Prevalence of optic disc drusen: A systematic review, meta‐analysis and forecasting study

Optic disc drusen (ODD) are calcium‐containing deposits in the optic nerve head, capable of causing visual field defects and sudden visual loss. The underlying pathophysiology remains inadequately understood and treatment options are missing. In this paper, we systematically reviewed prevalence studies of ODD in non‐selected populations to provide an overview of its prevalence, conducted meta‐analyses to determine modality‐specific prevalence estimates and performed a forecasting study to estimate current and future global population number of individuals with ODD. We searched 11 literature databases on 25 October 2022 for prevalence studies of ODD in non‐selected populations. Eight eligible studies provided data from a total of 27 463 individuals. Prevalence estimates were stratified according to diagnostic modalities: ophthalmoscopy 0.37% (95% CI: 0.10–0.95%), fundus photography 0.12% (95% CI: 0.03–0.24%), spectral domain optical coherence tomography with enhanced depth imaging 2.21% (95% CI: 1.25–3.42%) and histopathology 1.82% (95% CI: 1.32–2.38%). Using histopathology‐based summary prevalence estimate, we forecast 145 million individuals with ODD currently, a number expected to increase further due to world population growth. These numbers underscore the importance of including ODD in health education and highlight the necessity of continuing research in ODD.


| I N T RODUC T ION
Optic disc drusen (ODD) was first described in 1857 (Müller, 1857) and has since been histopathologically well-characterized as acellular deposits located in the optic nerve head anterior to the lamina cribosa (Hamann et al., 2018).Although these deposits are often benign and asymptomatic, visual field defects can be seen on examination in up to 87% (Lorentzen, 1966).ODD can also be the pathophysiological origin or increase the risk of various ischemic and vascular diseases of the retina and optic disc, that is non-arteritic anterior ischemic optic neuropathy (Fraser et al., 2021;Hamann et al., 2020;Purvin et al., 2004;Rueløkke et al., 2020), central retinal artery occlusion (Newman et al., 1989;Newsom et al., 1995;Uehara et al., 1982) and central or branch retinal vein occlusion (Austin, 1995;Chern et al., 1991;Green et al., 1981;Rothenbuehler et al., 2021).Finally, ODD may mimic the appearance of papilledema, potentially caused by a life-threatening raised intracranial pressure, and thereby putting patients at risk of unnecessary, extensive and invasive workups.Therefore, from a clinical point of view, diagnosing and recognizing ODD is of high priority for the patient as well as any ophthalmic health service.
Recognizing the importance of ODD, the intuitive next question is 'How prevalent is it?'.Prevalence of a disease plays an important part in design of education, research, advocacy and policy decisions (Lau et al., 2018).In this paper, our aim was to systematically review studies that investigate prevalence of ODD.Next, based on data extracted from eligible studies, we conducted meta-analyses for summary estimates.Data from our meta-analysis and statistical forecasting of population estimates were used to compute current and future global prevalence estimates of ODD.Taken together, we here provide the best current and, to our knowledge, the only global prevalence estimate of this important disease.
This study consisted of three parts.First, we systematically reviewed the literature in concordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al., 2009).We then conducted meta-analyses of prevalence estimates stratified according to different diagnostic modalities.For all aspects of the meta-analysis, we followed the recommendations of the Cochrane Handbook (JPT et al., 2022).Our protocol for the systematic review and the meta-analysis was registered a priori in the PROSPERO database (protocol nr.CRD42022371419).Finally, we performed a forecasting study by extrapolating the results of previous studies to population statistics and forecasts as calculated by the United Nations Population Division data.This is a well-described methodological approach for estimating current and future global prevalence estimates of various diseases (Møller-Lorentzen et al., 2020;Subhi et al., 2022;van Dijk et al., 2022).Ethical approval was not required for this study according to Danish law.

| Eligibility criteria
Eligible studies were defined as those with prevalence estimates of ODD in a non-selected population.Hence, we did not include studies of estimates in a specified disease population, fr example studies that only estimated the prevalence of ODD in patients with ischemic optic neuropathy were not considered for our review.We did not restrict on the definition of ODD or the approach for diagnosing ODD.Our review included population-based as well as clinic-based studies, as long as the population was not selected based on a disease group.The studies were expected to be observational in nature, but we did not restrict on this definition and allowed relevant data from any study design.However, single case studies, publications without original data, conference abstracts or animal studies were not considered eligible.We did not restrict on geography or journals, but for practical reasons we only considered studies reported in English language.
Record.We used search phrases and search strategies designed for the individual databases.Details of our search phrases and employed strategy used in the individual databases are provided in Data S1.
One author (H.M.) extracted references from individual databases and imported to EndNote X9.3.1 (Clarivate Analytics, Philadelphia, PA, USA) for systematic screening.One author (H.M.) examined all titles and abstracts to remove obviously irrelevant records and duplicate records.Remaining records were retrieved in full text for eligibility evaluation.Two authors (H.M. and L.M.) examined all full-text studies in an independent fashion for potential eligibility.Reference lists were reviewed for further potentially eligible studies.The authors then compared study eligibility, and a third author (Y.S.) was invited to discuss discrepancies and to reach a final decision.

| Data extraction and risk of bias within studies
We used data extraction forms to extract data from eligible studies regarding design, characteristics, methods and results.We anticipated that most studies would be cross-sectional and observational and therefore used the relevant items from the Agency for Healthcare Research and Quality checklist, which is the recommended tool for evaluating cross-sectional studies (Zeng et al., 2015).Two authors (H.M. and L.M.) independently extracted data and evaluated risk of bias within studies.The authors then compared their results, and a third author (Y.S.) was invited to discuss discrepancies and to reach a final decision.

| Outcome measures, data analysis and synthesis
Our primary outcome measure was the prevalence of ODD in a non-selected population.All studies were described qualitatively in text and in tables.For the metaanalysis, we used MetaXL 5.3 (EpiGear International, Sunrise Beach, QLD, Australia) for Microsoft Excel (Microsoft, Redmont, WA, USA).We used the metaanalyses to calculate pooled prevalence estimates within each diagnostic modality.Unit of analysis was per participant, and we did not allocate any analytic value to having ODD in one versus both eyes.We used the random-effects model to account for potential heterogeneity across studies.In prevalence meta-analyses, caution must be made when numbers are close to the extremes, that is 0% and 100%, since variance instability occurs leading to erroneous weighting of studies (Barendregt et al., 2013).This potential issue was accommodated by transforming all prevalence numbers for analyses using the double arcsine method and back transforming for interpretation (Barendregt et al., 2013).Due to low number of studies within each meta-analysis, we refrained from conducting heterogeneity statistics, Funnel plot for potential publication bias and sensitivity analyses for evaluation of robustness, because these analyses do not provide any meaningful interpretation when number of studies in analyses are <5.

| Forecasting study
Publicly available population statistics and projected population forecast were extracted using the United Nations Population Division data.We used the 2022 Revision of World Population Prospects, which is the official United Nations population estimates and projections, which considers the results of 1758 national population censuses conducted between 1950 and 2022, as well as information from vital registration systems and from 2890 nationally representative sample surveys.These data are used to construct population projections until the year 2100 that reflect a range of plausible outcomes at the global, regional and national levels.Prevalence estimates from our quantitative analyses were applied on extracted population statistics to calculate the best current and future estimate for the prevalence of ODD for each geographic region as defined by the United Nations Population Division: Africa, Asia, Europe, Latin America and the Caribbean, North America and Oceania.

| R E SU LT S
Our literature search identified 245 references, of which 107 were duplicates and 126 obviously irrelevant.Remaining 12 references were retrieved in full text for the evaluation of eligibility.Of these full-text studies, four did not fulfil our eligibility criteria and were excluded (Figure 1).Thus, eight studies were eligible for our qualitative review.One study used a convenience sampling method (Ghassibi et al., 2017), which is a markedly different sampling method than the other included studies.Therefore, we decided to include this study for the qualitative review but refrained from including it in the quantitative analyses due to heterogeneity compared to other studies.Thus, seven studies were eligible for our quantitative review.

| Study characteristics
In this review, we summarize data from eight eligible studies with a total of 27 463 participants (Bassi et al., 2019;Elbaz et al., 2019;Erkkilä, 1974;Friedman et al., 1975;Ghassibi et al., 2017;Malmqvist et al., 2020;Skougaard et al., 2020;You et al., 2009).Of these, four studies were designed as population-based, three clinicbased and one community-based.Two of the clinic-based studies were necropsies of consecutive patients.Studies were conducted in Europe (n = 4), Asia (n = 2), and North America (n = 2).Five studies were cross-sectionally designed and three were cohort studies; however, all three cohort studies provided cross-sectional data to evaluate the prevalence of ODD in their study sample.Further study characteristics including participant demographics are summarized in detail in Table 1.
Diagnosis of ODD was made using fundus images in three studies (Bassi et al., 2019;Elbaz et al., 2019;You et al., 2009), ophthalmoscopy in one study (Erkkilä, 1974), spectral domain optical coherence tomography (SD-OCT) with enhanced depth imaging (EDI) mode in two studies (Ghassibi et al., 2017;Malmqvist et al., 2020) and histopathological examinations in two studies (Friedman et al., 1975;Skougaard et al., 2020).Definitions of ODD varied across all studies.Details regarding diagnostic methods and definitions of ODD are summarized in detail in Table 2.

| Results of individual studies
Three studies examined optic discs using fundus photography (Bassi et al., 2019;Elbaz et al., 2019;You et al., 2009).Bassi et al. (2019) evaluated the prevalence of optic disc anomalies in adults in South India.This study also used 20° stereo-pair photography of the optic disc (Bassi et al., 2019).The prevalence of any optic disc anomaly was 1.1%, and the prevalence of ODD was 0.03% (Bassi et al., 2019).Elbaz et al. (2019) correlated fundus and optic nerve head findings with systemic parameters in adults in the German Gutenberg Health Study.This study reported a prevalence of ODD of 0.17% and a correlation between ODD and diabetes as well as previous myocardial infarction using logistic regression not adjusted for multiple testing (Elbaz et al., 2019).You et al. (2009) used data from the Beijing Eye Study in Northern China to evaluate correlation between ODD and ocular findings in adults.In this study, the presence of ODD was 0.19%, and its presence did not correlate with any demographic group (You et al., 2009).No correlation was found between ODD and visual acuity, refractive error, intraocular pressure, perimetry or lens opacity; however, eyes with ODD had significantly smaller optic disc area (You et al., 2009).Erkkilä (1974) used ophthalmoscopy with semicircular beam of light to examine schoolchildren aged 6-7 in Helsinki municipal primary schools and reported the prevalence of ODD.In this sample, the prevalence was 0.37% (Erkkilä, 1974).
Two studies examined optic discs using SD-OCT with EDI mode (Ghassibi et al., 2017;Malmqvist et al., 2020).Ghassibi et al. (2017) recruited healthy volunteers among spouses of glaucoma patients, volunteers in the hospital and their friends, and hospital workers' relatives or friends, and reported the prevalence of ODD in this sample and correlated to demographics, optic nerve head diameter and axial length.In this convenience sampling study from New York, 10.8% had ODD and the presence of ODD was not associated to any demographic group but was positively associated to a smaller optic nerve head diameter and a shorter axial length (Ghassibi et al., 2017).Malmqvist et al. (2020) used the Danish Copenhagen Child Cohort 2000 Eye Study to examine children that had already been examined at age 11-12 years (Malmqvist, Li, et al., 2018), again after 5 years at age 16-17 years, to evaluate presence of ODD, early risk factors for the development of ODD, and ocular correlates to prevalent ODD.This study found that the prevalence of ODD increased from 1.5% at age 11-12 years to 2.2% at age 16-17 years and that eyes which later developed ODD had prelaminar hyperreflective lines at age 11-12 years (Malmqvist et al., 2020).Furthermore, eyes with ODD had shorter scleral canal diameter (Malmqvist et al., 2020).
Two studies reported the prevalence of ODD using histopathology (Friedman et al., 1975;Skougaard et al., 2020).The sampling strategies of these studies were based on consecutive patients or eyes for histopathology (Friedman et al., 1975;Skougaard et al., 2020), which was considered the best possible for practical reasons.Friedman et al. (1975) reported the prevalence and location of ODD in a series of consecutive necropsies in New York.This study reported a prevalence of 2.04%, and that the majority of ODD were located deep in the optic nerve head (Friedman et al., 1975).Skougaard et al. (2020) reported the prevalence of ODD and its histopathological characteristics in a consecutive examination of enucleated eyes in Denmark.This study found ODD in 1.81%, and that half of all cases were only deep and that the remaining were either both deep and superficial or only superficial (Skougaard et al., 2020).

| Risk of bias of within studies
Risk of bias evaluation within individual studies showed that all studies clearly defined source of data and explained reasons for any exclusions from analyses where exclusions were made.Most studies declared eligibility criteria, time period of investigation and consecutive or populationbased sampling of participants.Quality assurance was defined as any assessments to confirm the validity of outcomes, that is multiple graders or multimodal imaging to confirm the diagnosis.This aspect was explained and explored in four of eight studies.Details of our risk of bias evaluation within studies are available as Table 3.

| of ODD using different diagnostic modalities
quantitative review is based on seven studies (Bassi et al., 2019;Elbaz et al., 2019;Erkkilä, 1974;Friedman et al., 1975;Malmqvist et al., 2020;Skougaard et al., 2020;You et al., 2009).One histopathological study on enucleated eyes provided data per eye (Skougaard et al., 2020), which we extrapolated to data per person for this analysis.Thus, this qualitative analysis is based on a total of 27 333 individuals.Outcomes were categorized according to the different diagnostic methodological approaches: ophthalmoscopy (n = 1 study), fundus photography (n = 3 studies), SD-OCT with EDI mode (n = 1 study) and histopathology (n = 2 studies).Since ophthalmoscopy and SD-OCT with EDI mode had single eligible study each, these estimates were reported as is.The summary estimate of ODD based on fundus photography was calculated to 0.12% (95% CI: 0.03-0.24%).The summary estimate of ODD based on histopathology was calculated to 1.82% (95% CI: 1.32-2.38%).Results are summarized in Table 4.

| Estimation of current and future global prevalence of ODD
For this part of our study, we used the prevalence estimate as calculated in our meta-analysis of the histopathologybased diagnosis.Based on these estimates, and global Small, whitish bodies in the optic nerve head.

Elbaz et al. (2019)
Fundus images in a darkened room and with the pupil's natural width.Three photographs were taken of each eye, starting with the right eye.Two images (30° and 45°) centred on the optic nerve, and 30° centred on the macula and the midperipheral fundus.
Crystalline yellowish white nodular lesions on the surface of the optic nerve head.Erkkilä (1974) Heine ophthalmoscopy with semicircular beam of light.
Anomalous optic disc elevation, even in the absence of visible superficial hyaline concrements, was taken to be an optic disc with drusen whenever the elevation was of a hard amorphous appearance or the disc margins were irregular and elevated, when transillumination of the elevated optic disc tissue was produced by indirect ophthalmoscopic illumination or drusen were found in the subject's contralateral eye or in the eye of another member of the family.

Friedman et al. (1975)
Microscopy of eyes fixed in 10 percent neutral buffered formalin, processed in either celloidin or paraffin, and sectioned at 15 μm (celloidin) or 9 or 10 μm (paraffin).As a matter of routine 20 or more sections were taken at random through the optic nerve and the macular area was studied unless a specific pathological process was present that warranted special study.
Serial horizontal and vertical cross-sectional scans (interval between scans ~30 μm) of the optic nerve head were obtained in both eyes of each enrolled subject using spectral domain EDI-OCT.The OCT was set to image a 15 × 10-degree rectangle for horizontal scans (and a 10 × 15-degree rectangle for vertical scans) centred on the optic disc.Each rectangle was imaged with 97 scans, and each scan had 20 OCT frames averaged.
Horizontal hyperreflective band(s) perpendicular to the OCT beam with or without a signal-poor core, except those that occur in pairs along the course of vessels representing the anterior and posterior vascular walls.The authors defined a quantitative criterion for the length of isolated hyperreflective bands (45 μm) to strengthen the reliability of ODD diagnosis.

Malmqvist et al. (2020)
A 20-degree 6-line radial OCT scan using spectral domain EDI-OCT in a high-resolution mode with averaging of 25 B-scans.
Hyporeflective round or ovoid elements with a full or partial hyperreflective margin within the optic nerve head on OCT.

Skougaard et al. (2020)
Sections of all enucleated eyes stained with haematoxylin and eosin, and where the optic nerve head was visible by microscopy, were assessed by microscopy.One central section of the optic nerve was evaluated in each eye.
Calcifications appearing as the combination of circular shape(s) with basophilic elements (haematoxylin and eosin stain) anterior to lamina cribosa.Calcifications not located in relation to the optic nerve and/or of nondefined shape, other than circular, were excluded from the analysis.

You et al. (2009)
Digital photographs of the cornea, optic disc, and fundus and retro-illuminated photographs of the lens were taken using a fundus camera after pupillary dilatation.
Small, whitish nodules in the optic nerve head with a diameter of 30-200 μm Abbreviations: EDI, enhanced depth imaging; OCT, optical coherence tomography; ODD, optic disc drusen; N/A, not available.
statistical forecasts, estimate that global prevalence of ODD is 145 million individuals, will grow steadily to 188 million individuals in year 2100.This estimate is based on the assumption that the underlying risk of developing ODD do not undergo any significant change with time, that the population projection assumptions are correct and that no significant racial differences exist in the prevalence of ODD.Details of current and future prevalence of ODD stratified according to global regions are available as Table 5.

| DI SC US SION
In this study, we systematically reviewed the prevalence of ODD using data from eight studies with 27 463 participants.We find that 1.82% of all individuals have ODD and that this translates to a current global prevalence of 145 million individuals with ODD worldwide, a number that is expected to increase to 188 million in year 2100.
These numbers can be used to explain the global impact of ODD and provide meaningful estimates for education, research, advocacy and policy design.
Studies used different diagnostic modalities, and it stood clear that modalities without insight into the depth of the optic nerve head, for example ophthalmoscopy and fundus photography, finds substantially lower prevalence estimates (0.37% and 0.12%, respectively) than diagnostic modalities that allow examination into the deeper structures of the optic nerve head, for example EDI-OCT (2.21%) and histopathology (1.82%).These findings also underscore the diagnostic inaccuracy of ophthalmoscopy and fundus photography for detecting ODD.Although other modalities also exist and may facilitate a diagnosis of ODD (ultrasonography, fundus autofluorescence, fluorescein angiography and computer tomography imaging), our review did not identify any prevalence studies based on these modalities.Youn et al. (2023) recently compared the diagnostic accuracy of various diagnostic modalities for detecting ODD.The SD-OCT with EDI mode demonstrated highest diagnostic accuracy (sensitivity: 95%, specificity: 99%) and examiner confidence compared to fundus photography (sensitivity: 38%, specificity: 97%), ultrasonography (sensitivity: 74%, specificity: 99%) and fundus autofluorescence (sensitivity: 84%, specificity: 100%) (Youn et al., 2023).Therefore, when interpreting prevalence studies of ODD, it should be acknowledged that detailed insight into the depth of optic nerve head, such as that provided of histopathology or SD-OCT with EDI mode, is important to avoid misdiagnosis and obtain accurate prevalence estimates.When using EDI-OCT for the diagnosis of ODD, using standardized protocols and definitions of diagnosis represents a key component for reproducible findings and standardized research findings (Malmqvist, Bursztyn, et al., 2018).For example, according to the Optic Disc Drusen Studies Consortium recommendation for the diagnosis of ODD, prelaminar hyperreflective lines may reflect early ODD changes, but they do not necessarily always represent ODD.Recently it was found, that almost one third of patients with papilledema due to idiopathic intracranial hypertension had prelaminar hyperreflective lines, that were demonstrated on EDI-OCT when the papilledema had subsided (Wibroe et al., 2021).2017), who included hyperreflective lines in the diagnostic definition of ODD, reported a much higher prevalence of ODD at 10.8%.Studies on ODD are often based on large, clinically evident cases.For example, when Lorentzen (1966) reported visual field defects in 87% of cases, it is worth noting that this study identified ODD based on fundoscopy, and therefore cases were likely to be large ODD, clearly visible and distinguishable from papilledema.In clinical practice, many ODD are not visible on fundoscopy and can be subtle and barely detectable even with EDI-OCT.In other words, there is a quantitative spectrum of ODD from undetectable tiny deep buried specks to large drusen, of which only the latter constitute a risk for impact on visual function.Thus, when understanding the impact of ODD in the general population, care should be exercised before extrapolating findings based on clinical studies of large ODD, on assumptions of the broad spectrum of ODD in the general population.
Strength and limitations need to be acknowledged when interpreting the results of our study.First, we strictly evaluated estimates in non-selected populations.This is important, since we are aware of several studies of ODD prevalence that include different selected populations and therefore do not provide a representative prevalence estimate of the general population.However, there is also a limitation here.From a theoretical point of view, patients who agree to participate in a general screening study, or patients undergoing necropsy or histopathology, may be more likely to have a condition and these studies are therefore theoretically prone to selection bias.Second, for the meta-analysis, Skougaard et al. (2020) only provided histopathological data per eye, which we then extrapolated to per person.Considering that the risk of ODD is higher in the second eye of the individual than that of a random sample from the population, this introduces a bias to the estimates.However, the impact of this bias may be small if any, since the other histopathological study which used data per person (Friedman et al., 1975) and the SD-OCT study with EDI mode (Malmqvist et al., 2020) both found an estimate at similar prevalence of ~2%.Third, all studies are based on populations in North America, Europe or Asia, and the summary estimate of histopathological studies are based on populations from New York and Denmark (Friedman et al., 1975;Skougaard et al., 2020).It remains unclear to which extent estimates obtained from these populations can be extrapolated to others as some racial differences may exist (Mansour & Hamed, 1991;Thurtell et al., 2012).Finally, a systematic review and meta-analysis is only as good as its studies.In this case, we were only able to include eight studies for the qualitative analysis and seven for the quantitative analysis, in which the latter was further stratified according to diagnostic modalities.Because of the small number of studies for each of the meta-analyses, we were unable to conduct meaningful analyses of any heterogeneity across studies, risk of bias across studies or sensitivity analyses.
In conclusion, we here systematically review and present summary estimates of the prevalence of ODD.Using forecasting statistics, we were able to calculate an estimate of the current and future global prevalence of ODD.These numbers underscore the importance of incorporating knowledge of ODD in health education and in the evaluation of optic nerve disease and highlight the importance of continuing research in ODD.However, careful interpretation of these prevalence numbers must be exercised since several sources of bias and lack of insight into potential interracial prevalence differences altogether may influence the accuracy of our estimates.

F U N DI NG I N FOR M AT ION
No funding is obtained for this study.

CON F L IC T OF I N T E R E ST STAT E M E N T
The authors declare that no potential conflicts of interest exist in relation to this work.

F
I G U R E 1 PRISMA flow diagram of study selection process.T A L E 1 Characteristics of the studies in review.cross-sectionalThe Chennai Glaucoma Study conducted in rural and urban South India of participants aged ≥40 years.
Details of diagnosis and definition of optic disc drusen in studies in review.ReferenceDiagnostic methods employed Definition of ODDBassi et al. (2019) Fundus images and one stereo-pair (non-simultaneous) of 20° optic disc photographs for each eye.
Gradable fundus photography was obtained from 6013 study participants.
USA Clinic-based cross-sectional Consecutive patients seen in necropsy in a single centre.Histopathological examination was performed in 737 patients.Abbreviations: OCT, optical coherence tomography; ODD, optic disc drusen; N/A, not available; USA, United States of America.
Studies are assessed Yes/No/Unclear/Not relevant (NR) on the following items from the Agency for Healthcare Research and Quality checklist: Defines source: Defines the source of information.Eligibility criteria: Lists inclusion and exclusion criteria or refers to previous publications.Time period: Indicates period used for identifying participants.Consecutive recruitment: Indicates whether subjects were consecutively recruited for the study.Quality assurance: Describes any assessments undertaken for quality assurance.Explains exclusions: Explains any patient exclusions from analyses where such exclusion is present.Prevalence is stated in % with 95% CI.Weight indicates weighting of the study in the analysis as calculated based on our random-effects model.