 0.50 dioptres (68.2% with ultrasound biometry vs. 72.7% with optical biometry).
7.1.5.2. Keratometry (manual and automated) and topography to measure corneal curvature
Very low-quality evidence from 1 RCT containing 46 participants found no statistically significant between group differences in mean absolute prediction errors for standard keratometry compared with corneal topography in people undergoing phacoemulsification cataract surgery. Statistically significant between group differences were observed in the proportion of individuals achieving postoperative refraction within 0.50 dioptres of the predicted refraction (34.8% with standard keratometry vs. 69.6% with corneal topography).
Overall, very low-quality evidence from 2 retrospective case series containing 186 participants showed smaller mean prediction errors and/or greater proportions of individuals within 0.50 dioptres of the predicted refraction in the topography group compared with the automated keratometry group in people undergoing phacoemulsification cataract surgery with a history of corneal refractive surgery. However, the direction of effect and/or whether statistically significant between group differences were observed depended upon the type of topography machine (e.g. Scheimpflug or Orbscan), topography reading (e.g. true net corneal power, equivalent K, 2.0mm or 4.0mm diameter central zone of the total mean power, simulated K), formulas (e.g. SRK-T, Haigis-L, American Society of Cataract and Refractive Surgery estimation) and point estimate (e.g. mean prediction errors, mean absolute prediction errors) used.
7.1.5.3. Health economic evidence
No health economic evidence was identified for this review question.
7.1.6. Evidence to recommendations
| Relative value of different outcomes | The Guideline committee agreed that the critical outcome for decision making was deviation from predicted refractive outcome, while resource use and costs were considered to be important. The committee noted that tolerances in axial length and corneal curvature measurement and formulas may impart a total refractive error of up to 1.00 dioptre. The committee noted that axial length is a major contributor to prediction errors such that for every 1.0mm measurement error, 3.00 dioptres refractive outcome error is introduced. However, the ratio for keratometry is 1:1, such that for every 1.00 dioptre corneal curvature error, 0.90 to 1.00 dioptres refractive outcome error is introduced. |
|---|---|
| Trade-off between benefits and harms | The committee noted that optical biometry is commonly used in routine NHS standard practice as it is user-friendly, convenient, fast, does not require direct contact with the individual’s eye and generates the results immediately. In addition, commonly used optical biometry machines have the capability of providing both axial length and keratometry measurements so additional corneal curvature measuring devices are not required. However, the committee noted that optical biometry is not appropriate in some individuals, for example, those with dense cataracts, and in those cases ultrasound biometry becomes necessary. The committee noted that in current UK practice, optical biometry machines may be used to measure keratometric readings, even in these situations where ultrasound biometry is required to measure axial lengths. In contrast, ultrasound biometry procedures are more complicated, requiring experienced technicians to minimise measurement errors resulting from for example, excessive corneal indentations that artificially shortens the length of the eye, or off-axis readings. Contact ultrasound biometry also requires an anaesthetic to be administered with a small risk of infection and abrasion, while immersion ultrasound biometry requires an eye water bath. However, the committee noted that ultrasound biometry is convenient as the machine is portable and therefore can be useful in tandem with handheld keratometers for individuals with limited mobility or reduced ability to comply (for example, reduced cognitive function). The committee noted that owing to the limited availability of expertise in ultrasound biometry in the NHS, there may be delays in undertaking the assessment and obtaining the results, particularly if the individual has to be referred to another centre. The committee highlighted that ultrasound and optical biometry may give different results, and this needs to be taken in to account when calculating the intraocular lens power. No statistically significant differences in absolute prediction errors were observed for ultrasound and optical biometry, irrespective of the type of ultrasound biometry (although the committee recognised that specific studies only comparing immersion and contact ultrasound biometry were excluded). The committee also noted that both ultrasound and optical biometry showed proportions of individuals exceeding the standard benchmarks for attaining a final spherical equivalent within 0.50 dioptres and 1.00 dioptre of the predicted refraction. The committee noted that automated keratometry is currently used in NHS standard practice to assess corneal curvature measurements in routine cataract surgery patients with regular corneas. However, keratometry may not be appropriate for some individuals. Therefore, corneal topography is a useful adjunct in patients with irregular corneas or a history of corneal refractive surgery. The committee also noted that corneal topography may be useful in circ*mstances where the cornea is abnormally flat (<41.00 dioptres) or steep (>47.00 dioptres) or if there is significant astigmatism (delta K >2.50 dioptres) to assist in planning of incision techniques. The committee agreed that there is a significant cost attached to the machinery for corneal topography, particularly in light of its relatively infrequent use in biometry. Moreover, a high level of skill is required to undertake corneal topography, the equipment may not be available in all ophthalmology departments and measurements take longer, requiring expertise in interpreting the data. The committee also noted that machines and techniques measure different points on the eye and use different readings. |
| Consideration of health benefits and resource use | For optical biometry, the main cost relates to the cost of the equipment, which is already in situ in NHS clinics. In addition, cost is offset by the volume of use and throughput, and the lower staff time and experience required. For ultrasound biometry, the main costs relate to higher staff time and experience required, because ultrasound biometry equipment is widely available in all departments. Access to ultrasound biometryexperienced technicians is rapidly declining as optical biometry becomes ubiquitous. The impact of this is resource limitation and the need to refer individuals to clinics that still have staff with the expertise to undertake ultrasound biometry. The committee agreed that ultrasound biometry should be available and able to be used in ophthalmic units to ensure that people with physical or cognitive impairment are not unfairly disadvantaged because they cannot travel to another unit where the service is offered. The committee noted that the evidence indicates that both ultrasound and optical biometry are effective. However, given the practical advantages of optical biometry, the most efficient way of implementation is as currently observed in standard NHS practice, where optical biometry is routinely used, and ultrasound biometry used in special circ*mstances. In spite of this, the committee agreed that it was important to maintain competence in ultrasound biometry within the NHS, for use in situations where optimal biometry is either not practical to do, or does not provide accurate results. The committee noted that the main costs attached to corneal topography is in the acquisition cost of the machine and the requirement for highly skilled and experienced staff to operate the equipment and interpret the results. |
| Quality of evidence | The committee noted that only 5 relevant randomised controlled trials were identified for the comparison on ultrasound vs. optical biometry. It noted that there was a larger body of evidence consisting of comparative case series that may provide further evidence for this comparison, but agreed that given the potential confounding factors of the observational studies, the best study design to consider the effectiveness of the different biometry techniques was the randomised controlled trial. The committee agreed that the overall quality of evidence was low because of the risk of bias associated with the limited reporting in the studies, and the lack of generalisability on the use of ultrasound biometry in the studies compared with standard NHS clinical practice. It noted that all studies used 1 experienced practitioner/technician to undertake all the ultrasound biometry measurements and therefore, inter-observer reliability would not be captured. It agreed that the accuracy and reliability of ultrasound biometry is heavily dependent on technicians’ experience and therefore was not confident that the observed findings would be reproducible in current NHS clinical practice, where ultrasound biometry is no longer routinely used, which has implications on staff training and expertise. The committee discussed the evidence and noted that the randomisation methods used in Fontes et al. (2011) were unclear and that the 2 groups were of very different sizes, suggesting the possibility of biased allocation. The minimum age in the optical biometry group was reported to be 11 years and while it was likely to be different pathology (e.g. congenital cataracts), there was agreement that this would have little impact on this particular review question as the eye at that age is at a mature size. Moreover, it was noted that the overall mean age for the optical biometry group was 70 years with a small standard deviation, suggesting that it is likely to be 1 or 2 outliers, which should not considerably affect the results. The committee also noted that the study applied the Holladay I formula which is not considered optimal in current UK practice but agreed that since both biometry groups used the same formula, the overall findings should not be affected. The committee discussed the issue of confounding with non-standardised keratometry, given that keratometric readings are also required in intraocular lens formulas. Rajan et al. (2002) did not undertake standardised keratometry and Fontes et al. (2011) did not report any details on keratometry measurements. The committee noted the generally small studies (1 randomised controlled trial and 2 retrospective case series) that were identified for the comparison on keratometry vs. topography. It agreed that the evidence was very low quality. Specifically for the randomised controlled trial, it noted the high risk of bias from the lack of reporting of specific methods, large imprecision in the point estimates and the limited generalisability given that the study was published in 1995 such that clinical practice, keratometry and topography technology have progressed. The committee discussed the evidence from the 2 retrospective studies that included the specific subgroup of individuals with a history of corneal refractive surgery undergoing phacoemulsification cataract surgery. The committee agreed that the evidence was very low quality noting its retrospective nature, and that practice may have changed over time. In addition, the committee agreed that mixed populations containing individuals with different types of refractive surgeries (e.g. laser-assisted in situ keratomileusis, photorefractive keratectomy, radial keratotomy) for varying indications (e.g. myopia, hyperopia) should not be pooled as different surgical techniques would impact upon measurements due to altered corneal shape and stability of keratometry (e.g. individuals with a history of radial keratotomy have diurnal fluctuations in corneal curvature measurements). Moreover, the indication of surgery would typically determine the appropriate intraocular lens formula that should be used. The committee also noted the variability in observed effect depending upon the type of topography machine, topography reading, formulas and point estimate used in the analysis. It agreed that it was difficult to determine the effectiveness of keratometry vs. topography given these confounding issues. The committee also noted that there was variation between the studies in the intraocular lens formulas and constants that were used. However, because the techniques used were the same within each study (and therefore comparative data from a study are done using a consistent technique), the committee did not believe this was likely to be a source of considerable bias. As a result of the particular poor quality evidence base on the optimal biometry techniques in people who have had previous corneal refractive surgery, the committee agreed it was appropriate to make a research recommendation for this group of patients. |
| Other considerations | The committee noted that there is no true gold standard for biometry (axial length) and keratometry (corneal curvature), but agreed that all instruments should undergo calibration checks as per manufacturer’s recommendations. The committee agreed that, in patients with a history of corneal refractive surgery, the specific machine used to measure corneal curvature is less relevant than choosing the most appropriate and effective method. Because of the wide range of methods offered to estimate the corneal power used to calculate intraocular lens power following corneal refractive surgery, it is general practice to use a consensus of several methods to obtain an average. The predictability of cataract outcome after corneal refractive surgery is less than that in previously untreated eye and the patients should be counselled accordingly preoperatively. The committee emphasised the importance of personalisation based on specific equipment and techniques used and other related issues such as, surgeon factors. The committee noted that as part of routine practice, both eyes are normally assessed in the same visit to validate biometry readings. It noted that although optical biometry readings are directly transferred by some instruments into intraocular lens calculation programmes, there is a possibility of transcription errors for both techniques depending on the operating protocols used in individual clinics. |
7.1.7. Recommendations
- 8.
Use optical biometry to measure the axial length of the eye for people having cataract surgery.
- 9.
Use ultrasound biometry if optical biometry:
is not possible or
does not give accurate measurements.
- 10.
Use keratometry to measure the curvature of the cornea for people having cataract surgery.
- 11.
Consider corneal topography for people having cataract surgery:
who have abnormally flat or steep corneas
who have irregular corneas
who have significant astigmatism
who have had previous corneal refractive surgery or
if it is not possible to get an accurate keratometry measurement.
7.1.8. Research recommendation
- 3.
What is the effectiveness and cost effectiveness of biometry techniques in adults undergoing phacoemulsification cataract surgery with a history of corneal refractive surgery?
Why this is important
The number of individuals undergoing corneal refractive surgery is increasing, and a significant number of these individuals will eventually develop age-related cataracts. The corneal changes resulting from different types of refractive surgeries provide a challenge in undertaking accurate biometry assessments, and may result in worse visual outcomes of surgery in this population compared with people without prior corneal refractive surgery. Robust evidence from randomised controlled trials is needed to inform the appropriate techniques that should be used in undertaking biometry including equipment, readings and formulas.
7.2. Intraocular lens formulas
7.2.1. Review question
What are the most appropriate formulas to optimise intraocular lens biometry calculation?
7.2.2. Introduction
The evolution of theoretical intraocular lens (IOL) formulas, based on geometrical optics, is universally accepted as an essential factor contributing to the improvement of predictability of the refractive outcome with modern cataract surgery. Implicit to the third generation formulas is the variation of the effective lens position (ELP), previously referred to as anterior chamber depth (ACD), with corneal power and, in particular, the axial length of the patient’s eye. Fourth generation formulas such as Olsen and Holladay II have further improved ELP accuracy by adding variables including lens thickness. Parallel with refinement of IOL formulas has been improvement of biometry measurements, particularly axial length, with devices employing infra-red laser interferometry such as the ‘IOLMaster’ and ‘Lenstar’.
In 2010, the Royal College of Ophthalmologists published cataract surgery guidelines recommending the most appropriate IOL formulas, available at that time, for given axial length. Although these guidelines were widely acknowledged, the National Biometry audit demonstrated lack of awareness of and poor compliance with these recommendations, and also emphasised the importance of customising A constants (a measure of lens power) to minimise prediction error.
Increasingly, patients undergoing cataract surgery are likely to have a history of corneal refractive laser surgery such as laser-assisted in situ keratomileusis (LASIK) and laser-assisted sub-epithelial keratomileusis (LASEK). This is important because such surgeries alter the relationship between the anterior and posterior corneal curvature and thereby renders inaccurate the basic assumptions regarding the power of the central cornea in IOL formulas. As a result, there is a risk of unpredictable under correction of the corneal power in people with myopia, which will result in the eye being hyperopic after cataract surgery.
The aim of this review was to determine the most appropriate IOL formulas that should be used in different circ*mstances in order to optimise intraocular lens calculation. The Guideline committee prioritised the following circ*mstances:
‘Virgin’ eyes without a history of corneal refractive surgery within various ranges of axial lengths, categorised (RCOphth, 2010) into:
Short: less than 22.00mm
Average length: 22.00 to 24.50mm
Medium long: 24.50 to 26.00mm
Very long: more than 26.00mm
People with a history of corneal refractive surgery, categorised into:
Refractive error: myopia vs. hypermetropia
Surgical procedure: laser-assisted in situ keratomileusis (LASIK), laser-assisted subepithelial keratomileusis (LASEK), photorefractive keratectomy (PRK) vs. radial keratotomy (RK)
The review focused on identifying studies that fulfilled the conditions specified in Table 13. For full details of the review protocol, see Appendix C. The main outcome for this review question was the predictive accuracy of the different IOL formulas, assessed by deviations from the predicted refractive outcome expressed as a spherical equivalent. As suggested by Gale et al. (2009), a benchmark standard of 85% of individuals achieving a final spherical equivalent within 1.00 dioptre of the predicted refraction and 55% of individuals within 0.50 dioptres was used to evaluate the clinical relevance of the review findings.
Table 13
PICO inclusion criteria for the review question on intraocular lens formulas.
No relevant randomised controlled trials (RCTs) comparing different IOL formulas in adults undergoing phacoemulsification cataract surgery to predict the accuracy of postoperative refraction were identified. Papers were excluded if they:
were guidelines/health technology assessment reports, narrative reviews, case studies/reports/series, reliability studies, diagnostic accuracy studies, non-comparative studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
focused on combination surgical procedures – that is, cataract surgery in tandem with other surgical procedures (for example, phacotrabeculectomy, canaloplasty, Descemet’s stripping automated endothelial keratoplasty)
did not provide adequate information to assess the status of ocular comorbidities or previous ocular surgeries
did not provide separate subgroup data of axial lengths in virgin eyes
were not published in the English language.
For flow charts of study inclusion and exclusion, see Appendix K. For the list of excluded studies with reasons, see Appendix F.
Protocol deviation
Given that no relevant RCTs were identified, the search was expanded to include comparative observational studies. Eighteen relevant observational studies that compared the predictive accuracy of different IOL formulas in a range of axial lengths of virgin eyes undergoing phacoemulsification cataract surgery were identified. Six observational studies in eyes with a history of corneal refractive surgery were included. Since these studies were in the form of intra-person comparisons (where every tested formula was calculated for each individual in the study), it was agreed the usual concerns associated with using nonrandomised data were not relevant here, and therefore observational studies were started as being high-quality evidence in the GRADE framework, and downgraded from that point.
7.2.3. Evidence review
In total, 18,080 references were found for a combined database search for all 4 related review questions on biometry and postoperative refractive errors, with 315 articles ordered for full-text review. Fourteen observational studies on virgin eyes undergoing phacoemulsification cataract surgery were included (Aristodemou et al., 2011; Bang et al., 2011; Carifi et al., 2015; Day et al., 2012; El-Nafees et al., 2010; Eom et al., 2014; Mitra et al., 2014; Moschos et al., 2014; Percival et al., 2002; Petermeier et al., 2009; Srivannaboon et al., 2013; Tsang et al., 2003; Wang and Chang, 2013; Wang et al., 2011). Six comparative observational studies on eyes with a history of corneal refractive surgery undergoing subsequent phacoemulsification cataract operations were included. All studies included people with myopic LASIK/LASEK or PRK (Fam and Lim, 2008; Huang et al., 2013; Kim et al., 2013; Saiki et al., 2013; Savini et al., 2010; Xu et al., 2014). The formulas used for eyes with prior corneal refractive surgery were categorised as historical data methods (where information on patient history is used as part of the calculation) and no historical data methods (where patient history is not used as part of the calculation).
At the update searches, 13 full text articles were evaluated and 4 comparative case series on virgin eyes undergoing phacoemulsification cataract surgery were included (Cooke and Cooke, 2016; Doshi et al., 2017; Kane et al., 2016; Ozcura et al., 2016).
7.2.3.1. Description of included studies
Details of the included studies are found in the evidence tables (see Appendix E).
7.2.3.1.1. Virgin eyes without a history of corneal refractive surgery
Of the 18 identified studies, 17 provided usable data (exception Tsang et al., 2003). All studies were comparative case series, 15 retrospective and 3 prospective (Doshi et al., 2017; El-Nafees et al., 2010; Srivannaboon et al., 2013). All studies with the exception of Petermeier et al. (2009) stated that the phacoemulsification cataract surgery was uneventful or people with intraoperative/postoperative complications had been excluded. Table 14 provides a summary of the key study characteristics.
Table 14
Summary of key characteristics of included studies for virgin eyes without a history of corneal refractive surgery.
7.2.3.1.2. People with a history of corneal refractive surgery
All studies were comparative case series, 5 retrospective and 1 prospective (Huang et al., 2013) including eyes with a history of myopic LASIK/LASEK/PRK. All studies with the exception of Fam and Lim (2008) stated that the phacoemulsification cataract surgery was uneventful. Table 14 provides a summary of the key study characteristics.
Table 15
Summary of key characteristics of included studies for eyes with a history of myopic LASIK/LASEK/PRK.
7.2.4. Health economic evidence
A literature search was conducted jointly for all review questions in this guideline by applying standard health economic filters to a clinical search for cataracts (see Appendix D). A total of 4,306 references was retrieved, of which none were retained for this review question. Health economic modelling was not prioritised for this review question.
7.2.5. Evidence statements
7.2.5.1. Virgin eyes without a history of corneal refractive surgery
7.2.5.1.1. Axial lengths less than 22.00mm
Evidence from 5 network meta-analyses and 1 pairwise comparison including data from up to 11 case series showed that the SRK/T formula had the lowest predictive accuracy for intraocular lens power calculations as assessed by mean absolute error and proportion of people achieving predicted target within 0.25, 0.50, 1.00 and 2.00 dioptres. Haigis and Hoffer Q formulas showed the highest predictive accuracy with lowest imprecision as assessed by mean absolute error and proportion of people achieving predicted target within 0.25, 0.50 and 2.00 dioptres. The overall quality was assessed to be very low to moderate (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.1.2. Axial lengths 22.00–24.50mm
Evidence from 5 network meta-analyses including data from up to 4 case series showed that the Barrett Universal II and SRK/T formulas were similarly effective in terms of predictive accuracy of intraocular lens power calculations as assessed by mean absolute error and proportion of people achieving predicted target within 0.25, 0.50, 1.00 and 2.00 dioptres. The overall quality was assessed to be moderate to high (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.1.3. Axial lengths 24.50–26.00mm
Evidence from 5 network meta-analyses including data from up to 6 case series showed that the Barrett Universal II formula was most effective in terms of predictive accuracy of intraocular lens power calculations as assessed by the proportion of people achieving predicted target within 0.25, 0.50, 1.00 and 2.00 dioptres. SRK/T formula was effective in terms of predictive accuracy of intraocular lens power calculations as assessed by the proportion of people achieving predicted target within 2.00 dioptres. The overall quality was assessed to be low to moderate. The overall quality was assessed to be very low to high (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.1.4. Axial lengths greater than 26.00mm
Evidence from 5 network meta-analyses and 1 pairwise comparison including data from up to 8 case series showed that the SRK/T and Haigis formulas had the highest predictive accuracy for intraocular lens power calculations as assessed by the proportion of people achieving predicted target within 0.25, 0.50 and 2.00 dioptres. The overall quality was assessed to be low to moderate (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.2. Eyes with a history of myopic LASIK/LASEK/PRK
7.2.5.2.1. Historical and no historical data methods
Evidence from 5 network meta-analyses and 1 pairwise comparison including data from up to 5 case series showed that it was not possible to distinguish between the formulas tested, as assessed by mean absolute error, mean prediction error, or the proportions of people achieving predicted target within 0.50 and 1.00 dioptres. The Haigis-L formula was more effective than the SRK/T formula, as assessed by the proportions of people achieving predicted target within 1.50 and 2.00 dioptres. The overall quality was assessed to be very low to low (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.2.2. No historical data methods
Evidence from 4 network meta-analyses including data from 4 case series showed that it was not possible to distinguish between the formulas tested, as assessed by the proportions of people achieving predicted target within 0.50 and 1.00 dioptres. The overall quality was assessed to be very low (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.2.3. Historical data methods
Data from 2 network meta-analyses and 1 pairwise comparison including data from up to 2 case series showed that it was not possible to distinguish between the formulas tested, as assessed by mean absolute error or the proportion of people achieving predicted target within 2.00 dioptres. The SRK/T formula had the lowest predictive accuracy, as assessed by the proportion of people achieving predicted target within 0.50 and 1.00 dioptres. The overall quality was assessed to be very low (see Appendix G for the GRADE tables and Appendix H for the results of the network meta-analyses).
7.2.5.3. Health economic evidence
No health economic evidence was identified for this review question.
7.2.6. Evidence to recommendation
| Relative value of different outcomes | The Guideline committee agreed that the critical outcome for decision-making was deviation from predicted refractive outcome, though resource use and costs were also considered to be important. The committee agreed that, of the different refractive outcomes presented, the proportion of people with 0.5 dioptres was likely to be the most relevant, as this was the clinically relevant outcome for which the largest amount of data was available (smaller errors are unlikely to have a meaningful impact on patients’ vision; larger ones are uncommon events regardless of formula). The committee highlighted the importance of selecting appropriate intraocular lens (IOL) formulas depending on the axial length of the eye and in specific circ*mstances where a history of corneal refractive surgery are likely to impact upon the shape of the cornea, such that resulting keratometry and/or topography measurements would require adjustments when applied to standard formulas. |
|---|---|
| Trade-off between benefits and harms | The committee agreed that the SRK/T formula was the most appropriate to use as the reference category in the analyses (where it was available), as it was the formula used most commonly across the different trials and outcome measures, and is one that is in use in clinical practice. The committee noted that individual IOL formulas used a range of variables in addition to IOL constants, with the simplest including only 2 measurements, that is, axial length and keratometric reading (for example, Hoffer Q, SRK/T), while others included 7 variables (for example, Holladay 2 uses axial length, keratometry, preoperative anterior chamber depth and refraction, lens thickness, age and horizontal white-to-white measurement). However, the committee agreed that these formulas were comparable when considered as complex interventions and noted that all required measurements including those for the Holladay 2 (with the exception of lens thickness) can be obtained using standard modern biometry machines. The committee noted the recently published Super Formula which uses other formulas (Haigis, Hoffer Q, Holladay 1 with and without Wang-Koch adjustment). The committee noted the general high levels of statistical imprecision observed across all the formulas and outcomes. For eyes without a history of corneal refractive surgery, the main results of the evidence synthesis were that the SRK/T formula performs poorly in eyes with short axial lengths (those less than 22.00mm) in contrast to eyes with very long axial lengths (those greater than 26.00mm), and the Hoffer Q performs poorly in eyes with very long axial length (greater than 26.00mm). The Haigis formula was among the best options for 3 of the 4 axial length subgroups. Eyes with short axial lengths For eyes with short axial lengths, the Hoffer Q formula was similarly effective to the Haigis in predictive accuracy. Barrett Universal II and SRK/T formulas were the best options for eyes with average or medium long axial lengths. While several newer formulas showed trends towards better predictability, the committee was hesitant to recommend these formulas because of the high levels of statistical imprecision and small study samples. Therefore, it agreed it would be more appropriate to make a research recommendation looking at the effectiveness of these newer formulas in larger studies. The committee noted that, for eyes with a history of corneal refractive surgery, the absolute levels of prediction error were worse than in eyes without previous surgery. For example, across all studies and formulas, in the non-surgery group, for axial lengths between 22.00 and 24.50mm, 70.1% of the prediction errors were less than 0.50 dioptres, while in eyes with prior surgery, only 31.1% of the prediction errors were less than 0.50 dioptres. The committee therefore agreed it was appropriate to make a research recommendation looking at the most appropriate formulas to use in people with prior corneal refractive surgery. Eyes with a history of corneal refractive surgery For eyes with a history of corneal refractive surgery, it was not possible to identify formulas that provided consistently better results than others, as there was considerable uncertainty and heterogeneity in the evidence base. The formulas used across multiple studies produced very different levels of accuracy in different studies, and the committee was not able to identify aspects of the study design or patient population that would adequately explain these levels of heterogeneity. The committee noted, however, that there was a pattern of formulas which did make adjustments performing better than those based on clinical history alone, implying that making an adjustment is better than not doing so, even if it was not possible to recommend which particular adjustment should be made. The committee also agreed that, given the clear evidence that predictions were less accurate in this group, this information should be communicated to patients before surgery, to ensure they are fully informed and have realistic expectations of the benefits they are likely to receive from surgery. |
| Consideration of health benefits and resource use | No health economic evidence was found for this review question, and it was not prioritised for de novo modelling work. However, the committee noted that various IOL formulas are available as a standard package within more recent biometry machines (which the committee confirmed were widely in use), but some of the newer formulas may require additional proprietary licenses, although this does not apply to those formulas which are recommended here. The committee did not consider the recommendations made would have significant resource implications. |
| Quality of evidence | The committee noted the lack of randomised controlled trials examining the effectiveness of different IOL formulas. The committee agreed that it may have been useful to consider narrower ranges of axial lengths in order to identify critical thresholds for the appropriate use of different IOL formulas. However, the committee noted that only 1 large UK-based study provided this level of detailed evidence (Aristodemou et al. 2011) for the Hoffer Q and SRK/T formulas, and that the reported findings for axial length subgroups in increments of 0.5 to 1.0mm were congruent with the overall network meta-analysis results observed for the 4 prioritised axial length classes. The committee also highlighted that focusing on narrower bands of axial lengths would impact upon the statistical power and precision of the findings. The committee agreed that strict selection criteria excluding studies that did not specify phacoemulsification cataract surgery or did not provide adequate information to assess the status of ocular comorbidities or previous ocular surgeries and/or separate subgroup data of axial lengths in virgin eyes were necessary to ensure that the included studies were adequately hom*ogeneous to be included in a network meta-analysis. However, the committee recognised that this meant 2 specific papers that have been relied on in other guidelines were excluded (MacLaren et al. 2007 and Narvaez et al. 2006). The committee noted that the sensitivity analyses based on the type of biometry undertaken and the use of IOL constant optimisation also showed little variation compared with the overall findings of all included studies. The committee agreed that the overall quality of evidence was very low to moderate, and noted that the evidence for people with prior corneal refractive surgery was of particularly low quality, consisting mainly of small retrospective studies (and with no evidence at all in eyes post radial keratotomy). The committee also noted that the formulas assessed in the included papers had all been derived from retrospective analyses, and none had been subject to prospective testing. The committee noted that, in some analyses, the ordering of effectiveness of the interventions differed between the analyses looking at mean absolute error and those looking at the proportion of people within 0.5D. They agreed this was likely to be because the mean difference results were being skewed by a small proportion of people having very large errors in prediction. The committee agreed the within 0.5D evidence was more appropriate for decision making, as once an error reaches a certain level the clinical outcome (of lens explantation and new lens insertion) is the same, regardless of the magnitude of the error. |
| Other considerations | The committee agreed that, given the lack of distinction in predictive accuracy of different IOL formulas for axial lengths ranging from 22.00 to 24.50mm and 24.50 to 26.00mm, it would be useful to group these bandings in the recommendations. |
7.2.7. Recommendations
- 12.
For people who have not had previous corneal refractive surgery, use 1 of the following to calculate the intraocular lens power before cataract surgery:
If the axial length is less than 22.00 mm, use Haigis or Hoffer Q.
If the axial length is between 22.00 and 26.00 mm, use Barrett Universal II if it is installed on the biometry device and does not need the results to be transcribed by hand. Use SRK/T if not.
If the axial length is more than 26.00 mm, use Haigis or SRK/T.
- 13.
Advise people who have had previous corneal refractive surgery that refractive outcomes after cataract surgery are difficult to predict, and that they may need further surgery if they do not want to wear spectacles for distance vision.
- 14.
If people have had previous corneal refractive surgery, adjust for the altered relationship between the anterior and posterior corneal curvature. Do not use standard biometry techniques or historical data alone.
7.2.8. Research recommendations
- 4.
How effective are newer intraocular lens formulas (for example, Barrett, Olsen, T2) compared with standard formulas for phacoemulsification cataract operations on eyes without a history of corneal refractive surgery, especially for long and short axial lengths?
Why this is important
Appropriately applied intraocular lens (IOL) formulas are paramount to improving predictive accuracy and patient satisfaction following cataract surgery and IOL implantation. Despite significant technological advancement in ophthalmology, it is widely recognised that many of the currently used IOL formulas were developed more than 20 years ago. Newer formulas are being published but there is a dearth of evidence comparing their effectiveness to standard formulas in people without a history of corneal refractive surgery. Methodologically robust randomised controlled trials are needed to address this research gap.
- 5.
What is the effectiveness of different intraocular lens formulas for eyes after prior corneal refractive surgery, as measured in a prospectively collected multi-centre study?
Why this is important
Appropriately applied intraocular lens (IOL) formulas are paramount to improving predictive accuracy and patient satisfaction following cataract surgery and IOL implantation. There are particular challenges in accurate prediction in people with a history of corneal refractive surgery, and there is a lack of evidence for the most effective formulas to use in this group, with a total absence of large, prospective studies. Methodologically robust randomised controlled trials are needed to address this research gap.
7.3. Intraocular lens constant optimisation
7.3.1. Review question
What is the effectiveness of strategies used to select intraocular lens constants in order to optimise biometry calculation?
7.3.2. Introduction
The aim of this review was to determine the effectiveness of strategies used to select intraocular lens (IOL) constants in order to optimise biometry calculation.
The review focused on identifying studies that fulfilled the conditions specified in Table 16. For full details of the review protocol, see Appendix C. The main outcome for this review question was the predictive accuracy of the different optimisation strategies, assessed by deviations from the predicted refractive outcome expressed as a spherical equivalent. As suggested by Gale et al. (2009), a benchmark standard of 85% of individuals achieving a final spherical equivalent within 1.00 dioptre of the predicted refraction and 55% of individuals within 0.50 dioptres was used to evaluate the clinical relevance of the review findings.
Table 16
PICO inclusion criteria for the review question on intraocular lens constant optimisation.
No randomised controlled trials (RCTs) comparing different strategies to optimise IOL constants in adults undergoing phacoemulsification cataract surgery to predict postoperative refraction were identified. Papers were excluded if they:
were guidelines/health technology assessment reports, narrative reviews, case studies/reports/series, reliability studies, diagnostic accuracy studies, non-comparative studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
focused on combination surgical procedures that is, cataract surgery in tandem with other surgical procedures (for example, phacotrabeculectomy, canaloplasty, Descemet’s stripping automated endothelial keratoplasty)
were not published in the English language.
For flow charts of study inclusion and exclusion, see Appendix K. For the list of excluded studies with reasons, see Appendix F.
Protocol deviation
Given that no relevant RCTs were identified, the search was expanded to include comparative observational studies. Nine relevant retrospective comparative case series that compared the predictive accuracy of different IOL constant optimisation strategies in virgin eyes without a history of corneal refractive surgery undergoing phacoemulsification cataract operations were identified. Since these studies were in the form of intra-person comparisons (where both optimisation and non-optimisation were considered for each individual in the study), it was agreed the usual concerns associated with using non-randomised data were not relevant here, and therefore observational studies were started as being high-quality evidence in the GRADE framework, and downgraded from that point.
7.3.3. Evidence review
In total, 18,080 references were found for a combined database search for all 4 related review questions on biometry and postoperative refractive errors, with 315 articles ordered for full-text review. Nine observational studies on virgin eyes undergoing phacoemulsification cataract surgery were included (Aristodemou et al., 2011; Charalampidou et al., 2010; Day et al., 2012; Eom et al., 2013; Fam et al., 2009; Lee et al., 2015; Petermeier et al., 2009; Sharma et al., 2014; Wang et al., 2011).
No additional relevant studies were identified in the update searches undertaken at the end of the guideline development process.
7.3.3.1. Description of included studies
All 9 identified studies were retrospective comparative case series with sample sizes ranging from 50 to 8,108 eyes. With the exception of Petermeier et al. (2009), all studies stated that the phacoemulsification cataract surgery was uneventful or those with intraoperative/postoperative complications had been excluded. All studies used optical biometry to undertake preoperative assessments (IOLMaster in 8 studies and Lenstar in Lee et al. 2015). One study (Aristodemou et al., 2011) tailored the use of IOL formula based on the individual eye’s axial length. An extensive range of IOL constant optimisation methods was examined including the use of User Group for Laser Interference Biometry (ULIB) website to download IOL constants or personalise constants, back-calculating to achieve a prediction error of zero, using optimised axial length and/or keratometry readings, using IOL constants derived from biometry machines (IOLMaster, Lenstar), use of manufacturers’ IOL constants, traditional A constants and optimising the axial length compared with using the IOLMaster axial lengths. Further details of the reported methods for optimisation are found in the evidence tables (see Appendix E).
7.3.3.2. Evidence review strategy
Separate data for excluded IOL formulas (that is, Binkhorst II, Holladay 1, SRK I and SRK II) reported in studies were not extracted or analysed. Where a study included multiple IOLs and reported both separate data for each IOL and combined data, the individually reported IOL data were preferentially used. Where a study reported results for multiple IOL formulas, the IOL formula that is recommended for the mean axial length of that study was preferentially extracted and analysed (see section 7.2 on intraocular lens formulas). Where a study reported several versions of the optimisation method, for example, using the entire sample to calculate individualised IOL constants vs. using half the sample and extrapolating to the full population; or the use of 3 optimised constants vs. 2, the option that would more likely provide the optimal optimisation was preferentially selected, that is, total sample individualised and 3 optimised constants.
7.3.4. Health economic evidence
A literature search was conducted jointly for all review questions in this guideline by applying standard health economic filters to a clinical search for cataracts (see Appendix D). A total of 4,306 references was retrieved, of which none were retained for this review question. Health economic modelling was not prioritised for this review question.
7.3.5. Evidence statements
Evidence from 5 network meta-analyses and 1 pairwise comparison including data from up to 7 retrospective case series suggested that the use of standard IOL constants may be suboptimal in maximising the predictive accuracy of intraocular lens power calculations as assessed by mean absolute error and proportion of people achieving predicted target within 0.25, 0.50, 1.00 and 1.50 dioptres. The proportions of individuals achieving postoperative refraction within 0.50 and 1.00 dioptres were lower in groups using standard IOL constants (46.3% and 83%) compared with optimised constants (75.2% and 94.1%) in 5 and 6 low-quality retrospective case series (8,698 and 8,749 eyes) respectively. The overall quality was assessed to be low (see Appendix G for the GRADE table and Appendix H for the results of the network meta-analyses).
Evidence from 5 network meta-analyses and 1 pairwise comparison including data from up to 7 retrospective case series showed that, of the 7 different IOL constant optimisation methods assessed, none were significantly better than each other in improving predictive accuracy of intraocular lens power calculations. Two methods, surgeon’s personalisation using the Users Group for Laser Interference Biometry (ULIB) framework and optimising individual IOL constants by back-calculating the prediction error to zero showed trends of being effective in improving the proportion of eyes achieving the predicted target within 1.00 and 0.25 dioptres respectively. The overall quality was assessed to be low (see Appendix G for the GRADE table and Appendix H for the results of the network meta-analyses).
7.3.5.1. Health economic evidence
No health economic evidence was identified for this review question.
7.3.6. Evidence to recommendations
| Relative value of different outcomes | The Guideline committee agreed that the critical outcome for decision making was deviation from predicted refractive outcome, while resource use and costs were considered to be important. The committee noted that intraocular lens (IOL) constant optimisation was one of several strategies to improve postoperative refractive outcomes, involving adjustments specific to the IOL and individual surgeons. They highlighted that IOL manufacturers’ tolerance for lens accuracy is variable and can range from 0.25 to 0.40 dioptres tolerance. Surgeon variables include the size and method of insertions such that small variations can result in systematic differences in postoperative refractive outcomes. The committee emphasised the importance of personalisation based on specific biometry equipment and techniques used, multiple preoperative assessment staff and other related issues such as surgeon factors. |
|---|---|
| Trade-off between benefits and harms | The committee noted that, historically, it was difficult to obtain audit data of prediction errors that can be used to inform IOL constant optimisation, as there were no robust mechanisms for returning postoperative outcome data, particularly where patients were discharged for postoperative refraction by community optometrists. However, such data are currently much more accessible with the availability of automated biometry with electronic storage of results. The committee agreed that the time taken to submit audit information is not overly onerous and therefore it would be useful to encourage departments to undertake such practice, to facilitate quality data sets that can be used to improve the accuracy of IOL constant optimisation. The committee recognised that this practice would need to be maintained as IOLs change over time. The committee noted that, generally, UK surgeons and departments do not formally calculate optimised constants, but rather use informal processes, for example, surgical teams apply adjustments (over- or under-estimates) based on reflection of their experience, type of IOL used (for example, standard vs. multifocal) and patient preference (for example, to be over- rather than under-corrected). The committee agreed that, overall, the evidence synthesis was suggestive that, compared with standard IOL constants, optimisation of IOL constants is likely to improve the predictive accuracy of postoperative refractive outcomes. However, this finding was subject to substantial statistical uncertainty: although there was a trend towards improved accuracy in all outcomes, credible intervals from the network meta-analyses tended to be very wide and only 1 comparison produced results that satisfied conventional definitions of statistical significance (adjusting the prediction error to zero for the proportion of eyes within 0.25 dioptres of the predicted postoperative refraction). The committee understood that this uncertainty was substantially caused by statistical heterogeneity in the underlying evidence, leading to large random-effects terms in the synthesis models. In particular, it was notable that the 2 trials that examined comparable strategies for calibrating prediction error to zero (Aristodemou et al., 2011 and Day et al., 2012) gave incongruent results, with substantial and significant accuracy gains in Aristodemou et al. (2011) but not in Day et al. (2012). The committee discussed that this discrepancy may have arisen because Day et al. (2012) was a small study restricted to eyes with short axial lengths (less than 22.00mm), whereas Aristodemou et al. (2011) was a much larger study including eyes of all sizes. The effect of this discrepancy was to ‘dilute’ the strongly significant gains demonstrated in the larger, more representative study, as the synthesis models had to estimate a broad, uncertain distribution of effects in order to fit the heterogeneous data. The committee therefore concluded that Day et al. (2012) had a disproportionate effect in the network meta-analyses, and considered putting additional weight on the findings of Aristodemou et al. (2011), due to the study’s greater power and more inclusive population. For these reasons, the committee arrived at the view that surgeons should consider personalising their IOL constants. The evidence was not sufficiently unambiguous to make a firm (‘offer’) recommendation, but there was no prospect of patient harm resulting from the approach, and it should not be onerous for surgeons to incorporate this step into their audit routines (that is, the anticipated opportunity cost – in terms of surgeon time – is negligible). However, the committee agreed that no specific distinction could be made on the best optimisation strategy, given that all the credible intervals overlapped each other across all the assessed outcomes (mean absolute error and proportion of eyes within 0.25, 0.50 and 1.00 dioptres). Therefore, the committee agreed a recommendation that urges surgeons to consider personal optimisation, but leaves the specific strategy to the individual’s discretion. |
| Consideration of health benefits and resource use | No health economic evidence was found for this review question, and it was not prioritised for de novo modelling work. The committee did not consider the recommendation made would have significant resource implications. |
| Quality of evidence | The committee noted the lack of randomised controlled trials examining the effectiveness of different IOL constant optimisation strategies. They highlighted that all the identified studies were retrospective in design such that assumptions were made that the preoperative data were accurate. The postoperative refractive outcome was used to back-calculate the likely outcomes given that various optimisation strategies had been applied. The committee noted that, with the exception of 1 large UK based study, the studies were small. The committee noted that 3 studies specifically stated that an autorefractor had been used to assess the postoperative refractive outcome. This is different to clinical practice in that auto-refraction is used as a baseline measurement and does not guide lens selection/corrective lens prescription. However, the committee agreed that, due to lack of detailed reporting, it was unclear as to whether other studies had only assessed subjective refraction postoperatively. The committee noted the general lack of descriptive detail of the optimisation methods applied in most of the studies, particularly ambiguity regarding the use of the Users Group for Laser Interference Biometry (ULIB) framework, which made it difficult to implement in clinical practice. The committee noted that, in many instances, the comparator arms may have also involved the use of optimised constants (for example, IOL constants available from optical biometry machines) but, because of the limited detail provided by the studies, it was unclear whether optimisation occurred. However, it agreed that these comparator arms could be grouped together in 1 category of standard IOL constants since it was clear that an optimisation strategy was being applied in the other arms, and given the retrospective nature of the study designs, all optimisation methods were compared with the original calculations undertaken on the same optical biometry machine. While the committee recognised that various confounding factors (for example, type of IOL and IOL formulas) were kept constant within studies, and that sensitivity analyses undertaken involving the removal of the study on light-adjustable lens had not affect the overall findings, it agreed that this specific study (Conrad-Hengerer et al. 2011) should be excluded from the evidence base because refraction could not be determined as being stable or accurate at the point of measurement. The committee agreed that the remaining 9 studies were adequately hom*ogeneous to be included in a network meta-analysis and that the overall quality of evidence was low to moderate. They agreed that whilst the exclusion of participants with complications during surgery is likely to have led to overestimates in the effectiveness of biometry overall, there was no reason to believe this will have led to differences in the comparative effectiveness of the approaches. |
| Other considerations | No other considerations were identified for this review question. |
7.3.7. Recommendations
- 15.
Surgeons should think about modifying a manufacturer’s recommended intraocular lens constant, guided by learning gained from their previous deviations from predicted refractive outcomes.
7.4. Other considerations in biometry
7.4.1. Review question
What other factors should be considered such as, who should undertake biometry and when should preoperative biometry be assessed?
7.4.2. Introduction
The aim of this review was to identify other factors that should be considered to minimise the risk of biometry errors and postoperative refractive errors and in particular the following:
who should undertake biometry
when should preoperative biometry be assessed
second eye prediction refinement.
The review focussed on identifying studies that fulfilled the conditions specified in Table 17. For full details of the review protocol, see Appendix C. The main outcome for this review question was the predictive accuracy of the different methods, assessed by deviations from the predicted refractive outcome expressed as a spherical equivalent. As suggested by Gale et al. (2009), a benchmark standard of 85% of individuals achieving a final spherical equivalent within 1.00 dioptre of the predicted refraction and 55% of individuals within 0.50 dioptres was used to evaluate the clinical relevance of the review findings.
Table 17
PICO inclusion criteria for the review question on other factors.
Randomised controlled trials (RCTs) and observational studies comparing different methods of reducing the risk of biometry errors and postoperative refractive errors in adults undergoing phacoemulsification cataract surgery were included. Papers were excluded if they:
were guidelines/health technology assessment reports, narrative reviews, case studies/reports, case series with less than 10 people, reliability studies, diagnostic accuracy studies, non-comparative studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
focused on combination surgical procedures that is, cataract surgery in tandem with other surgical procedures (for example, phacotrabeculectomy, canaloplasty, Descemet’s stripping automated endothelial keratoplasty)
were not published in the English language.
For flow charts of study inclusion and exclusion, see Appendix K. For the list of excluded studies with reasons, see Appendix F.
7.4.3. Evidence review
In total, 18,080 references were found for a combined database search for all 4 related review questions on biometry and postoperative refractive errors, with 315 articles ordered for full-text review. Four unique observational studies were included (Aristodemou et al., 2011; Covert et al., 2010; Jabbour et al., 2006; Jivrajka et al., 2012), all focusing on second eye prediction refinement, that is using the first eye prediction error to adjust the intraocular lens (IOL) calculation for the second eye. Since these studies were in the form of intra-person comparisons (where every tested strategy was calculated for each individual in the study), it was agreed the usual concerns associated with using non-randomised data were not relevant here, and therefore observational studies were started as being high-quality evidence in the GRADE framework, and downgraded from that point. No relevant studies were identified for the other two listed factors, that is, staffing and timing of preoperative assessments.
No additional relevant studies were identified in the update searches undertaken at the end of the guideline development process.
7.4.3.1. Description of included studies
Details of the included studies are found in the evidence tables (see Appendix E).
7.4.3.1.1. Second eye prediction refinement
The 4 case series including a total of 2,291 participants (4,582 eyes; range n=97 to 1,867) undergoing bilateral sequential phacoemulsification cataract surgery were carried out in the UK (Aristodemou et al., 2011), USA (Covert et al., 2010; Jivrajka et al., 2012) and Germany (Jabbour et al., 2006). All but 1 study (Jivrajka et al., 2012) specifically stated that the surgery was conducted in 1 hospital. Timing between the first and second eye surgeries was not reported by Aristodemou et al. (2011), while Covert et al. (2010) reported a mean of 36.7 days, Jabbour et al. (2006) reported a median of 3 months and Jivrajka et al. (2012) provided a range of 1 to 3 months. All but 1 study (Jivrajka et al., 2012) used a retrospective design to develop and/or test various correction factors based on the first eye prediction error. One study did not report any baseline characteristics (Aristodemou et al., 2011). Two studies reported mean age, one specifically at the time of first eye surgery (69.9 years, Covert et al., 2010) and the other was unclear in terms of timing (77.57 years, Jivrajka et al., 2012). Three studies reported similar distributions of female patients (51% in Jivrajka et al., 2012 to 64% in Jabbour et al., 2006), mean axial lengths ranging from 23.15mm (Jabbour et al. 2006) to 24.0mm (Covert et al., 2010) and mean keratometric readings ranging from 43.48 dioptres (Jabbour et al., 2006) to 44.00 dioptres (Covert et al., 2010). Two studies specifically excluded people who had corneal astigmatism >3.00 dioptres (Aristodemou et al., 2011; Jabbour et al., 2006). All studies applied exclusion criteria based on concurrent procedures and/or ocular comorbidities. No studies provided information on specific diagnosis.
All but 1 study undertook biometry and keratometry measurements using the IOLMaster; Jabbour et al. (2006) used 2 ultrasound biometers and 2 identical Bausch & Lomb keratometers. Only 2 studies provided some information on the biometry assessors; Covert et al. (2010) noted that a trained ophthalmic technician carried out measurements, while Jabbour et al. (2006) highlighted that readings were taken by 2 different operators. All studies used different formulas. Aristodemou et al. (2011) used the Hoffer Q, Holladay I and SRK/T formulas based on the axial lengths of paired eyes, Covert et al. (2010) used the SRK-II and Holladay (1998) formulas, Jabbour et al. (2006) used the SRK/T and axial length vergence formulas while Jivrajka et al. (2012) used the Haigis formula.
All phacoemulsification cataract surgery was undertaken by the same surgeon for 3 studies, Aristodemou et al. (2011) did not provide any details. Only 3 studies reported timing of postoperative refractive assessment which varied from at least 4 weeks (Aristodemou et al., 2011; Covert et al. 2010) up to 8 weeks (Jivrajka et al., 2012). All but 1 study (Jivrajka et al., 2012) reported that subjective refraction was used at postoperative assessment.
All but 1 study (Jabbour et al., 2006) found that 50% was the optimal correction factor to take into consideration when applying the first eye prediction error.
The quality of the evidence ranged from very low to low (see Appendix G for the GRADE tables and Appendix H for the meta-analysis results).
7.4.4. Health economic evidence
A literature search was conducted jointly for all review questions in this guideline by applying standard health economic filters to a clinical search for cataracts (see Appendix D). A total of 4,306 references was retrieved, of which none were retained for this review question. Health economic modelling was not prioritised for this review question.
7.4.5. Evidence statements
7.4.5.1. Second eye prediction refinement
Very low-quality evidence from 1 retrospective case series of 412 people found a small statistically significant between group difference in mean absolute prediction errors in favour of the 50% adjusted 2nd eye prediction group compared with the unadjusted 2nd eye prediction group.
Statistically significant between group differences were only observed in the proportion of individuals achieving postoperative refraction within 0.50 dioptres (80.3% with 50% adjusted 2nd eye prediction vs. 73.3% with unadjusted 2nd eye prediction) in 2 low-quality retrospective case series. No statistically significant differences were observed in the proportion of individuals achieving postoperative refraction within 1.00 dioptre (3 low quality case series; 96.3% with 50% adjusted 2nd eye prediction vs. 94.7% with unadjusted 2nd eye prediction).
7.4.5.2. Health economic evidence
No health economic evidence was identified for this review question.
7.4.6. Evidence to recommendations
| Relative value of different outcomes | The Guideline committee agreed that the critical outcome for decision making was deviation from predicted refractive outcome, while resource use and costs were considered to be important. |
|---|---|
| Trade-off between benefits and harms | The committee discussed the implications of using first eye prediction error to inform calculations of intraocular lens power of the second eye in terms of adequate timing between the first and second eye surgeries to ensure that the refractive error of the first eye had stabilised. The committee noted that individuals undergoing bilateral simultaneous cataract surgery may be disadvantaged by recommending that second eye prediction is adjusted based on first eye prediction. However, the committee agreed that given the potential benefit of improved prediction of the second eye and subsequent improved patient outcomes such as satisfaction, the use of first eye prediction error to inform second eye prediction should be considered by healthcare professionals where appropriate, such as in cases where first eye prediction error does not result in ‘refractive surprise’ or require lens exchange. The committee agreed that, although the evidence base was of low quality, it did suggest that second eye prediction adjustment did lead to improved refractive outcomes, and it was highly unlikely there would be any negative outcomes that could result from doing so which would counterbalance these small gains. |
| Consideration of health benefits and resource use | No relevant health economic evidence was identified and de novo health economic modelling was not prioritised for this review question. |
| Quality of evidence | The committee noted that no relevant studies were identified to inform who should undertake biometry or when the preoperative biometry assessment should take place. The committee agreed that the evidence for the use of first eye prediction errors to inform second eye prediction refinement was generally of low quality because the majority of studies (3 out of 4) were retrospective in design, applying theoretical calculations, with no consideration of practical, clinical and individual implications such as anisometropia. However, the committee noted that the only small prospective study on 97 people showed similar evidence of beneficial effect of using 50% adjusted first eye prediction error to inform calculations of intraocular lens power of the second eye. The committee also noted that in 1 retrospective study, 50% adjusted refinement was shown to be beneficial even in situations where the intraocular lens constants in the formula were already optimised. The committee agreed that it would be useful to provide a clinical guide on the maximum threshold level of prediction error from the first eye for use in second eye prediction, in order to minimise the risk of anisometropia. However, the committee noted that the evidence reviewed did not facilitate recommendation with this detailed information. |
| Other considerations | The committee noted that currently individuals are routinely refracted postoperatively at 4–6 weeks but the outcome data are not necessarily provided to ophthalmology departments to enable consideration of adjustment for second eye surgery intraocular lens calculations. The committee noted that there is evidence to suggest that there is limited uptake of guidelines on the appropriate use of formulas, and therefore a recommendation to adjust second eye prediction based on first eye prediction errors would be useful in improving patient care. The committee noted that a range of professionals may undertake biometry but it is exceedingly important that staff are appropriately trained and experienced. |
7.4.7. Recommendations
- 16.
Consider using 50% of the first-eye prediction error in observed refractive outcome to guide calculations for the intraocular lens power for second-eye cataract surgery.
7.5. Risk stratification and risk factors for increased cataract surgical complications
7.5.1. Review questions
What is the effectiveness of risk stratification techniques to reduce surgical complications?
What are the risk factors associated with increased surgical complications in cataract surgery?
7.5.2. Introduction
The aim of this review was to determine the effectiveness of preoperative risk stratification techniques, and the identification of risk factors associated with an increase in surgical complications. The reviews for these two separate issues focused on identifying studies that fulfilled the conditions specified in Table 18 and Table 19, respectively. For full details of the review protocol, see Appendix C. The main outcomes for this review were surgical complication rates.
Table 18
PICO for effectiveness of preoperative risk stratification techniques.
Table 19
PICO for risk factors that are associated with surgical complications.
Papers were excluded if they:
were narrative reviews, case studies/reports, commentaries, editorials/letters or opinion pieces.
were studies on procedural safety surgical checklists e.g. WHO, case reports/case studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
reported studies conducted entirely in non-OECD countries
were not published in the English language.
Papers were excluded if they:
were narrative reviews, case studies/reports, commentaries, editorials/letters or opinion pieces
were studies on procedural safety surgical checklists e.g. WHO, case reports/case studies
included animals, healthy eyes, other ocular conditions besides cataracts or mixed primary populations of people with different eye pathologies
reported studies conducted entirely in non-OECD countries
were not published in the English language.
For flow charts of study inclusion and exclusion, see Appendix K. For the list of excluded studies with reasons, see Appendix F.
7.5.3. Evidence review
In total, 9,823 references were found from a combined database search for both review questions, and full-text versions of 67 citations that seemed potentially relevant to this topic were retrieved and screened at full-text. Four observational studies were included for risk stratification (Blomquist et al., 2010; Muhtaseb et al., 2004; Osbourne et al., 2006 and Tsinopoulos et al., 2013). Twelve studies (11 observational studies and 1 systematic review) were included for risk factors (Artzen et al., 2009; Beatty et al., 1998; Blomquist et al., 2012; Briszi et al., 2012; Chatziralli et al., 2011; Chen et al., 2010; Gonzalez et al., 2014; Keklikci et al., 2009; Ling et al., 2004; Narendran et al., 2009; Robbie et al., 2006 and Rutar et al., 2009).
No additional relevant studies were identified in the update searches undertaken at the end of the guideline development process.
7.5.3.1. Description of included studies
Summaries of the included studies for the review questions are given in Table 20 and Table 21, with full evidence tables available in Appendix E and GRADE tables available in Appendix G.
Table 20
Summary of included studies – risk stratification techniques.
Table 21
Summary of included studies – risk factors for surgical complications.
7.5.4. Health economic evidence
A literature search was conducted jointly for all review questions in this guideline by applying standard health economic filters to a clinical search for cataracts (see Appendix D). A total of 4,306 references were retrieved, of which 0 were retained for this review question. Health economic modelling was not prioritised for this review question.
7.5.5. Evidence statements
7.5.5.1. Risk stratification techniques
Moderate-quality evidence from 1 retrospective cohort study of 1,883 participants found that those with a cataract risk score of >6 as determined using the Najjar–Awwad risk stratification algorithm have a clinically meaningfully increased risk of complications during cataract surgery.
Low- to high-quality evidence from 1 case-control study of 11,913 participants and 1 prospective cohort study of 1000 participants found that those with increasing potential complication scores as determined using the Muhtaseb risk stratification algorithm had clinically meaningfully higher odds of developing complications during cataract surgery.
Low-quality evidence from 1 case-control study of 11,913 participants found that those with increasing potential complication scores as determined using the Habib risk stratification algorithm had clinically meaningfully higher odds of developing complications during cataract surgery.
Very low- to low-quality evidence from 1 RCT of 953 participants could not distinguish rates of posterior capsule rupture or rates of all intraoperative complications between those in the risk stratified or unstratified arms of the trial.
Low-quality evidence from 1 RCT of 953 participants found, in the subgroup of participants operated on by trainee resident surgeons, clinically meaningfully lower odds of adverse events in the risk stratified as opposed to unstratified arm of the trial.
7.5.5.2. Risk factors
7.5.5.2.1. Risk of suprachoroidal haemorrhage
Low- to very low-quality evidence from 1 case-control study of 558 participants found those with a posterior capsule rupture, using cardiovascular drugs, with glaucoma, with increased preoperative intraocular pressure and those who undergo conversion from phacoemulsification to extracapsular cataract extraction during surgery had higher odds of developing a suprachoroidal haemorrhage during cataract surgery.
Very low-quality evidence from 1 case-control study of 99 participants could not differentiate preoperative intraocular pressure between those who did and did not develop a suprachoroidal haemorrhage during cataract surgery.
7.5.5.2.2. Risk of floppy iris syndrome
Low- to moderate-quality evidence from 1 retrospective cohort study of 59 participants found that those with a preoperative pupil diameter of ≤ 6.5 mm had higher odds of developing floppy iris syndrome during cataract surgery, but could not differentiate the odds between those receiving or not receiving prophylactic intracameral lidocaine-epinephrine.
Moderate- to high-quality evidence from 1 systematic review of 17,588 eyes found that people with hypertension had higher odds of developing floppy iris syndrome during cataract surgery, but could not differentiate the odds for people with diabetes mellitus.
Moderate-quality evidence from 1 systematic review of 17,588 eyes found that people using tamsulosin had higher odds of developing floppy iris syndrome during cataract surgery.
Moderate- to high-quality evidence from 1 systematic review of 17,588 eyes found that people using alfuzosin, terazosin or doxazosin had higher odds of developing floppy iris syndrome during cataract surgery.
7.5.5.2.3. Risk of posterior capsule rupture, vitreous loss or both
Moderate-quality evidence from 1 prospective cohort study of 55,567 participants found that those with the following preoperative characteristics had higher odds of developing posterior capsule rupture during cataract surgery:
Glaucoma
Diabetic retinopathy
Brunescent / white cataract
No fundal view / vitreous opacities
Pseudo exfoliation / phacodonesis
Pupil size (small)
Axial length ≥ 26.0mm
Low- to moderate-quality evidence from 1 prospective cohort study of 55,567 participants found that, when compared with those operated on by a consultant, people who were operated on by the following surgical grade had higher odds of developing posterior capsule rupture during cataract surgery:
Fellow
Specialist registrar
Senior house officer
but could not differentiate the odds for associate specialist or staff grade surgeons.
Low- to moderate-quality evidence from 1 prospective cohort study of 55,567 participants found that, when compared with those aged under 60 at the time of surgery, people over 70 had higher odds of developing posterior capsule rupture during cataract surgery, but could not differentiate the odds for ages 60–69.
Moderate-quality evidence from 1 prospective cohort study of 55,567 participants found that people who used doxazosin or were unable to lie flat for the operation had higher odds of developing posterior capsule rupture during cataract surgery.
7.5.5.2.4. Risk of developing intraoperative complications
Low-quality evidence from 1 prospective cohort study of 1,441 participants could not distinguish rates of intraoperative complications between those in different age groups.
Very low- to moderate-quality evidence from 1 retrospective cohort study found that people with preoperative white cataract or dense nuclear sclerosis had higher odds of developing intraoperative complications during cataract surgery but could not differentiate the odds for the following characteristics:
Small pupil (< 6.0 mm)
Anterior chamber depth < 2.5 mm
Axial length > 26.0 mm
Pseudoexfoliation syndrome
Posterior synechia
Restless patient
Moderate-quality evidence from 1 retrospective cohort study of 2,434 participants found that those with the following preoperative conditions had higher odds of developing intraoperative complications during cataract surgery:
Worse corrected distance visual acuity (logMAR)
Prior pars plana vitrectomy
Dementia
Zonular dehiscence
Very low-quality evidence from 1 retrospective cohort study of 320 participants could not distinguish rate of intraoperative complications between those with better and worse preoperative visual acuity.
Low- to very low-quality evidence from 1 prospective cohort study of 4,335 participants found that those with a preoperative visual acuity more than 1 logMAR had higher odds of developing intraoperative complications during cataract surgery than people with a preoperative visual acuity less than or equal to 0.3 logMAR.
Very low-quality evidence from 2 case-control studies of 1,255 participants found that having a preoperative white cataract increased the odds of developing intraoperative complications during cataract surgery.
Very low-quality evidence from 1 case-control study of 655 participants found that those with preoperative characteristics of phacodonesis or a brunescent/hard cataract had higher odds of developing intraoperative complications during cataract surgery, but could not differentiate the odds for those with corneal pathology or ocular comorbidity.
7.5.5.3. Health economic evidence
No health economic evidence was identified for this review question.
7.5.6. Evidence to recommendations
| Relative value of different outcomes | The committee agreed there were two important pieces of information which would inform any recommendations made. The first was whether risk-stratification algorithms were able to accurately predict individuals at higher risks of complications (and whether these algorithms stratified the risk of all complications or whether they were able to identify people at higher risk of specific complications). Secondly, it would be important to have information on whether the use of risk-stratification algorithms in practice leads to a reduction in overall complication rates, as the use of such algorithms would only be justified if it were to result in clinical benefit. |
|---|---|
| Trade-off between benefits and harms | The committee noted that the evidence presented and risk factors identified were largely in line with current clinical opinion and covered the major risk factors, including those posed from patients taking medication such as tamsulosin. The committee also noted that evidence regarding conversion rates from phacoemulsification to extracapsular cataracts extraction were likely to have changed over time with much fewer extracapsular cataract extraction operations taking place now. The committee agreed the evidence presented showed both that riskstratification algorithms worked (that is, they were able to predict people at higher risk of complications) and that the use of an algorithm in an RCT demonstrated the potential to reduce complication rates. Specifically, the use of a risk-stratification algorithm, and the assignment of more complex cases to a more experienced surgeon led to a significant reduction in complication rates for trainee surgeons, without there being a significant increase in rates for the more experienced surgeons. The committee also agreed that since this evidence came from only 1 RCT using 1 particular risk stratification algorithm, that is was appropriate for this recommendation to be made at only the ‘consider’ level. The committee agreed there were a couple of unintentional downsides that could potentially occur as a result of the widespread adoption of risk-stratification. Firstly, while the assignment of more complex cases to more experienced surgeons should reduce the overall complication rate, it may result in more experienced surgeons having worse adverse event rates, which can cause problems when these rates are used to judge surgeon performance. It was noted that surgeon-specific complication rates are risk-adjusted when results are submitted and analysed, but this can only be done in cases where patients have been preoperatively risk-stratified. Secondly, the committee agreed that there is still a need to train the next generation of cataract surgeons and it could hamper teaching opportunities if they were not able to experience more complex surgeries. As a result of this, the committee agreed that it was appropriate that specific precautions were taken to maximise clinical outcomes in people at a high risk of complications. The surgeon training needs identified above meant that the committee agreed that it would not be appropriate to say these cases should not be assigned to surgeon in training, as this could lead to greater harms in the long-term from future surgeons not being fully trained, but felt it was appropriate to recommend that trainee surgeons should only undertake these more complex cases under the close supervision of an experienced surgeon. The committee also agreed there was another important subgroup of people, those with only 1 functional eye, where the consequences of a complication could be very severe and therefore again it was appropriate that trainee surgeons should only undertake these cases under the close supervision of an experienced surgeon. |
| Consideration of health benefits and resource use | The committee agreed that the use of risk-stratification algorithms was already widespread, and that the information needed did not represent anything that should not be considered as a part of the normal preoperative process. Therefore, there was not expected to be any substantial increase in resource use from these recommendations. The committee agreed the only way in which a significant increase in resource use might result was if the use of riskstratification led to a higher proportion of cases being assigned to more experienced surgeons, but it was noted that the trial evidence did not seem to suggest this would be a likely result. |
| Quality of evidence | The committee agreed that, while the evidence on individual risk factors was generally of low quality, the fact it supported existing clinical opinion meant this was not a great cause of concern. It also noted that a number of risk-stratification algorithms had been tested in large groups of individuals with fairly consistent results, and this provided additional support to the committee’s recommendations. It was, however, noted that there was only a single study which compared two risk-stratification algorithms against each other, and in the absence of more such data it was not felt to be appropriate to recommend the use of any specific algorithm, only that the one used should have been previously validated. |
| Other considerations | The committee agreed that some of the information coming out of this review also had implications for the conversations that should be had when an individual is deciding whether or not to have surgery. The committee discussed the evidence which indicated that patients with white, brunescent or hard cataracts were at an increased risk of intraoperative complications. It agreed that surgeons like to have greater illumination of the eyes features during surgery and, in patients with a denser cataract, it can be harder to see what you are doing, making the procedure more complex. It also noted that denser cataracts are harder to break up, take longer during surgery and those with denser cataracts tend to have worse visual acuity outcome after surgery. The committee also noted that, due to the progressive nature of most cataracts, a delay in the time of surgery may result in the cataract having hardened and therefore the person being at a higher risk of complications. The committee agreed that it was appropriate that individuals considering surgery should be given this information, as it may affect the risk–benefit balance of surgery for people with good vision in the other eye. The committee also agreed it was important that people be informed of the results of their individual risk-stratification, as these may impact the overall risk–benefit balance of surgery. It was, however, noted that making sure these results are communicated clearly and understood by the person was important, as otherwise they may cause unnecessary concern. In particular, there was concern that telling individuals they are at a higher risk of complications may cause them concern, even when the absolute level of risk is still very low. Therefore, the information provided to the individual should not just talk about their relative risk of complications compared with other people, but also the absolute level of risk, as this was felt to be easier to understand and more informative for most people. |
7.5.7. Recommendations
- 17.
Consider using a validated risk stratification algorithm for people who have been referred for cataract surgery, to identify people at increased risk of complications during and after surgery.
- 18.
Explain the results of the risk stratification to the person, and discuss how it may affect their decisions.
- 19.
To minimise the risk of complications during and after surgery, ensure that surgeons in training are closely supervised when they perform cataract surgery in:
people who are at high risk of complications or
people for whom the impact of complications would be especially severe (for example, people with only 1 functional eye).
- 20.
Explain to people who are at risk of developing a dense cataract that there is an increased risk of complications if surgery is delayed and the cataract becomes more dense.
