Article Text

Original research
Correlation between ellipsoid zone thickness and the presence of subretinal drusenoid deposits in age-related macular degeneration
  1. Raymond Matthew Bellis1,
  2. Yang Fei1,
  3. Brandon Le1,
  4. Gerardo Ledesma-Gil2,
  5. Oscar Otero-Marquez1,
  6. Yuehong Tong1,
  7. Katy Tai1,
  8. Richard B Rosen1,
  9. Gareth M C Lema1,
  10. Roland Theodore Smith1
  1. 1Ophthalmology, New York Eye and Ear Infirmary of Mount Sinai, New York, New York, USA
  2. 2Retina, Instituto de Oftalmologia Fundacion Conde de Valenciana IAP, Mexico City, Mexico
  1. Correspondence to Dr Roland Theodore Smith; RTS1MD{at}GMAIL.COM


Purpose Subretinal drusenoid deposits (SDDs) in age-related macular degeneration (AMD) are associated with systemic vascular diseases that compromise ocular perfusion. We demonstrate that SDDs are associated with decreased ellipsoid zone (EZ) thickness, further evidence of hypoxic damage.

Methods Post hoc analysis of a cross-sectional study. 165 AMD subjects (aged 51–100; 61% women). Spectral-domain optical coherence tomography was obtained in both eyes. Masked readers assigned subjects to three groups: drusen only, SDD+drusen (SDD+D) and SDD only. EZ thickness was measured subfoveally and 2000 µm nasally, temporally, superiorly and inferiorly from the fovea. Univariate testing was performed using two-tailed t-tests with Bonferroni correction.

Results The mean EZ thickness differences between the SDD+D and drusen-only groups were (in μm) 1.10, 0.67, 1.21, 1.10 and 0.50 at the foveal, nasal, temporal, superior and inferior locations, respectively (p=0.08 inferiorly, otherwise p≤0.01); between the SDD-only and drusen-only groups, the differences were 3.48, 2.48, 2.42, 2.08 and 1.42 (p≤0.0002). Differences in EZ thicknesses across all subjects and between groups were not significantly different based on gender, race or age.

Conclusion Subjects with SDDs (±drusen) had thinner EZs than those with drusen only, and the inferior EZ was least affected. EZs were thinnest in SDD-only subjects. This thinning gradation is consistent with progressive destruction of highly oxygen-sensitive mitochondria in the EZ from hypoxia. These findings support the reduced ophthalmic perfusion hypothesis for the formation of SDDs secondary to high-risk systemic vasculopathy.

  • Macula
  • Retina
  • Oxidative Stress
  • Degeneration
  • Imaging

Data availability statement

Data are available on reasonable request.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

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.


  • Numerous retinal pathologies, including age-related macular degeneration (AMD), cause damage to the ellipsoid zone (EZ) of the photoreceptors. In particular, subretinal drusenoid deposits (SDDs) in AMD have been associated with decreased EZ integrity and reflectivity, and recently, with high-risk systemic vascular disease (HRVDs).


  • This study demonstrates a quantitative link between SDDs and decreased EZ thickness that is also physiologically linked with an HRVD-driven mechanism for SDDs of insufficient ocular perfusion and hypoxic photoreceptor damage.


  • Clinicians can better provide multidisciplinary patient care for retinal, cardiovascular and neurovascular disorders by understanding this recently discovered connection between AMD and systemic vascular disease.


Age-related macular degeneration (AMD) is the leading cause of blindness in the industrialised world.1 While the hallmark AMD finding has historically been the presence of drusen—lipid-laden deposits located under the retinal pigment epithelium (RPE)—attention has been called in recent years to the presence and clinical relevance of subretinal drusenoid deposits (SDDs). Also known as reticular pseudodrusen, SDDs are found in the subretinal space above the RPE and are best identified with advanced retinal imaging such as spectral-domain optical coherence tomography (SD-OCT).2 Intermediate AMD (iAMD) is defined phenotypically by one or more large drusen >125 µm and/or pigmentary abnormalities.3 Along with drusen, SDDs are the second marker of iAMD4 and are associated with an increased risk and rate of progression relative to drusen to both atrophic and neovascular forms of late AMD.5 Moreover, it has been shown that persons with SDDs are more likely to have visual impairment and increased mortality in a 15-year follow-up period when controlling for age and sex.6 Recently, Ledesma-Gil et al found that SDDs, but not drusen, were associated with high-risk vascular diseases (HRVDs),7 which may explain poorer survival in persons with SDDs. Moreover, this work suggests a direct link between impaired perfusion and the SDD phenotype of AMD—in fact, it establishes a new AMD paradigm in which SDDs are the direct result of hypoperfusion.

Until recently, no widely accepted evidence-based mechanism for the development of SDDs existed; Querques et al theorised that SDDs represented accumulations of unphagocytised photoreceptor outer segments above the RPE which could implicate primary RPE dysfunction in the pathogenesis of SDDs,8 but this has not been validated. Curcio et al proposed a model of SDD formation characterised by apical RPE secretion of cholesterol-containing lipoprotein particles.9 However, Greferath et al found no lipid labelling in histological analysis of SDDs in donor eyes that had been identified by in vivo imaging.10 Moreover, Curcio et al found photoreceptor markers abutting, not in, SDDs on histopathology, whereas Greferath et al found photoreceptor outer segment proteins in SDDs. Both of these points favour damaged photoreceptors as the substrate for SDDs, in support of the hypoperfusion hypothesis and in opposition to the RPE secretion model.

Li et al recently observed increased choriocapillaris (CC) flow defects, decreased CC thickness and decreased mean choroidal thickness in eyes with SDDs compared with those with drusen,11 specifically supporting choroidal and CC insufficiency in the hypoperfusion theory. This proposed mechanism aligns well with the fact that SDDs are more commonly observed in the superior perifovea,12 where the retina may be more affected by physiologically decreased blood flow secondary to gravity. Given that the photoreceptors in the retina require the highest amount of oxygen per tissue area to maintain optimal function,13 it follows that a reduction in perfusion would have measurable adverse effects on photoreceptor cell health and vision.14

On SD-OCT, there are four hyper-reflective bands in the outer retina.15 The second of these bands moving outwards is the ellipsoid zone (EZ) (figure 1) within the photoreceptors.16 The EZ is densely packed with mitochondria and is, therefore, vital for photoreceptor health and function.17 Given the high reflectivity of the EZ for infrared light and resulting back-scattering of this light seen as brightness on SD-OCT, changes in the integrity and intensity of the EZ occur in various retinal pathologies, such as multiple evanescent white dot syndrome, macular holes, retinitis pigmentosa and AMD.18–20

Figure 1

Measurement of ellipsoid zone (EZ) thickness, left eye, subject with drusen but no subretinal drusenoid deposits (SDDs). (A) Near-infrared reflectance (NIR) imaging centred in the macula. The four cardinal positions for EZ thickness measurement are approximately 2000 μm superior, inferior, nasal and temporal to the fovea. The superior and inferior cardinal points are identified by the yellow calliper measurements. (B) Spectral-domain optical coherence tomography imaging at the fovea, with the EZ identified as the second hyper-reflective band in the outer retina (red line). Nasal and temporal cardinal points are identified at the termination points of the yellow horizontal line, and the EZ thickness is measured at the marked locations (yellow calliper marks). All EZ measurements are taken in areas with preserved EZ integrity. Drusen are identified as hyper-reflective extracellular accumulations deep to the retinal pigment epithelium (blue arrows).

The width (distance from temporal to nasal boundary) of the EZ in retinitis pigmentosa has been correlated with visual prognosis.21 To our knowledge, however, the thickness (distance from superior to inferior boundary) of the EZ has not been previously studied in retinal disease. We, therefore, propose EZ thickness as a new surrogate for photoreceptor health and postulate in particular that persons with SDDs have reduced EZ thickness compared with those without SDDs. Exploring this association may help researchers and clinicians further understand the pathogenesis of the SDD phenotype of AMD, its link to HRVDs and its associated clinical manifestations.


Study design

200 iAMD subjects were recruited from 2 tertiary vitreoretinal referral centres in New York City, New York, USA: Vitreous Retina Macula Consultants of New York and Department of Ophthalmology, New York Eye and Ear Infirmary of Mount Sinai. The study was conducted between August 2019 and November 2021, with a 14-month interruption due to the COVID-19 pandemic. Demographic information, history of retinal diseases and history of cardiovascular diseases were collected using a study questionnaire. Cardiovascular disease information was verified using medical records.

Inclusion criteria

Subjects were aged 51–100 years, diagnosed with iAMD in at least one eye. All subjects had the capacity to sign informed consent and complete the study questionnaire.

Exclusion criteria

Other confounding retinal or macular diseases, including retinal degenerations, retinal vascular diseases, prior retinal surgery or intravitreal injections and/or inconclusive medical or macular diagnoses.

Vascular history

Subjects self-reported relevant cardiovascular histories including stroke/transient ischaemic attack, myocardial infarction, coronary artery bypass graft surgery, angina, arrhythmia, positive stress test, positive cardiac catheterisation, valvular disease and congestive heart failure. These histories and diagnoses were cross-referenced against medical records for accuracy, and these subjects were classified as having HRVD. All other subjects were deemed to not have HRVD.

Ophthalmic imaging

Volume SD-OCT scans (27 lines, automated retinal tracking, 16 scans averaged per line, good quality at least (29–34) per the device specifications) were obtained on the Heidelberg Spectralis HRA+OCT (Heidelberg Engineering, Heidelberg, Delaware, USA). Subjects without good-quality SD-OCT scans were removed. SD-OCT images were analysed for SDDs following a published protocol.22 Subjects were assigned to three groups: drusen only, SDD+drusen (SDD+D) or SDD only (without drusen). An eye was considered to have drusen based on the standard grading for iAMD, namely one large (>125 µm) or many medium (between 63 and 125 µm) drusen noted on SD-OCT. An eye was considered to have SDDs if the SD-OCT images met the criteria for either stages 1, 2 or 3 SDD staging as defined by Cymerman et al.23 Placement of subjects into these groups was done by two masked graders with a senior retinal specialist to adjudicate cases of ambiguity.

The EZ was identified subfoveally as the next hyper-reflective band moving outward after the external limiting membrane (figure 1). The thickness of the EZ was defined to be the distance between the upper edge and lower edge of this band. Two blinded readers independently measured the EZ thicknesses manually using digital callipers on the Heidelberg Eye Explorer OCT image display at the foveal centre and then approximately 2000 µm nasally, temporally, superiorly and inferiorly to the fovea in each eye. The device measurement tool recorded distances in integers and could not discriminate distances <1 µm. The decision to take measurements 2000 µm away from the fovea in each cardinal direction was to evaluate the thickness of the EZ close to but within the perimeter of the macula, which is known to be approximately 5500 µm in diameter.24 Care was taken to ensure that the EZ was intact at all locations where thickness measurements were performed. In cases where drusen or SDDs abutted the EZ at a given measurement location such that adequate measurements could not be obtained, the measurement site was adjusted laterally to a point as close as possible to the intended site where EZ integrity was maintained and adequate measurement was possible. Collection of EZ thickness measurements was done for right eyes and left eyes separately, and the measurements obtained from the two readers were averaged. Figures 1 and 2 show representative examples of near-infrared reflectance and SD-OCT images of drusen-only and SDD+D subject eyes, respectively.

Figure 2

Measurement of ellipsoid zone (EZ) thickness, left eye, subject with both drusen and subretinal drusenoid deposits (SDDs). (A) Near-infrared reflectance (NIR) imaging centred in the macula. The four cardinal positions for EZ thickness measurement are approximately 2000 μm superior, inferior, nasal and temporal to the fovea. The superior and inferior cardinal points are identified by the yellow calliper measurements. Homogeneous hyporeflective lesions are seen throughout the macula, typical of SDDs. (B) Spectral-domain optical coherence tomography imaging at the fovea, with the EZ identified as the second hyper-reflective band in the retina (red line). Nasal and temporal cardinal points are identified at the termination points of the yellow horizontal line, and the EZ thickness is measured at the marked locations (yellow calliper marks). All EZ measurements are taken in areas with preserved EZ integrity, however, thickness of the EZ is reduced compared with the drusen-only subject. SDDs are identified as multiple smooth confluent hyper-reflective lesions (‘ribbons’ pattern, pink arrows) above the retinal pigment epithelium (RPE) and distorting the EZ. Drusen are identified as hyper-reflective extracellular accumulations deep to the RPE (blue arrows).

Statistical analysis

Drusen-only subjects were considered the index group to which the other two groups (SDD+D and SDD only) were compared. Univariate statistics for continuous variables were two-tailed t-tests with a significance level set at α=0.05. Since the same dataset was used to test multiple hypotheses (ie, EZ thickness in five locations), the Bonferroni correction was employed to reduce the risk of type I error. Given the five location hypotheses per dataset, corrected α=0.05/5=0.01. Age was reduced into a binary (≤75 years and >75 years) to test the association between age and EZ thickness. The association between subject groups and HRVD status was evaluated using Pearson’s χ2 test with α=0.05. Reliability and repeatability of EZ thickness measurements were assessed using the intraclass correlation coefficient (ICC) with a two-way random effects model based on absolute agreement. Interpretation of the ICC followed guidelines published by Koo and Li.25 Microsoft Excel 365 was used for data analysis and statistical testing.

Patient and public involvement

Patients and the public were not involved in any way in the design, conduct, reporting or dissemination plans of this research.


Patient selection, demographics and clinical characteristics

From the 200 subjects in the cross-sectional study, 35 had images that were unusable either due to poor quality or had EZ thicknesses that could not be measured at any of the 5 locations; these subjects were, therefore, excluded. 165 subjects remained with 330 potential eyes. Of these, 300 eyes were deemed of sufficient quality for analysis (at least one eye per subject).

The drusen-only group contained 163 eyes, the SDD+D group contained 93 eyes and the SDD-only group had 44 eyes. 65 subjects were male and 100 were female, while 141 were white and the remaining 24 identified as either black, Hispanic, Asian or other. The mean ages (in years) in the drusen-only, SDD+D and SDD-only groups, respectively, were 78.34, 80.74 and 79.33, with respective SD 9.24, 8.32 and 7.97. Although the subjects were not matched, the associations of age (≤75 vs >75), gender and ethnicity (white vs non-white) across drusen only, SDD+D and SDD-only subjects showed no significant differences in univariate analysis.

In the SDD-only group, 14 subjects had HRVDs (58.3%) while 10 did not. In the SDD+D group, 25 subjects had HVRDs (46.3%) while 29 did not. In the drusen-only group, 26 subjects had HVRDs (29.9%) while 61 did not (table 1). The correlation between HRVD status and SDD status is statistically significant when comparing both SDD-only subjects to drusen-only subjects and SDD+D subjects to drusen-only subjects (p≤0.0002).

Table 1

Association between high-risk vascular diseases (HRVDs) and subretinal drusenoid deposits (SDDs)

EZ findings

The mean EZ thicknesses for the drusen-only group at the foveal, nasal, temporal, superior and inferior locations were (in μm) (SD) 15.95 (2.94), 14.19 (1.97), 14.48 (2.24), 14.67 (2.31) and 14.25 (2.14), respectively, whereas in the SDD+D group the thicknesses were (in μm) 14.85 (2.34), 13.52 (2.09), 13.27 (1.98), 13.57 (1.74) and 13.75 (2.28), respectively (table 2). The mean differences between the two groups were (in μm) 1.10, 0.67, 1.21, 1.10 and 0.50. While the difference in the inferior location was not statistically significant (p=0.08), EZ thicknesses in the remaining locations were significantly different between SDD+D and drusen-only subjects (p≤0.01). Across all five locations, the EZ thickness was decreased by an average of 6.2% between drusen-only and SDD+D subjects.

Table 2

Ellipsoid zone measurements and comparisons in the drusen only, subretinal drusenoid deposit+drusen (SDD+D) and SDD-only groups

The mean EZ thicknesses for the SDD-only group at the foveal, nasal, temporal, superior and inferior locations were (in μm) (SD) 12.47 (2.89), 11.70 (1.75), 12.06 (1.79), 12.59 (2.04) and 12.83 (2.13), respectively. The mean differences from the drusen-only group were (in μm) 3.48, 2.48, 2.42, 2.08 and 1.42, with significant EZ thinning in the SDD only cases compared with drusen-only cases (p≤0.0002). Across all five locations, the EZ thickness was decreased by an average of 16.2% between drusen-only and SDD-only subjects.

Due to poor quality or inability to accurately locate or measure the EZ, thicknesses were not recorded with equal frequency at all locations. A cumulative 258 subfoveal, 292 nasal, 296 temporal, 294 superior and 295 inferior measurements were captured.

Between the two readers, the ICC was 0.56, indicating moderate reliability and repeatability.


We have demonstrated that the EZ is significantly thinner in subjects with SDDs than in those without SDDs, consistent with a predicted loss of reflective mitochondria. Moreover, we discovered a graded decrease in EZ thickness moving from eyes with drusen only to those with SDD+D to those with SDDs only and no soft drusen. This step-down pattern held true across all five measurement locations in the retina, which we postulate is consistent with a reduction in ophthalmic perfusion throughout.

Previous research on SDDs has uncovered several findings that corroborate the perfusion hypothesis—mean choroidal thickness on SD-OCT has been found to be less in persons with SDDs and drusen compared with drusen alone, and CC flow deficits on SD-OCT angiography are more pronounced in those with SDDs compared with those with conventional drusen.26–28 It is well known that the CC perfuses the RPE and photoreceptors29; since photoreceptors have the highest ATP and oxygen demand in the eye30 and the choroid has the highest rate of blood flow per unit weight of any tissue in the body,31 it follows that photoreceptors would be highly sensitive to any disruption in choroidal blood supply. Moreover, our results show that the AMD subjects with HRVDs had higher odds of having SDDs compared with drusen only. While HRVD status in this study was limited to self-reported histories and medical record diagnoses with no quantitative perfusion metrics, this correlation further supports our proposed pathogenesis of SDDs and consequent thinning of the EZ.

While our study showed a decrease in EZ thicknesses across all five locations, the inferior location notably had the least difference between drusen-only subjects and those with either SDDs+D or SDDs only. This implies that ischaemic damage to the photoreceptors appears to be relatively spared in the inferior retina, consistent with the influence of gravity directing more blood flow inferiorly.32 This finding is further strengthened by the fact that, in persons with both SDDs and drusen, SDDs are observed with higher frequency in the superior-temporal retina and lower frequency inferiorly,2 33 which is in line with our finding that the largest difference in EZ thickness between the drusen-only and SDD+D subjects occurred in the temporal location.

In addition, the largest mean difference between EZ thicknesses was found in the fovea when comparing drusen-only subjects to those with SDDs only. It is well established that the density of cones is highest in the central fovea and decreases peripherally.34 Moreover, previous research has shown that AMD subjects with SDDs have a significant reduction in cone density,35 even though SDDs are abundant outside the fovea and tend to follow rod distribution. Therefore, in eyes with SDDs, the damage to the EZ appears to be affecting both cones and rods, and the loss is greater in the fovea where photoreceptor density is greater. However, it is thus not clear why the SDDs themselves are less abundant in the fovea; this may be due to some other biological difference between the rods and cones.

The EZ is tightly packed with photoreceptor mitochondria, which are the highest retinal oxygen consumers and most susceptible to hypoxia. Our hypothesised vascular chain of events is that HRVDs lead to impaired ocular perfusion, hypoxic damage to the EZ, and finally SDD formation. These data presented in this study have shown that SDDs are associated with EZ thinning and HRVDs are associated with SDDs, however, the overall hypothesised chain of events is purely theoretical. Direct evidence that systemic perfusion metrics are correlated with ocular perfusion metrics would provide strong evidence for the first link of the chain. Indeed, it has been shown that the ocular pulse amplitude, a surrogate for perfusion, is correlated with cardiac output and cardiac index,36 and recent work has directly linked the presence of SDDs at the end of the chain to a decrease in cardiac index.37 Thus, future research should be directed to demonstrating correlations between key indicators in systemic vasculopathies and the presence of SDDs and thinning of the EZ. These latter outcomes are consistent with photoreceptor cell death and mitochondrial damage, respectively, from decreased ocular perfusion.

In this scenario, the presence of SDDs can be considered a ‘canary in a coal mine’, a sign of insufficient oxygen on SD-OCT warning clinicians to have a high index of suspicion for HRVDs and impaired ophthalmic perfusion. Conversely, persons with cardiac or carotid pathologies, impaired kidney function, hypertensive choroidopathy or other systemic diseases that are associated with poor perfusion could benefit from ophthalmological screening for the presence of the SDD form of iAMD. Because the best treatment for AMD over the life span is health management such as smoking cessation and antioxidant supplementation,38 early detection is key so that these interventions can be implemented before permanent vision loss occurs.

This study has several limitations. To our knowledge, there is no accepted normal range for EZ thickness, and therefore, our results must be expressed as differences in thickness measurements across groups. However, this study also provides absolute thickness measurements and may pave the way for future standardisation and research applications. Another limitation relates to the measurement methodology itself—in order to meticulously place the calliper bars on OCT images, it was necessary to magnify the images such that delineation between hyper-reflective bands was not always clear. As a result, intrareader and inter-reader thickness measurements can vary by several microns which can skew results, as noted by only moderate reliability between the two readers. Future work should be aimed at validating our results using ultra-high resolution OCT images and several readers with superior inter-rater reliability. Additionally, as this study was cross-sectional and not prospective, no subjects had follow-up imaging to determine whether EZ thickness changed as their iAMD progressed. Future research may shed light on the dynamics of EZ thickness and SDD concentration relative to the stage and severity of AMD. As this study was a post hoc analysis, we are unable to provide CC blood flow data which would directly assist in quantifying ocular perfusion and supporting our hypothesis. Future research endeavours in this field should employ the use of OCT angiography to compare blood flow rates in eyes with and without SDDs. Finally, the results from this moderately sized, mostly Caucasian elderly population invite replication in larger and diverse cohorts.

Strengths of this study at two tertiary retina referral centres include rigorous patient selection and AMD phenotyping with high-quality imaging for SDDs and drusen. Although EZ thicknesses are easily measured on routine SD-OCT images, EZ thicknesses in the study were measured manually by two individuals to demonstrate repeatability. We invite further efforts to study the association between EZ thickness and the presence of SDDs, and further correlate the findings with states of perfusion dysfunction. The measurement of EZ thickness can be implemented as an important quantitative indicator of photoreceptor health for clinicians and researchers.

In summary, AMD subjects with SDDs have thinner EZs than those with drusen only. This may represent hypoxic damage to the oxygen-sensitive photoreceptors—in particular, the EZ. This is consistent with, and adds further support for, the proposed mechanism of SDDs—reduced ophthalmic perfusion due to high-risk cardiovascular disease and stroke. Understanding the connection between SDDs and these vascular diseases will help clinicians reduce associated blindness, morbidity and mortality.

Data availability statement

Data are available on reasonable request.

Ethics statements

Patient consent for publication

Ethics approval

This post hoc analysis of a cross-sectional study of randomly selected iAMD subjects adhered to the tenets of the Declaration of Helsinki. The study was approved by the Icahn School of Medicine at Mount Sinai Institutional Review Board (IRB approval no. 19-00437) and informed consent was obtained from all study participants, allowing the review of relevant medical records.


Supplementary material

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Contributors RMB, YF, BL and RTS: design, acquisition of data, analysis and interpretation of data, manuscript preparation. GL-G and OO-M: acquisition of data, analysis and interpretation of data. YT and KT: acquisition of data. GMCL and RBR: manuscript preparation. RTS: responsible for overall content as guarantor.

  • Funding Regeneron Pharmaceuticals Investigator-Initiated Study (Grant # 19-0910-00001-01), Research to Prevent Blindness Challenge Grant (Grant # 22-0007-00001-01) (RTS and RBR).

  • Competing interests RBR is a consultant to OptoVue, Boehringer-Ingelheim, Astellas, Genentech-Roche, NanoRetina, OD-OS, Regeneron, Bayer, Diopsys and Teva. He has personal financial interests in Opticology, Guardian and CellView. None are relevant to the content of this paper. The remaining authors have no relevant disclosures.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

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

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.