Ophthalmic statistics note 13: method agreement studies in ophthalmology—please don’t carry on correlating…

Good clinical care depends on accurate and precise measurement and is essential throughout medicine. Many decisions about patient care in ophthalmology are based on changes in the measurement of characteristics over time (intraocular pressure in glaucoma, central retinal thickness in medical retina studies, Kmax readings in keratoconus) or differences between the measurements for an individual and the ‘normal’ population (or more strictly speaking those without the condition of interest). We want to know the true value of the characteristic, but this is obtained with measurement error. Measurement error is the difference between the known measured quantity and its unknown true value. It has two main components: random error due to chance and systematic error (sometimes known as systematic bias) not due to chance. Precision is related to random error while accuracy is related to systematic error. Measurement error may cause inappropriate patient management or harm.1 No method of measurement is completely without measurement error. Often less invasive and cheaper methods of measurement have more measurement error. One has to balance the desire for an accurate measurement, against being overly invasive to the patient or too time-consuming to incorporate in a routine clinical setting. Medicine moves at a rapid rate and new measurement technologies frequently arrive in the marketplace. While once there was a single method to measure intraocular pressure (Goldmann Applanation Tonometry), now there are many different methods and different machines on the market for example, non-contact tonometers, hand-held devices, etc. Newer methods may be more accurate and precise than older methods or may be cheaper or quicker to use. Some methods may be more convenient to use, but can have greater measurement error. The regulatory framework for devices within the UK is very different to that of medicines and while safety of the new instrument must be demonstrated, …