The first real-life long-term evaluation of the effect of population screening on blindness from open-angle glaucoma
More than 20 years of follow-up of the largest screening for open-angle glaucoma conducted so far
The results suggest that bilateral blindness in the screened population was reduced by half
To evaluate the effect of population screening on low vision and blindness from open-angle glaucoma.
Retrospective cohort study.
A large population-based screening for glaucoma was conducted in Malmö, Sweden, from 1992 to 1997. A total of 42,497 subjects were invited, of which 32,918 were screened, and 9,579 were non-responders (ie, did not participate). The records of glaucoma patients who had visited the Department of Ophthalmology at Malmö University Hospital from January 1, 1987, to December 31, 2017, were reviewed. Patients diagnosed at or after the screening were assessed for moderate or severe vision impairment, here called low vision, or blindness by the World Health Organization definition. Selection bias was corrected by creating a group of potential screening participants from a comparison group of clinical patients. Main outcome measures were the risk ratios of the cumulative incidence for bilateral low vision or blindness caused by glaucoma in screened patients compared with the potential participants.
The cumulative incidence of blindness was 0.17% in the screened population versus 0.32% among the potential participants; and for low vision 0.25% versus 0.53%. The risk ratio (95% confidence interval) between the two was 0.52 (0.32-0.84) for blindness and 0.46 (0.31-0.68) for low vision. There were no differences between the proportions of potential confounders in the comparison group and those in the non-responders.
The results suggest that population screening may reduce bilateral low vision and blindness caused by glaucoma by approximately 50%.
Glaucoma is one of the leading causes of irreversible blindness worldwide. , Late presentation is the major risk factor for developing blindness from open-angle glaucoma, which suggests that screening for the disease (ie, earlier detection with subsequent treatment) may reduce the prevalence of glaucoma-induced blindness. In developed countries, approximately 50% of glaucoma cases are undiagnosed. Most of the criteria for a screenable disease are fulfilled by open-angle glaucoma, with the exception of the cost-effectiveness. , Several reports have been published regarding the potential benefit of population screening in reducing the risk of impairment due to glaucoma, but the results of those investigations represent predictions made using models based on prevalence data and presumed effects of glaucoma treatment. No study has yet reported real-life long-term effects of screening on rates of glaucoma blindness.
In the 1990s, the largest screening ever performed for open-angle glaucoma was conducted, which included 44,243 subjects, of whom 32,918 lived in Malmö. Twenty years later, that screening offered a unique opportunity to investigate the effect of screening on impairment from glaucoma, the intent of the present study.
The population screening for open-angle glaucoma was conducted from 1992-1997 in the cities of Malmö and Helsingborg, Sweden, with populations at that time of 250,000 and 135,000 citizens, respectively. The purpose was to recruit subjects with previously undetected glaucoma for possible inclusion in the Early Manifest Glaucoma Trial (EMGT). In Malmö, all residents born between 1918 and 1932 and all women born between 1933 and 1939 were invited to the screening, excluding any individuals who had visited the Department of Ophthalmology at Malmö University Hospital during the year prior to screening.
The screening procedure and study design have been described previously in detail. In short, the screening included measurement of intraocular pressure (IOP), refraction, visual acuity (VA), fundus photography, and questions regarding medical and family history and current medications. Monoscopic fundus color photographs were obtained through dilated pupils using the non-mydriatic TRC-NW3 fundus camera (Tokyo Optical Company, Tokyo, Japan) and Kodachrome 64 slide film (Kodak, Rochester, New York, USA). The picture angle was 50 degrees with the optic disc in the center of the image. All images were inspected at 10-fold magnification and searched for signs of glaucomatous disc or retinal nerve fiber layer damage by 1 glaucoma expert with considerable experience in glaucoma screening. Individuals who screened positive for at least 1 of the following criteria returned for additional examinations: 1) IOP >25 mm Hg in at least 1 eye; 2) suspected or evident glaucomatous changes in the optic disc, retinal nerve fiber defects, or optic disc hemorrhages visible in fundus photographs; 3) exfoliation syndrome; or 4) manifest glaucoma in at least 1 first-degree relative.
The post-screening examinations included repeated perimetry using the Humphrey Field Analyzer (Carl Zeiss Meditec, Dublin, California, USA) Full Threshold program, measurement of IOP using Goldmann applanation tonometry, dilated slit-lamp examination, and ophthalmoscopy.
Patients included in the EMGT were 50-79 years old and had newly diagnosed, previously untreated chronic primary open-angle glaucoma (POAG) or exfoliation glaucoma (PEXG), confirmed by 2 consecutive Humphrey Full Threshold tests with glaucomatous field defects. Defects found in the 2 field tests had to affect the same sector of the Glaucoma Hemifield Test and be classified as “outside normal limits,” or if “borderline,” with obvious localized glaucomatous changes of the optic disc in an area corresponding to the field defect.
Subjects with POAG or PEXG who fulfilled the other criteria for inclusion in the EMGT were asked to participate in the trial. Screened individuals with suspected or manifest glaucoma who did not fulfill the inclusion criteria or who refused to participate in the trial were recommended and offered treatment and follow-up care at the Department of Ophthalmology at Malmö University Hospital or with an ophthalmologist in private practice of their own choice.
The main population of this study consisted of subjects of the screen-invited birth cohorts, men and women born between 1918 and 1932, and women born between 1933 and 1939 who were living in the city of Malmö from 1992-1997 and had been diagnosed with POAG or PEXG, either at or after the screening or study start. Patients with secondary or angle-closure glaucoma were not eligible for the current study and neither were patients with POAG or PEXG diagnosed before the screening or study start. With few exceptions, glaucoma could be defined by criteria equivalent to those in the EMGT (ie, 2 consecutive Humphrey field analyzer threshold visual field test results with glaucomatous defects that had to affect the same sector of the field). The Glaucoma Hemifield Test had to be “outside normal limits,” or if “borderline,” with obvious localized glaucomatous changes of the optic disc corresponding to the field defect.
The present study compared the frequency of bilateral impairment from glaucoma among 3 groups. The “screened” group included those who attended the screening. The second group, the “non-responders to screening,” included those who had been invited to the screening but chose not to attend. The third group consisted of subjects who were a few years older or younger than the invited screening cohorts, designated the “uninvited comparison group,” which included men and women born between 1915 and 1917 and men born between 1933 and 1935. Subjects in this third group were not invited to the screening and would reveal what the proportion of impaired subjects would be without screening. , In the selection of patients into the uninvited comparison group, we tried to mimic the selection process used for the screened and non-responders as closely as possible. Inclusion and exclusion criteria were the same, and only the study start and end dates were adjusted. An overview of the selection process of the patients eligible for the present study is shown in Figure 1 .
The study start date for the screened glaucoma patients was defined as each patient´s individual screening date. Each screened age cohort had been examined consecutively according to date of birth. Non-responders to screening were assigned a study start date which was the same as the screening date for screened subjects having the same birthdate. This strategy resulted in almost identical start and follow-up times. The study end for both groups was December 31, 2017. For the uninvited comparison group, the start and end of the study period were adjusted to match age and follow-up time to those of the non-responders, who kept their original study start date. In the nonresponder age cohorts born between 1930 and 1932, the study end was shortened by 3 years to match ages and follow-up times of the respective uninvited age cohorts ( Figure 2 ).
The EMGT (US National Institutes of Health Clinical Trials.gov identifier NCT00000132; registered Sept 23, 1999) and the screening were approved by the Ethics Committee at Lund University, Sweden, in 1992, and approval was also given to the current study in 2006, with extensions most recently approved in 2019. Advertisements were published in local newspapers to allow glaucoma patients who had visited the department not to be included in the study. The EMGT was also approved by the Committee on Research Involving Human Subjects at the State University of New York at Stony Brook, USA. All studies followed the tenets of the Declaration of Helsinki.
Data were collected from medical records on visits to the Department of Ophthalmology at Malmö University Hospital from January 1, 1987, up to December 31, 2017, and were reviewed to ascertain the date of diagnosis. The start date for review was set a few years before the screening was initiated to ensure that only previously undiagnosed subjects were included in the current investigation.
Data for the size of the population of screen-invited cohorts were retrieved from screening records and protocols in paper format. The uninvited comparison group population was not part of the screening, and therefore, the mid-year number of subjects in this population was obtained from the Swedish Central Bureau of Statistics ( www.scb.se ) and adjusted for the proportion of individuals in the screening population who were not invited to the screening.
Ages for all glaucoma patients were recorded at the start and at the end of the study. In addition, sex, date of diagnosis, date of death or loss to follow-up, VA, and the amount of visual field loss, which was noted as mean deviation (MD) for each eye at the last visit, were recorded. The presence of exfoliative material in either eye on the anterior lens capsule or pupillary margin was also noted. In the cases with unilateral exfoliation syndrome, the patient was considered to be affected.
Furthermore, for each patient, the final data obtained before study end were assessed for possible impairment in either eye, as well as its cause and date of occurrence. The term vision impairment, defined according to World Health Organization (WHO) criteria ( Table 1 ), includes reversible causes of decreased vision (eg, uncorrected refractive errors). This study recorded best-corrected visual acuity and therefore used the terms low vision, corresponding to moderate and severe vision impairment of the WHO criteria, and blindness. Numbers of low vision in this report include blindness.
|Category||Visual acuity in the better eye|
|Worse than:||Equal to or better than:|
|1 Mild vision impairment||0.5 (20/40)||0.3 (20/70)|
|2 Moderate vision impairment a||0.3 (20/70)||0.1 (20/200)|
|3 Severe vision impairment a||0.1 (20/200)||0.05 (20/400)|
|4-6 Blindness||0.05 (20/400)|
|Category||Remaining central visual field in the better eye|
|Less than:||Equal to or more than:|
|3 Severe vision impairment a||20°||10°|
Bilateral low-vision was reported, based on best-corrected visual acuity and/or visual field status in the best eye. Thus, a subject with 1 blind eye and low vision in the other eye was considered to have low vision. Each eye was counted as having low vision/blindness when VA got below the threshold for low vision or blindness (ie, <20/70 or <20/400, respectively) and did not improve on subsequent visits. The same was done for visual field data, with total extent of the central field of <20 degrees but ≥10 degrees, defined as low vision, and central field extent <10 degrees defined as blindness. A temporal aspect was applied: the disease that first caused low vision or blindness was registered as the main cause for that eye. If 2 or more diseases contributed to the visual loss in 1 eye, and the date of visual loss was not known, the disease that was judged to contribute most extensively to the impairment was registered as the main cause.
The constriction of visual fields was determined by a previously described approach. , Briefly, the diameter of the remaining visual field was calculated by drawing pseudoisopters on the numerical threshold decibel (dB) map on Statpac single-field analysis printouts from the Humphrey perimeter, midway between test point locations with threshold sensitivity values of 10 dB or better and points with sensitivity less than 10 dB.
In 96% of patients in this study, low vision and blindness from glaucoma were determined based on both visual field data and VA. However, in the 4% of patients in whom visual field data were lacking, the classification was based solely on VA.
The main risk factors for rapid progression and/or glaucoma blindness were recorded from the non-responders and the uninvited comparison group (ie, presence of exfoliation syndrome; age at death; and at study start: bilateral glaucoma; level of visual field loss by mean deviation; and untreated IOP level). , The untreated IOP was noted as the mean of the measurements recorded within 3 months before either the date of diagnosis or the starting date of IOP-reducing treatment, the latter for patients who were treated before receiving a diagnosis of glaucoma.
To determine whether care practice had differed among groups, data were analyzed from a random sample of 100 patients from each of the 3 groups. The number of visits per patient within 3 years after diagnosis or referral to our department was registered, in addition to the number of patients who had been treated with laser trabeculoplasty or other glaucoma surgical procedures.
Bilateral low vision and blindness due to glaucoma were the main outcome variables. Patients were considered to have been lost to follow-up if they moved to other parts of Sweden or out of the country, and as a result, their ophthalmological data were not available to this study.
Subjects who attend a screening sometimes have a different risk of developing the disease endpoint than non-responders to screening. , The self-selection bias that may occur when comparing the outcome of screening attenders to non-responders can be corrected by the method of Duffy and associates, in which a noninvited comparison group was included. This correction is based on the assumption that the population of the uninvited comparison group would have the same proportions of screening attenders and non-responders as in the screen-invited population ( Figure 3 ). , The number of subjects and persons with low vision or blindness from glaucoma was calculated in both the potential non-responders and the potential participants in the uninvited comparison group. Results were rounded to the nearest integer.
Means of normally distributed continuous variables between groups were compared by Student´s t -test. The nonparametric Mann-Whitney U test was performed for comparisons when appropriate. Associations between categorical variables were assessed using the χ 2 test or by the z-test in case of 2 independent population proportions. The follow-up time was defined as the number of years from a patient’s study start date to the study end date. The reverse Kaplan-Meier method was used to calculate median follow-up time, where events and censoring are reversed. Thus, the events of death or bilateral blindness from any cause were treated as censored, and censored patients (ie, those lost to follow-up and those alive at study end) were treated as events. Distributions of follow-up time were compared between groups by using the log rank test. Rates and risk ratios for the cumulative incidence of low vision and blindness were calculated. Sample size and power calculations were not made for this study, because no samples were taken. The screening invited full age cohorts, and the study included all glaucoma patients visiting the authors’ department from the time of screening/study start to study end. The Statistical Package for the Social Sciences version 24.0 software (IBM, Armonk, New York, USA) was used. Risk ratios with confidence intervals (CI) were calculated using OpenEpi version 3.0.1 software (updated April 6, 2013, accessed October 25, 2019; OpenEpi, Minneapolis, Minnesota, USA).
A total of 32,918 subjects born between 1918 and 1939 were screened; 9,579 individuals refused the screening invitation; and 4,117 subjects were not invited to the screening, because they had visited the Department of Ophthalmology during the year prior to screening. Thus, the screening had an attendance rate of 77.5%. Most attendees were of European origin.
Among all individuals invited to the screening, 1,846 patients were identified who had diagnoses of POAG or PEXG at the screening or during the follow-up period; 1,575 patients were screened, and 271 were non-responders to screening. The diagnosis was set at or within 6 months of the screening in 427, and at more than 6 months after the screening in 1,148 of the screened patients. Of the latter patients, 155 received diagnoses of ocular hypertension or suspected glaucoma at the screening and were subsequently followed by regular checkups at the authors’ department or an ophthalmologist in private practice. A flowchart of the selection of patients eligible for the present study is presented in Figure 1 .
The mean age of all screen-invited glaucoma patients (ie, screened and non-responders) at the time of the screening was 67.6 years; 67.5 years for the screened, and 68.0 years for non-responders. Maximum follow-up time for each age cohort ranged from 22-25 years, and the median follow-up time for all invited patients was 23 years. Among the invited patients, 51% had PEXG, and 49% had POAG. The proportion of PEXG between screened and non-responders differed somewhat: 50% (95% CI: 0.48-0.53) and 56.5% (95% CI: 0.50-0.62), respectively ( P = 0.048, z-score). Characteristics of the screened and non-responder patients are presented in Table 2 .