Jump to comment:

• RE: Global variations and time trends in the prevalence of primary open angle glaucoma (POAG): a systematic review and meta-analysis. Kapetanakis, et al. 100:1 86-93 doi:10.1136/bjophthalmol-2015-307223
  Alicja R Rudnicka
  Published on: 22 March 2016
• Re: Global variations and time trends in the prevalence of primary open-angle glaucoma: a systemic review and meta-analysis
  Ching-Yu Cheng
  Published on: 22 March 2016

• Published on: 22 March 2016

  RE: Global variations and time trends in the prevalence of primary open angle glaucoma (POAG): a systematic review and meta-analysis. Kapetanakis, et al. 100:1 86-93 doi:10.1136/bjophthalmol-2015-307223

  • Alicja R Rudnicka, Reader in Medical Statistics
  • Other Contributors:
    • Venediktos V Kapetanakis, Michelle PY Chan, Paul J Foster, Derek G Cook, Christopher G Owen

  Dear Editor,

  We thank Cheng and Tham for their interest in our systematic review of global variations and time trends in primary open angle glaucoma (POAG).[1] The comments raise some important issues relating to the assimilation of evidence from observational studies, particularly in summarising estimates of chronic disease prevalence.

  Cheng and Tham have questioned our inclusive approach, which resu...

  Show More

  Dear Editor,

  We thank Cheng and Tham for their interest in our systematic review of global variations and time trends in primary open angle glaucoma (POAG).[1] The comments raise some important issues relating to the assimilation of evidence from observational studies, particularly in summarising estimates of chronic disease prevalence.

  Cheng and Tham have questioned our inclusive approach, which resulted in us having a greater number of studies, number of POAG cases and population denominator, and led to more precise estimates of POAG prevalence compared to their earlier review.[2] The purpose of our more complex approach was to explicitly model and quantify the heterogeneity between population surveys, which used different survey methods and case definitions of POAG. By quantifying these differences we are able to provide estimates of POAG prevalence standardised to studies which use optimal methods, i.e., that did not rely on intraocular pressure and routinely performed visual field assessments on all participants, while allowing studies with suboptimal methods to contribute to overall estimates by accounting for the differences in study methods. This is particularly important when attempting to appreciate time trends in POAG, given changes in study methods over time. Moreover it allows greater global representation of studies, allowing more recent studies, particularly from less developed countries often with suboptimal methods, to contribute. This would not have been the case if a more exclusive approach had been adopted.

  Cheng and Tham also raise the issue of examining response rates. However, we examined differences in response rates in our earlier review,[3] and found that these had little effect on our prevalence estimates. We acknowledge in the paper the difficulty of extracting response rates, as these are not routinely reported, and sometimes confused with participation rates, i.e., the number who took part and completed an examination out of those who agreed to take part. In general, our experience is that assessment of study quality of observational studies is complex, compared to established methods in experimental studies. Attempts were made to judge the quality of studies on criteria such as response rates and methods used. However, it is difficult to extract measures of study quality in a meaningful way; for example well- designed studies may have low response rates. Our experience suggests that subjective assessments of study quality can sometimes be arbitrary and may not capture characteristics which may influence the outcome of interest. Hence, we have chosen an approach which seeks to understand variability by explicitly modelling differences in study methods. We believe that this makes for a fuller assessment, providing evidence based guidance for the conduct of future studies. For instance, our work suggests that a recent shift away from routine visual field testing, perhaps as an inadvertent consequence of recent ISGEO guidelines,[4] may have artefactually influenced prevalence estimates; such an effect would not have been identified and quantified by a more exclusive approach.

  Cheng and Tham also question whether our more inclusive strategy may have led to lower global estimates of POAG prevalence; 2.2% (95% CrI 1.8% to 2.8%) in our study [1] versus 3.1% (1.7% to 5.3%) in their earlier review.[2] However, we would like to point out that these estimates are not dissimilar, with our more precise estimate being encapsulated within the much wider confidence interval of their earlier estimate.[2] That our estimate is marginally lower may in part be attributable to our different ethnic/regional categorization of studies, particularly amongst Asians. The earlier review simply treated Asians as one population.[2] However, we found evidence suggesting that the age specific prevalence differed across Asian populations,[1] which needs to be accounted for when deriving global estimates, especially given the large population size of Asia. Moreover, our review includes more recent studies from Asian populations, where POAG numbers are high given the large population, but age-specific prevalence is actually lower compared to other ethnic groups. We feel that our modelling approach is justified and provides an important corrective to earlier work. For example, the earlier review suggested urban-rural differences in disease prevalence,[2] where we showed no effect as urban- rural associations were inextricably linked to ethnicity and the ethnic specific associations with age.[1]

  We hope that these estimates provide greater certainty and transparency so we understand how the global and regional estimates of POAG prevalence are derived. Greater precision allows for more accurate and appropriate planning of health care service needed to care for those with the condition. We also believe that our work will assist with the design of future population based studies, which seek to describe the population prevalence of glaucoma.

  Authors: Dr Alicja R Rudnicka,* Dr Venediktos V Kapetanakis,* Miss Michelle P Y Chan,# Professor Paul J Foster,#** Professor Derek G Cook,* Professor Christopher G Owen

  *Population Health Research Institute, St George's, University of London, Cranmer Terrace, London, SW17 0RE, UK

  #Division of Genetics and Epidemiology, UCL Institute of Ophthalmology, 11-43 Bath Street, London EC1V 9EL, UK

  **NIHR Biomedical Research Centre Moorfields Eye Hospital NHS Foundation Trust, 162 City Road, London EC1V 2PD, UK.

  References:

  1. Kapetanakis VV, Chan MP, Foster PJ, Cook DG, Owen CG, Rudnicka AR. Global variations and time trends in the prevalence of primary open angle glaucoma (POAG): a systematic review and meta-analysis. Br J Ophthalmol 2015.

  2. Tham YC, Li X, Wong TY, Quigley HA, Aung T, Cheng CY. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology 2014; 121(11):2081- 2090.

  3. Rudnicka AR, Mt-Isa S, Owen CG, Cook DG, Ashby D. Variations in primary open-angle glaucoma prevalence by age, gender, and race: a Bayesian meta-analysis. Invest Ophthalmol Vis Sci 2006; 47(10):4254-4261.

  4. Foster PJ, Buhrmann R, Quigley HA, Johnson GJ. The definition and classification of glaucoma in prevalence surveys. Br J Ophthalmol 2002; 86(2):238-242.

  Conflict of Interest:

  None declared

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 22 March 2016

  Re: Global variations and time trends in the prevalence of primary open-angle glaucoma: a systemic review and meta-analysis

  • Ching-Yu Cheng, Associate Professor, Head of Ocular Epidemiology Reserach Group & Statistics Unit
  • Other Contributors:
    • Yih-Chung Tham

  Dear Editor,

  We read with interest the article by Kapetanakis et al.[1] The authors presented a systemic review of published population-based surveys to provide estimates of global and regional prevalence of primary open- angle glaucoma (POAG). The authors claimed that this review provided the most precise estimates of POAG prevalence thus far, as they adopted a more "inclusive" approach, compared to previous re...

  Show More

  Dear Editor,

  We read with interest the article by Kapetanakis et al.[1] The authors presented a systemic review of published population-based surveys to provide estimates of global and regional prevalence of primary open- angle glaucoma (POAG). The authors claimed that this review provided the most precise estimates of POAG prevalence thus far, as they adopted a more "inclusive" approach, compared to previous review.[2] As a result of this inclusive approach, more publications were included in this review with greater total number of cases and participants, and thus the authors believed that these estimates provided greater certainty compared to previous reviews.[2 3]

  Nevertheless, while an 'inclusive' approach adopted in this review generated a 'seemingly' more precise estimate (slightly narrower credible intervals), it may not necessarily provide estimates which 'accurately' reflect the trends of POAG prevalence. This is mainly because, the review also included older studies with improper definition/ diagnosis of POAG which used raised IOP (i.e. greater than 21mmHg) as a key diagnostic criterion for POAG. It has been widely recognized that using IOP as a diagnostic criterion is not appropriate and may miss POAG cases with lower levels of IOP.[4] This point has also been reiterated and acknowledged by the authors themselves in the manuscript. Furthermore, in Table 2, the authors showed that reliance on IOP in POAG diagnosis was associated with lower POAG prevalence. This finding befittingly concurs with the point that studies which used IOP as a diagnostic criterion would indeed underestimate the prevalence of POAG. Thus, inclusion of older studies with improper POAG definition in the meta-analysis would naturally result in an underestimated overall pooled estimate despite having a seemingly narrower range of credible intervals (mainly due to higher number of POAG cases). This is evident when comparing the global POAG prevalence in this review (2.2%) with the prevalence estimated in our group's recent meta- analysis project (3.05%).[2]

  The idea to include more studies in order to have 'greater power' but at the expense of compromised study quality inclusion criteria, is fundamentally flawed in the context of conducting a reliable meta- analysis. Estimation errors caused by inclusion of poor quality studies may not be sufficiently accounted for by using modeling method to adjust for study design factor. Thus, the authors' claim that 'more precise and accurate estimates were produced in this review' appears to be unconvincing and overly stated.

  Secondly, response rate was not taken into account when identifying studies to be included in this meta-analysis. It was noticed that studies of low response rates (i.e. less than 60% or 70%) were also included in the analysis. It should be noted that population-based studies with suboptimal response rate can give rise to sampling bias even though random sampling method was adopted. Thus, estimates from these studies may not be accurately representative of their respective underlying populations. Inclusion of such studies may further introduce error to the overall pooled estimate. Furthermore, if greater weights were attributed by studies which relied on IOP diagnostic criterion or by studies with poor response rate in this meta-analysis, the impact on the overall pooled estimate would be even greater. With regards to this, a forest plot ought to be included in this meta-analysis to illustrate the effect size and the weight rendered by each study. Such illustration with additional sensitivity analyses (stratified by study design factors or response rate) may help to further validate or refute the approach adopted in this analysis.

  Taken together, we humbly opine that a more 'inclusive' approach but without adequate screening of study quality, may not necessarily yield more accurate estimates. The seemingly more precise estimate (i.e. slightly narrower credible intervals) was simply due to increase in sheer number of POAG cases from studies of inadequate quality. Robust and comprehensive assessments of study quality are central to a valid, high quality meta-analysis.

  Thank you.

  References:

  1. Kapetanakis VV, Chan MP, Foster PJ, et al. Global variations and time trends in the prevalence of primary open angle glaucoma (POAG): a systematic review and meta-analysis. Br J Ophthalmol 2015.

  2. Tham YC, Li X, Wong TY, et al. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta- analysis. Ophthalmology 2014;121(11):2081-90.

  3. Quigley HA, Broman AT. The number of people with glaucoma worldwide in 2010 and 2020. Br J Ophthalmol 2006;90(3):262-7.

  4. Quigley HA. Glaucoma. Lancet (London, England) 2011;377(9774):1367 -77.

  Conflict of Interest:

  None declared

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top