Jump to comment:

• Reply to 'Big data, selection bias and clinical significance
  Julie Crewe, Tim Threlfall, Antony Clark, Paul Sanfilippo and David Mackey
  Published on: 8 January 2018
• Big data, selection bias and clinical significance
  Jingjing Shen, Minqian Shen and Yuanzhi Yuan
  Published on: 16 December 2017

• Published on: 8 January 2018

  Reply to 'Big data, selection bias and clinical significance

  • Julie Crewe, Research fellow Lions Eye Institute, Centre for Ophthalmology and Vision Science, University of Western Australia, Australia.
  • Other Contributors:
    • Tim Threlfall, Research Fellow
    • Antony Clark, Consultant Ophatlmologist
    • Paul Sanfilippo, Research Fellow
    • David Mackey, Consultant Ophthalmologist/Director

  We thank the authors for their comments and insights on our paper ‘Pterygia are indicators of an increased risk of developing cutaneous melanomas’.
  We agree that there was selection bias within the pterygium cases. All cases were identified as hospital in-patients and therefore represent that small select portion of the population who were receiving treatment for (or removal of) their pterygium. We made no assumptions about whether these cases were more or less severe than untreated pterygia or whether the removals were performed for cosmetic or other reasons. The strengths of this study are that it included all in-hospital cases treated in Western Australia over a 30 year time period, without prejudice.
  While melanoma is an uncommon problem in China, it is a major health issue in Australia and New Zealand. The clinical relevance of this study should be viewed against the background of the world’s highest incidence rates of cutaneous melanoma that currently exist, with up to 60 cases per 100,000 population1-3 in Australia and New Zealand. Contrast this with the incidence rates of East Asian countries of approximately 0.7 cases per 100,000 population.4 Both non-melanoma and melanoma skin cancers are a major health priority for cancer prevention research. We are not suggesting that pterygium be used as the sole indicator for a population wide screening program. There are well established major screening and surveillance programs in place in Australia. Our study a...

  Show More

  We thank the authors for their comments and insights on our paper ‘Pterygia are indicators of an increased risk of developing cutaneous melanomas’.
  We agree that there was selection bias within the pterygium cases. All cases were identified as hospital in-patients and therefore represent that small select portion of the population who were receiving treatment for (or removal of) their pterygium. We made no assumptions about whether these cases were more or less severe than untreated pterygia or whether the removals were performed for cosmetic or other reasons. The strengths of this study are that it included all in-hospital cases treated in Western Australia over a 30 year time period, without prejudice.
  While melanoma is an uncommon problem in China, it is a major health issue in Australia and New Zealand. The clinical relevance of this study should be viewed against the background of the world’s highest incidence rates of cutaneous melanoma that currently exist, with up to 60 cases per 100,000 population1-3 in Australia and New Zealand. Contrast this with the incidence rates of East Asian countries of approximately 0.7 cases per 100,000 population.4 Both non-melanoma and melanoma skin cancers are a major health priority for cancer prevention research. We are not suggesting that pterygium be used as the sole indicator for a population wide screening program. There are well established major screening and surveillance programs in place in Australia. Our study adds a new variable to the calculation of risk for individuals (along with skin colour, hair colour, latitude and sun exposure). We do not believe our findings will cause increased anxiety to people who develop pterygium but will be a useful aid to increasing awareness of the health risks associated with high levels of solar radiation that are a part of life in Australia.
  Table 5 age groups are correct as shown but could also be written as: <50 (aged 0 to 49.99) years and ≥50 (aged 50 or more) years.
  The suggested correction (>50 years and <49 years) would exclude all people aged 50 years which would be incorrect for the data shown.
  Julie Crewe
  Tim Threlfall
  Antony Clark
  Paul Sanfilippo
  Professor David Mackey

  References
  1 MacLennan R, Green AC, McLeod GRC, Martin NG. Increasing Incidence of Cutaneous Melanoma in Queensland, Australia. JNCI: Journal of the National Cancer Institute. 1992;84(18):1427-32 DOI:10.1093/jnci/84.18.1427
  2 Marks R. The Changing Incidence and Mortality of Melanoma in Australia. In: Dummer R, Nestle FO, Burg G, editors. Cancers of the Skin: Proceedings of the 8th World Congress. Berlin, Heidelberg: Springer Berlin Heidelberg; 2002. p. 113-21.
  3 Garbe C, Leiter U. Melanoma epidemiology and trends. Clinics in Dermatology. 2009;27(1):3-9 https://doi.org/10.1016/j.clindermatol.2008.09.001
  4 Makredes M, Hui SK, Kimball AB. Melanoma in Hong Kong between 1983 and 2002: a decreasing trend in incidence observed in a complex socio–political and economic setting. Melanoma Research. 2010;20(5):427-30 DOI:10.1097/CMR.0b013e3283281072

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top
• Published on: 16 December 2017

  Big data, selection bias and clinical significance

  • Jingjing Shen, ophthalmologist 1. Zhongshan Hospital Affiliated to Fudan University 2. Center for Evidence-based Medicine, Fudan University
  • Other Contributors:
    • Minqian Shen, ophthalmologist
    • Yuanzhi Yuan, ophthalmologist

  Title Page

  Title:
  Letter to the Editor

  The article in question：
  Crewe JM, Threlfall T, Clark A, Sanfilippo PG, Mackey DA. Pterygia are indicators of an increased risk of developing cutaneous melanomas. Br J Ophthalmol 2017.

  Authors:
  Jingjing Shen
  Minqian Shen
  Yuanzhi Yuan

  Corresponding author:
  Yuanzhi Yuan

  Address：#180 Fenglin Rd., Department of Ophthalmology, Zhongshan Hospital Affiliated to Fudan University, Shanghai 200032, P.R. China
  Email: yuan.yuanzhi@zs.hospital.sh.cn
  Phone: +86-186 1688 1220 or +86-21-64041990 ext. 2684

  Dear Editor,

  We read with great interest the paper by Crewe et al.1 The authors showed that patients with pterygium had higher risk of cutaneous melanomas (CM) in a large retrospective matched-cohort study in Western Australia (WA), and suggested pterygium as an indicator for CM. The finding was interesting. However, we doubt the conclusion and its clinical relevance and public health significance.

  Compared to control group, patients with pterygium had a 20% or 24% increased risk of developing CM in terms of odds ratio(OR) or incidence rate ratio (IRR), respectively. The incidence rate difference（IRD）, however, was only 27.7/100 000 person-years (PY) (Table 5., by subtracting the IR of the control group from that of the pterygium group, i.e. (186.5-158.8)/100 000 PY). The rate difference corres...

  Show More

  Title Page

  Title:
  Letter to the Editor

  The article in question：
  Crewe JM, Threlfall T, Clark A, Sanfilippo PG, Mackey DA. Pterygia are indicators of an increased risk of developing cutaneous melanomas. Br J Ophthalmol 2017.

  Authors:
  Jingjing Shen
  Minqian Shen
  Yuanzhi Yuan

  Corresponding author:
  Yuanzhi Yuan

  Address：#180 Fenglin Rd., Department of Ophthalmology, Zhongshan Hospital Affiliated to Fudan University, Shanghai 200032, P.R. China
  Email: yuan.yuanzhi@zs.hospital.sh.cn
  Phone: +86-186 1688 1220 or +86-21-64041990 ext. 2684

  Dear Editor,

  We read with great interest the paper by Crewe et al.1 The authors showed that patients with pterygium had higher risk of cutaneous melanomas (CM) in a large retrospective matched-cohort study in Western Australia (WA), and suggested pterygium as an indicator for CM. The finding was interesting. However, we doubt the conclusion and its clinical relevance and public health significance.

  Compared to control group, patients with pterygium had a 20% or 24% increased risk of developing CM in terms of odds ratio(OR) or incidence rate ratio (IRR), respectively. The incidence rate difference（IRD）, however, was only 27.7/100 000 person-years (PY) (Table 5., by subtracting the IR of the control group from that of the pterygium group, i.e. (186.5-158.8)/100 000 PY). The rate difference corresponds to a number needed to harm (NNH) of 3610 (the reciprocal of the rate difference), which means that if 3610 individuals with pterygium that needed to be excised are followed for one year, approximately one additional CM could be detected. However, the annual cases of pterygium treatment in WA were less than 1/3 of this NNH value (figure 2; around 1000 pterygium cases were treated in WA hospitals in 2013) 1. In other words, only one person could gain benefit if all cases of pterygium treatment in WA be carefully followed up for more than 3 years. It could potentially result in unnecessary biopsies, lead to overdiagnosis and overtreatment, and cause enormous waste of resources,2 not to mention the considerable anxiety exposed to the patients and their relatives year by year.

  Giving the strategy of the exposure group (the patients with pterygium) sampling, there was obvious selection bias in Crewe’s study. The reported population prevalence of pterygium varies from 2.8% to 7.8% in Australia.3-5 But only the cases with hospital treated pterygium were included in the study, which overrepresented the serious pterygium cases with higher cumulative risk of ultraviolet radiation exposure, thus may be predisposed to CM. On the other hand, patients may have their pterygia removed for cosmetic reasons. An increased detection rate for CM could also be observed in such patients as they might be alert to any changes of their appearance. Thus, the selection of the exposure group in this cohort study could falsely increase the association between pterygium and CM. As acknowledged in the paper, only a small proportion of pterygium cases were included in the study. If, however, the potentially biased conclusion be extrapolated to general pterygium cases, it could do even more harm than good.

  A real-world study with big data is appealing because of its representation of the wider population. However, even a small effect, if any, can be statistically significant with a large sample size. Caution should especially be taken in interpreting the findings and their clinical relevance and significance.

  Moreover, a typo in Table 5 needs to be corrected. The age groups should be labeled as “>50 years” and “<49 years.”

  Jingjing Shen.1,2
  Minqian Shen. 1,2
  Yuanzhi Yuan.1,2

  1. Zhongshan Hospital Affiliated to Fudan University
  2. Center for Evidence-based Medicine, Fudan University

  Financial Disclosures: The authors have no financial disclosures.

  Reference:
  1. Crewe JM, Threlfall T, Clark A, Sanfilippo PG, Mackey DA. Pterygia are indicators of an increased risk of developing cutaneous melanomas. Br J Ophthalmol 2017.
  2. Force USPST, Bibbins-Domingo K, Grossman DC, et al. Screening for Skin Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2016;316(4):429-35.
  3. Pham TQ, Wang JJ, Rochtchina E, Mitchell P. Pterygium, pinguecula, and 5-year incidence of cataract. Am J Ophthalmol 2005;139(6):1126-8.
  4. Landers J, Henderson T, Craig J. Prevalence of pterygium in indigenous Australians within central Australia: the Central Australian Ocular Health Study. Clin Exp Ophthalmol 2011;39(7):604-6.
  5. McCarty CA, Fu CL, Taylor HR. Epidemiology of pterygium in Victoria, Australia. Br J Ophthalmol 2000;84(3):289-92.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top