Article Text

Download PDFPDF

Risk model for intraoperative complication during cataract surgery based on data from 900 000 eyes: previous intravitreal injection is a risk factor
  1. Poya Hård af Segerstad
  1. Dept of Clinical Sciences Lund, Ophthalmology, Lund University, Skåne University Hospital, Lund, Sweden
  1. Correspondence to Dr Poya Hård af Segerstad, Skåne University Hospital, Klinikg 18, Lund, Sweden; p.res{at}hardaf.se

Abstract

Background/aims The aim of this study was to develop a risk model for intraoperative complication (IC) during cataract surgery, defined as posterior capsule rupture and/or zonular dehiscence, and to include previous intravitreal therapy (pIVT) in the model.

Methods This retrospective register-based study covered patients reported to the Swedish National Cataract Register (SNCR) between 1 January 2010 and 30 June 2018. Odds ratios (ORs) were used to quantify association strength of each variable with IC. Data from the SNCR were cross referenced with the Swedish Macula Register to include data on pIVT. Variables statistically significant in the univariate analyses (p<0.05) were included in a multivariate logistic regression model.

Results The inclusion criteria were met by 907 499 eyes. The overall rate of IC was 0.86%. Variables significantly associated with IC were best corrected visual acuity ≥1.0 LogMAR (OR (adjusted): 1.75, p<0.001), age ≥90 years (OR: 1.25, p<0.001), male sex (OR: 1.09, p<0.01), pseudoexfoliation (OR: 1.33, p<0.001), glaucoma (OR: 1.11, p<0.05), diabetic retinopathy (OR: 1.35, p<0.001), pIVT (OR: 1.45, p<0.05), surgeon’s experience <600 surgeries (OR: 2.77, p<0.001), use of rhexis hooks (OR: 6.14, p<0.001), blue staining (OR: 1.87, p<0.001) and mechanical pupil dilation (OR: 1.52, p<0.001).

Conclusion The risk model can be used in the preoperative setting to predict the probability of IC, to facilitate planning of surgery and improving patient communication. Patients who have undergone intravitreal therapy prior to cataract surgery have an increased risk of IC during cataract surgery.

  • lens and zonules
  • epidemiology
  • macula

Data availability statement

The deidentified patient data used in this article are not publicly available and may be obtained from the Swedish National Cataract Register (www.kataraktreg.se) and the Swedish Macula Register (www.makulareg.se), given approval from the appropriate Swedish authorities.

Statistics from Altmetric.com

Introduction

Cataract is one of the most common causes of visual impairment, with an estimated global prevalence of 17.2%.1 The procedure of cataract surgery, phacoemulsification, is known to be safe with a low complication rate,2–4 and it is therefore difficult to identify factors that may lead to intraoperative complication (IC). Despite the low rate of complications, additional surgical procedures may be required when they do occur, compromising both patient satisfaction and the final visual outcome.5 Large dataset studies are therefore required to identify factors associated with IC,4 6 and previous intravitreal injection has recently been identified as a risk factor.7 8 Intravitreal injections of anti-vascular endothelial growth factor are mainly performed as treatment for choroidal neovascularisation in eyes with age-related macular degeneration.9 With an increasingly aged population in Western countries, the number of cataract surgeries and intravitreal injections can be expected to increase.10–12 The Swedish National Cataract Register (SNCR) includes data on patients who have undergone cataract surgery in Sweden since its inception in 1992, and has a high coverage rate.13 In 2010, the Swedish personal ID number was included in the register, making it possible to cross reference it with data in other databases.14 Similarly, the Swedish Macula Register (SMR) contains data on eyes subjected to treatment for choroidal neovascularisation, collected since 2008.15

The primary aim of this study was to develop a risk model for IC during cataract surgery, based on data from 907 499 eyes reported to the SNCR. The model could be used in the preoperative clinical setting to facilitate planning of cataract surgery and improving patient communication. The secondary aim was to cross reference the dataset from SNCR with the SMR to include data on previous intravitreal therapy (pIVT) in the model.

Materials and methods

Data on cataract surgery performed between 1 January 2010 and 30 June 2018 were extracted from the SNCR.13 The following data were considered relevant and were used in this study: Swedish personal ID number, surgeon’s ID, date of surgery, patient’s right or left eye, sex, preoperative best-corrected visual acuity (BCVA), axial length, comorbidities (pseudoexfoliation, glaucoma, macular degeneration, diabetic retinopathy, corneal endothelial degeneration, uveitis, previous refractive surgery and previous vitrectomy), intraoperative difficulties (use of rhexis hooks, blue colour staining and mechanical pupil dilation), indication for surgery (poor visual acuity, anisometropia, elevated intraocular pressure (IOP), other visual disturbance and ‘other’), and whether communication occurred between the anterior and posterior segments during surgery, referred to as ‘IC’ in this study. Data on intravitreal therapy with no date restrictions (Swedish personal ID number, date of injection and right or left eye) were extracted from the SMR.15 Data on procedures other than phacoemulsification and patients <20 years at time of surgery were excluded. The SNCR data centre assisted in calculating the patient’s age at the time of cataract surgery and age at the time of intravitreal therapy from the datasets. Surgeon’s experience was defined as the accumulated number of surgeries performed per surgeon. As the surgeon’s ID was included in the SNCR from 2007, the data centre also assisted in calculating the accumulated number of surgeries performed per surgeon from 1 January 2007 to 31 December 2009, before it was included in the dataset. They also anonymised the data by replacing the personal ID number and the ID of the cataract surgeon with randomised numbers. The data from each register were cross referenced and eyes that had received intravitreal therapy prior to cataract surgery were identified. Visual acuity reported as Snellen chart scores was converted to LogMAR prior to analysis using the formula Embedded Image.16 Infinite LogMAR values derived from Snellen scores of 0.00 were rounded to 2.0 (n=34).

Statistical analysis

To identify cut-off levels at which each continuous variable (age, BCVA, surgeon’s experience and axial length) significantly affected the risk of IC, the variables were divided into intervals: 10 years for age, 0.1 for LogMAR, 200 surgeries and 1 mm in axial length. The OR of IC was then calculated as the ratio between the odds of IC within each interval and the odds of IC outside each interval, before the variables were converted into discrete variables and divided at the level found to affect risk outcome (OR ≥1.5). ORs were used to quantify association strength in order to estimate the univariate association of each variable with IC. Nominal variables statistically significant in the univariate analyses (p<0.05) were included in a multivariate logistic regression model. Variables found to be non-significant in the multivariate analysis were removed from the model. ORs were converted into probability to make them easier to use in the clinical setting. All statistical analyses were performed using R v3.5.1 (www.R-project.org, Vienna, Austria).17

Results

The inclusion criteria were met by 907 499 eyes in 572 536 patients in the SNCR. After cross referencing, 35 739 injections of 3451 eyes in 3 168 patients were identified in the SMR as having undergone intravitreal therapy prior to cataract surgery. The mean age at the time of cataract surgery was 74.4 years, the proportions of male eyes and patients were 40.3% and 58.9%, respectively. The mean age at the time of intravitreal therapy was 81.0 years, the proportions of male eyes and patients were 37.3% and 37.9%, respectively. Patient characteristics and missing data are presented in table 1. The amount of missing data was generally low, apart from the variables axial length, pseudoexfoliation, uveitis, previous refractive surgery, previous vitrectomy and indications for surgery, as these data were collected during limited periods. Surgeon’s experience was dichotomised at 600 surgeries as the OR of IC fell below 1.5 at that level. In a similar fashion, BCVA was dichotomised at 1.0 LogMAR, and age was trichotomised at 40 years and 90 years (see figure 1). Axial length was found to be non-significant for all values and was therefore excluded from further analyses.

Figure 1

(A) OR for complication as a function of the surgeon’s experience expressed as the number of surgeries performed, in intervals of 200. (B) OR for complication as a function of best-corrected visual acuity in LogMAR, in intervals of 0.10. Empty areas in the graph indicate missing data due to an artefact after conversion from Snellen scores to LogMAR values. (C) OR for complication as a function of patient age in years, in intervals of 10 years.

Table 1

Characteristics of the patients and surgeon’s experience identified in the Swedish National Cataract Register and the Swedish Macula Register

Unadjusted risk factors

Patient characteristics associated with increased risk of IC were BCVA ≥1.0 LogMAR (corresponding to Snellen scores ≤0.1), age <40 years, age ≥90 years and male sex, together with elevated IOP and ‘other’ indications for surgery. Ocular comorbidities associated with increased risk were previous vitrectomy, pIVT (dichotomised ‘yes’ or ‘no’), pseudoexfoliation, glaucoma and diabetic retinopathy. pIVT was also calculated as the number of injections prior to cataract surgery; OR per injection: 1.03, 95% CI 1.01 to 1.05 (p<0.01). Surgeon’s experience <600 surgeries was also associated with increased risk. Intraoperative difficulties were associated with the highest risk for IC: use of rhexis hooks, blue staining and mechanical pupil dilation, which indicate zonular instability, white cataract and small pupil, respectively. To explore confounding effects between glaucoma and pseudoexfoliation, the univariate risk of IC was calculated after exclusion of pseudoexfoliation from glaucoma eyes. Glaucoma remained significant (OR: 1.49, 95% CI 1.29 to 1.74, p<0.001). Anisometropia and low visual acuity were the only protective univariate factors. The ocular comorbidities macular degeneration, corneal endothelial degeneration, uveitis, previous refractive surgery and the indication ‘other visual disturbance’ were found to be non-significant. The results of the univariate analyses are summarised in figure 2.

Figure 2

Unadjusted ORs (blue) and adjusted ORs (red). Variables found to be significant in univariate analysis were included in the multivariate logistic regression model. The variable age group <40 years, and all indications for surgery did not remain significant. Previous vitrectomy was not included in the multivariate logistic regression model due to the high number of missing data. Variables that cross the vertical line at a value of 1 are non-significant. Error bars indicate 95% CIs.

Adjusted risk factors

Variables found to be significant in the univariate analyses were included in a multivariate logistic regression model. Cases with missing data (177 179) were omitted, resulting in 730 320 eyes being included in the multivariate logistic regression model. All parameters remained significant except age group <40 years, and all indications for surgery. Previous vitrectomy could not be included due to the high number of missing data, and was excluded from the final model. pIVT remained significant when included as the number of injections (OR per injection: 1.02, 95% CI 1.00 to 1.04, p<0.05). The results of the multivariate logistic regression analysis are summarised in figure 2.

In order to determine the probability of IC prior to surgery, the adjusted risk factors for a patient were combined by multiplication of the ORs of each variable, which resulted in an OR product. A similar approach has been used in previous studies.4 This was converted in to the probability of complication by using the formula Embedded Image; where the superscript int refers to the logistic regression intercept (–5.20), and ORp denotes the OR product. The equation is plotted in figure 3. For example, a male patient aged over 89 years, with pseudoexfoliation, glaucoma and a small pupil size who is believed to require mechanical pupil dilation has an OR product of 3.06. This gives a predicted probability of IC of 1.7%. If this patient was operated on by a surgeon who had performed <600 surgeries, the OR product would be 8.47 and the probability of an IC would be 4.5%. Using this model, the probability of ICs was calculated for each eye in the dataset, and was found to be less than 1.0% for 84% of the eyes in the dataset. As can be seen in figure 4, the number of eyes declined exponentially with increasing risk. The mean and median risk of the dataset were 0.82%±0.99% and 0.56%, respectively. The mean probability of IC for eyes with no missing data that were reported as having complication (n=6001) was 1.85%±3.18%, compared with 0.81%±0.95% for eyes without complication (n=724 319).

Figure 3

Probability (or risk) of complication as a function of the OR product. The probability of complication can be read off the y-axis after calculating the product of the ORs.

Figure 4

Frequency of the probability of intraoperative complication in the eyes in the dataset (note the logarithmic y-scale). The risk of complications was low for most eyes.

Discussion

The main purpose of this retrospective database study was to develop a robust risk model for IC during cataract surgery and to investigate the impact of pIVT on the risk. In line with the findings of Lee et al7 and Shalchi et al,8 it was found that pIVT was associated with an increased risk after adjusting for age, BCVA, sex, glaucoma, pseudoexfoliation, diabetic retinopathy, ‘other’ ocular comorbidity, the surgeon’s experience and intraoperative difficulties (rhexis hooks, blue staining and mechanical pupil dilation). Over 900 000 eyes reported to the SNCR from 1 January 2010 to 30 June 2018, were cross referenced with eyes reported to the SMR that had undergone intravitreal therapy prior to cataract surgery. Both registers have an exceptionally high degree of coverage.18 19 In our dataset, 0.38% (3451 of 907 499 eyes) were treated for choroidal neovascularisation before cataract surgery, which is in contrast to the prevalence reported in previous studies by Lee et al7 and Shalchi et al,8 where the corresponding values were 2.9% and 1.6%, respectively. This could partly be explained by the fact that the SMR only includes eyes treated for choroidal neovascularisation, while diabetes and other indications were included in the data used in the above-mentioned studies. Eyes previously subjected to intravitreal therapy were included in the present study irrespective of the number of injections. When investigating the correlation between the number of injections and the risk of ICs, both the univariate and multivariate analyses showed that a higher number of injections was associated with increased risk. This information can be used in the model using ORn, where OR is the multivariate OR per injection (OR: 1.02, 95% CI 1.00 to 1.04, p<0.05), and n the number of previous intravitreal injections. Lee et al7 also found that the number of injections was associated with increased risk. This is in contrast to the findings of Shalchi et al,8 who reported no association between the number of injections and the risk of complication.

Swedish personal ID number was introduced in the SNCR from 2010. Therefore, the data on eyes were extracted from 1 January 2010 (to 30 June 2018) except for surgeon’s experience. The surgeon’s experience was not a parameter collected in the SNCR, but was included in the analyses by counting the number of surgeries per surgeon from the year surgeon ID was introduced, 1 January 2007. A limitation of this method was that all the surgeons started at ‘0’ surgeries in 2007, regardless of their previous experience. Number of performed surgeries per surgeon was therefore calculated from 1 January 2007 to 31 December 2009 before being included in the eye dataset. By this approach, we eliminated the possibility that surgeons that gained experience during 2007–2009 would be assigned surgeon’s experience ‘0’. Due to the high coverage rate of the SNCR, surgeons having ‘0’ surgeries in the dataset were most likely beginners. The risk of IC was seen to decrease significantly as a result of gained experience, and reached a steady value below the OR of 1.0 after about 1800 surgeries. This was in line with Böhringer et al20 who reported a decreasing rate of posterior capsule rupture from 4% in novice surgeons (with less experience than 300 procedures), to 0.5%–1% in surgeons with more than 1500 procedures experience.

The patient’s age, and BCVA at which the risk of IC was significantly affected (OR >1.5) were in line with those reported in some previous studies.4 21 22 Narendran et al4 reported the OR for posterior capsule rupture, vitreous loss or both to be 2.18 in patients aged above 90 years. Zetterberg et al21 reported the OR for capsule complication to be 1.82 in eyes with BCVA <0.1 Snellen (>1.0 LogMAR), while Theodoropoulou et al22 reported an OR of 2.19 in eyes with BCVA >1.2 LogMAR. In the present study, no specific cut-off value was found for axial length (data not shown). This is believed to be mainly due to the chosen interval of 1.0 mm of axial length when calculating the OR for IC. It is possible that comparing OR for larger intervals of axial length than 1.0 mm might have given different results.

IC occurred in 0.86% (7832 of 907 499 eyes) of all reported cataract surgeries. This is over a factor of ten higher than the value of 0.071% (1199 of 1 687 635 eyes) reported by Lundström et al3 using data from the European Registry of Quality Outcomes for Cataract and Refractive Surgery (EUREQUO). The reason for the considerable difference in reported complications between the Swedish and European data is not clear, however, there is a difference in the way in which the data are reported. The SNCR includes data from almost all cataract surgeries in Sweden, while the EUREQUO is based on voluntary reporting, which may lead to incomplete coverage and lower rates of reported complications.

The same approach as used by Zetterberg et al21 was used to control for confounding effects between glaucoma and pseudoexfoliation. Glaucoma was found to be significant in the present study, which is in contrast to their findings. This is believed to be due to the higher numbers of eyes in the present study. As OR is difficult to interpret and seldom used in the clinical setting, they were converted to probability, which is easier to understand and communicate. The single eye with the highest probability of ICs in the present study was that of a female aged less than 90 years, with a BCVA greater than 1.0 LogMAR, glaucoma and diabetic retinopathy, operated on by an inexperienced surgeon (<600 surgeries) using rhexis hooks, blue staining and mechanical pupil dilation. The probability of IC in this eye was 51%, but no IC was actually reported. The eye with the theoretically highest risk would yield a complication probability of 71%.

The results of this study can be used in planning cataract surgery, and to improve communication with the patient prior to surgery. Patients with a probability of complications above a certain level could be allocated to more experienced cataract surgeons, and those with the highest risk to cataract surgeons with vitreoretinal competency, so that capsule complication and lens implantation could be addressed during the same session. The thresholds used to categorise patients according to risk will depend on the demographics and resources of individual healthcare providers. The strengths of this study lie in the high number of eyes and the high coverage rate of the unique nationwide registers, while previous studies have relied on local databases. In addition, the combination of risk factors in this study have not been presented before, in particular the use of pIVT data from the SMR together with the commonly known factors such as pseudoexfoliation. Moreover, the span of more than 8 years data collection contributes to the robustness of the model, while the translation from OR to probability of IC enables use of the model in a clinical setting prior to surgery. One of the limitations is the selection of variables reported to the registers. For instance, IC is defined only as ‘communication between the anterior and posterior segment’, meaning posterior capsule rupture, or dropped nucleus. However, other types of complication can occur intraoperatively, such as wound complications, corneal oedema, iris prolapse, iris bleeding, anterior capsule complications and expulsive bleeding, which are not included in the definition. Also, some of the parameters cannot always be foreseen preoperatively, such as the use of rhexis hooks, blue staining, or mechanical pupil dilation, although they are considered to be proxy markers for zonular instability, white cataract and small pupil. The use of these measures is highly dependent on the surgeon; some rarely use rhexis hooks, while others use blue staining routinely in the early phases of training. Furthermore, errors in entering the data, misunderstandings and miscommunication are inevitable in the reporting process, and the data used in this study had not been verified against the patients’ medical records.

Even though outside the scope of this study, the main reason for the increased risk of IC in eyes with pIVT is most likely iatrogenic damage to the lens or zonulae, following poor injection technique, which should be explored in future studies. In conclusion, it was found that pIVT increased the risk of IC during cataract surgery, based on data from 907 499 eyes reported to the SNCR from 1 January 2010 to 30 June 2018. A risk model was also presented that can be used to quantify the probability of IC. The application of this model in the clinical setting prior to surgery allows patients to be allocated to surgeons with the appropriate level of experience and competence, as well as improving communication with the patient regarding the risks involved.

Data availability statement

The deidentified patient data used in this article are not publicly available and may be obtained from the Swedish National Cataract Register (www.kataraktreg.se) and the Swedish Macula Register (www.makulareg.se), given approval from the appropriate Swedish authorities.

Ethics statements

Ethics approval

The study was approved by the Ethics Committee of Lund University, and was conducted according to the principles described in the Declaration of Helsinki.

Acknowledgments

The author thanks Mr Axel Ström, MSc, Forum Söder, for extensive statistical assistance.

References

Footnotes

  • Contributors Concept, design, statistical analysis, interpretation of data and manuscript writing was performed by PHaS.

  • Funding This study was partly supported by the Foundation for Visually Impaired in former Malmöhus Län (180414), The Eye Foundation (Ögonfonden (190426)) and The Cronqvist Foundation (SLS-879931). The funding organisations had no role in the design nor conduct of this research.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.