Evidence review

Clinical and technical evidence

Regulatory bodies

A search of the Medicines and Healthcare Products Regulatory Agency website revealed no manufacturer Field Safety Notices or Medical Device Alerts for this device. There were 2 reports representing 1 adverse event identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE) from 2003 to the present. The event, which did not result in patient harm, seemed to be related to the replacement of the Icare probes with a non‑FDA approved version; however, it was unclear who supplied or manufactured the replacement probes.

Clinical evidence

A literature search found 116 full‑text journal articles that reported on Icare tonometers. Studies were included if they used the Goldmann applanation tonometer (GAT) as the comparator. Retrospective studies, studies from developing countries, studies done outside of the European Union, studies involving fewer than 150 eyes (except Dahlmann‑Noor et al. 2013, which was set in the UK and so may be more relevant to NHS practice) and primary studies published before 2010 (see Cook et al. 2012) were excluded. A total of 6 studies, including 1 systematic review and 5 cross‑sectional studies, have been included in this briefing.

Cook et al. (2012) did a systematic review and meta‑analysis to assess the agreement of intraocular pressure (IOP) measurements taken using tonometers available in clinical practice and GAT. A total of 102 studies involving 11,582 patients (15,525 eyes) were included in the review; 99 studies were included in the meta‑analysis. The included studies each directly compared 1 of 8 different types of tonometer with GAT (14 studies used Icare [models not reported]; 32 used dynamic contour tonometers; 4 used handheld applanation tonometers; 26 used non‑contact tonometers; 3 used Ocuton S; 12 used ocular response analysers; 14 used Tono‑Pen; 20 used transpalpebral tonometers). About 50% of all measurements from other tonometer types and 52% of all Icare measurements were estimated to be within 2 mmHg of the GAT measurements, based on pooled values across all relevant studies. Icare, which ranked fourth out of the 8 tonometers in terms of its accuracy, had a mean difference of 0.9 mmHg (95% confidence interval [CI] 0.4 to 1.4) when compared with GAT. Non‑contact and handheld applanation tonometer measurements were the most similar to GAT measurements, with 66% and 59% of IOP measurements estimated to be within 2 mmHg of the GAT measurement respectively. The authors concluded that non‑contact tonometers and handheld applanation tonometers achieve the closest values to GAT (see table 2).

Avitabile et al. (2010) was a single‑centre, cross‑sectional study set in Italy. The aim of the study was to evaluate the effect of refractive errors and central corneal thickness (CCT) on the measurement of IOP by Icare (model not stated) and its agreement with GAT measurements. Reproducibility of IOP measurements was also analysed for each method. Healthy volunteers (n=327) were recruited among students, staff and relatives of patients referred to a university eye clinic, and allocated to 1 of 4 groups: emmetropic (perfect vision, n=78), hyperopic (far‑sightedness, n=83), myopic (near‑sightedness, n=87), and astigmatic (blurred vision due to an irregular curve of the cornea, n=79). The order in which IOP was measured using Icare or GAT was randomised, and IOP was measured twice for each device by 2 experienced ophthalmologists (who each did 1 measurement sequence for each device). The reproducibility of IOP measurements was high with both Icare and GAT and no significant difference was found between the IOP readings obtained by the 2 different operators for either method. The mean difference showed that Icare measurements were slightly higher than GAT in emmetropic (0.6±1.5 mmHg, p=0.000), hyperopic (0.7±1.5 mmHg, p=0.000) and astigmatic (0.6±1.2 mmHg, p=0.000) eye groups, with the greatest mean difference reported in myopic eyes (1.6±1.8 mmHg, p=0.000). The difference between Icare and GAT was greater in higher IOP readings (p<0.001). In all groups, increasing IOP values were correlated with increasing CCT (p<0.001) but the discrepancy between Icare and GAT values was correlated with refraction (p<0.001). The authors concluded that the Icare measurements were reproducible in healthy volunteers and were slightly higher than GAT measurements of IOP in all groups (see table 3).

Dahlmann‑Noor et al. (2013) did a single‑centre, cross‑sectional study in England. The aim of the study was to examine the agreement between IOP measurements using Icare (model unknown) and GAT, in children with glaucoma (n=102) recruited from a tertiary care centre. Two different observers measured IOP using Icare (the first observer taking 2 measurements and the second taking 1 measurement), and a third observer measured IOP using GAT. The children's preferred method of measurement was also recorded by the third observer. The amount of available data varied for the different comparisons (see table 4). The mean difference between the 2 Icare readings taken by observer 1 was not statistically significant (p=0.427), nor was the mean difference between Icare readings taken by observers 1 and 2 (p=0.8). Icare generally gave higher readings than GAT with a mean difference between GAT and Icare readings of 3.3 mmHg (p<0.001). There was an association between the extent of the difference between the 2 methods and the level of the measurement, with smaller differences being seen with lower IOP measurements. There was increased disagreement (larger discrepancies) between IOP measurements using Icare and GAT with higher CCT values. Eleven children preferred GAT, 70 preferred Icare and 21 gave no preference. The authors concluded that there was poor agreement between Icare and GAT in children with glaucoma and that Icare often overestimated IOP.

Marini et al. (2011) did a single‑centre, cross‑sectional study in Italy to test the agreement between IOP measurements taken with Icare (model unknown) and GAT in people with glaucoma and ocular hypertension (n=347). IOP was measured using Icare first (IC1), then GAT, and finally a second Icare measurement was taken (IC2). The mean IOP reading taken with IC2 was significantly lower than IC1 values (p<0.001); however, it was significantly higher than for GAT (p=0.011). A significant linear correlation was found between CCT and IC1 as well as CCT and IC2, in which a 4.6 and 4.1 mmHg increase in IOP was seen for each 100 micrometre increase in CCT respectively. Smaller differences between Icare and GAT measurements were seen with lower IOP measurements (p<0.001). The authors concluded that the agreement between the methods was acceptable for low IOP measurements but not for high IOP values (see table 5).

Moreno‑Montanes et al. (2015) did a multicentre, cross‑sectional study in Spain. The aim was to compare patient‑obtained IOP measurements using Icare ONE and clinician‑obtained values using Icare PRO with GAT measurements. The usability of Icare ONE was also assessed. People (n=150; 60 people without glaucoma and 90 people with glaucoma or OHT) were recruited from routine clinical visits at 2 departments of ophthalmology. All patients had best‑corrected visual acuity of 10/20 or better. Three measurements were taken using Icare ONE and 1 measurement each was taken using Icare PRO and GAT. The order in which Icare PRO and Icare ONE measurements were taken was randomised. GAT was the last measurement taken in all people, after the anaesthetic drops were given. For all participants, the mean IOP values were 16.6±4.43 mmHg with GAT, 17.5±5.42 mmHg with Icare ONE (p=0.32 compared with GAT), and 16.6±4.77 mmHg with Icare PRO (p=0.75 compared with GAT). The IOP values were within 3 mmHg of the GAT values in 67.1% of eyes with Icare ONE and in 79.6% of eyes with Icare PRO. Icare ONE results were significantly lower than GAT results for lower IOP values (p<0.001). The differences in IOP values between Icare ONE and GAT (p=0.08) and between Icare PRO and GAT (p=0.06) were not related to CCT. Using Icare ONE was classified as very easy by 37 participants (24.7%), easy by 79 participants (52.7%), complicated by 21 participants (14%) and very complicated by 13 participants (12.5%). The perception of increased difficulty using Icare ONE correlated with increased age (p=0.003). The authors concluded that results from Icare PRO were more similar to GAT results than those from Icare ONE (see table 6).

Rosentreter et al. (2013) carried out a single‑centre, cross‑sectional study in Germany. The study aimed to evaluate the agreement between IOP measurements obtained using an Icare tonometer (model unknown) and GAT or Pascal dynamic contour tonometry (DCT) in patients with corneal abnormalities. The authors also examined the influence of CCT, corneal diameter, corneal radius and axial length on IOP measurements. One experienced ophthalmologist took 3 measurements using each method in 99 patients (171 eyes with different corneal abnormalities and 26 eyes with normal vision). About 42% of Icare measurements were estimated to be within 2 mmHg of the GAT measurements. Icare and GAT readings were not significantly influenced by CCT, axial length, corneal diameter or corneal radius. In the eyes with corneal abnormalities, IOP measurements were difficult to obtain with GAT and DCT because of sutures interfering with the tip of the tonometers, corneal surface irregularities and corneal scars, whereas IOP was measureable with Icare with all corneal abnormalities. The agreement between Icare, GAT and DCT was clinically acceptable in corneal dystrophy and keratoconus but poor in eyes after keratoplasty. The authors concluded that although there was an acceptable agreement between the 3 methods, Icare significantly underestimated IOP in all groups compared with GAT and DCT (see table 7).

Recent and ongoing studies

Four ongoing or in‑development trials on Icare for measuring IOP were identified in the preparation of this briefing:

  • Technology‑based eye care services (TECS) Compare (NCT02558712): a prospective interventional study that aims to compare the TECS protocol to the standard face‑to‑face ophthalmological examination. Participants are adults without acute or chronic ocular issues, who are patients at an eye clinic in the USA. The study is not yet open for participant recruitment and has a planned closing date of September 2017.

  • Comparison of different portable tonometers (NCT01325324): a prospective interventional study of Icare Pro, TONO‑Pen AVIA, Perkins Tonometer and PASCAL Hand Held Dynamic Contour Tonometer. Participants are healthy adults without glaucoma or other optic neuropathies. The planned study completion date was December 2015; however, the entry states that the study is currently enrolling by invitation only in Switzerland.

  • The intraocular pressure measured by different tonometers in corneal edema (NCT01998568): a cross‑sectional study that aims to assess the effect of corneal oedema on IOP values measured by GAT in comparison with dynamic contour tonometer, Icare and Tono‑Pen. Participants are adults with clinical central corneal oedema. The study, which began in November 2013, is currently recruiting participants in Thailand. It had a planned closing date of February 2016.

  • The effect of corneal biomechanical properties on rebound tonometer in patients with normal tension glaucoma (ISRCTN16912051): a cross‑sectional study to examine the effect of the structure and functioning of the cornea on IOP measurements taken with Icare, ocular response analyser and GAT. Participants are adults with or without glaucoma from a university hospital in South Korea. The study, which began in January 2013, is currently recruiting patients. The study has a planned closing date of January 2023.

Costs and resource consequences

The costing report from NICE's guideline on glaucoma estimates that 172,000 referrals are made each year to hospital eye services in England for people with ocular hypertension or suspected chronic open‑angle glaucoma (COAG). This indicates the potential usage of Icare tonometers in the NHS. In addition, a third of these people are expected to need long‑term and repeated monitoring of IOP.

According to the manufacturer, Icare tonometers are being used at 115 sites in the UK, 109 of which are NHS hospitals. They are also used in primary care settings and in people's homes, to measure and monitor IOP.

If any of the Icare tonometers were adopted, there would be no need to change the way current services are organised or delivered. No other additional facilities or technologies are needed alongside the technology.

No published evidence on the resource consequences of adopting Icare tonometers was identified in the systematic review of evidence.

Strengths and limitations of the evidence

Excluding the systematic review (which did not outline individual study designs), most of the evidence considered in this briefing was from single‑centre, prospective cross‑sectional studies. In the 5 individual studies summarised, participants were children (n=102) or adults (n=923 people; n=1021 eyes), who had either normal vision or had glaucoma, ocular hypertension or corneal abnormalities. No randomised controlled trials were identified. All included studies were carried out in Europe and all studies compared IOP measurement using Icare with GAT, the reference standard.

The evidence shows variations in the reported accuracy and repeatability of Icare tonometers. The systematic review by Cook et al. (2012) reported heterogeneity in the results of the included studies. This was because the studies used varying numbers of observations for the tonometers and reference standard, included more than 1 eye per participant (resulting in clustering of data), and there was a lack of standardisation in reporting. In addition, the analysis did not take into account the effect of CCT, which varies across a population and influences IOP measurement (giving higher IOP values when the cornea is thick and underestimating it when the cornea is thin). These all limited the extent to which the authors could accurately represent the evidence or reach meaningful conclusions about the accuracy of each available tonometer. Also, Cook et al. (2012) did not report the repeatability of GAT measurements (or whether these were included in each Icare study reviewed), which is a useful metric in evaluating comparators. Additionally, no sub‑analysis of the 14 Icare studies included in the review was done and, therefore, no conclusions can be drawn regarding their heterogeneity.

The limits of agreement between Icare and the comparator in the studies by Moreno‑Montanes et al. (2015), Marini et al. (2011) and Rosentreter et al. (2013) were very large. The repeatability for Icare in the Dahlmann‑Noor et al. (2013) study was also larger than that recorded in the broader literature for GAT (about ±3.5 mmHg), although Icare repeatability was more similar to GAT in the Marini et al. (2011) study. Both the mean difference between Icare and GAT and limits of agreement with GAT will be affected by the population examined, such as the effect of CCT and refractive errors. The mean difference between GAT and Icare of ±1.6 mmHg reported in Avitabile et al. (2010) is not necessarily clinically meaningful for all refractive errors and is similar to the variability inherent to GAT; however, it may be relevant to people with myopic eyes who have a higher risk of glaucoma. Corneal astigmatism could have contributed to varied results in this study.

Dahlmann‑Noor et al. (2013) was the only study done in the UK and so these results may be more relevant to the NHS. This was also the only study that reported the use of a sample size calculation (determined by recommendations for Bland–Altman analysis). It is unclear whether the other studies that did not report the use of a sample size calculation were adequately powered to detect differences in the outcomes. However, the studies all included large sample sizes (minimum of 150 eyes), which should increase the probability of detecting a difference between groups where such a difference exists (type II error) and should also make it less likely that a significant finding was actually a false positive.

The operators performing GAT in the study by Moreno‑Montanes et al. (2015) were masked to the Icare ONE and Icare PRO readings. The operators in the other 4 prospective studies were not blinded to the intervention. This may introduce performance bias. Another source of potential bias is the experience of the user and the resulting proficiency with the procedure. Avitabile et al. (2010) and Dahlmann‑Noor et al. (2013) used 2 observers to measure IOP in their studies. Avitabile et al. (2010) specified that the measurements were taken by experienced ophthalmologists but Dahlmann‑Noor et al. (2013) did not specify the experience of the 2 observers taking Icare measurements. In both studies, there was no significant difference in the inter‑observer variability in Icare readings. None of the other studies indicated the users' level of experience.

Avitabile et al. (2010) randomised the order in which patients had Icare and GAT. Moreno‑Montanes et al. (2015) randomised the order in which patients had Icare PRO and Icare ONE to eliminate order‑effect bias. However, they did not randomise the order of Icare and GAT. The remaining 3 cross‑sectional studies did not randomise the order in which patients received the intervention methods.

The results from Moreno‑Montanes et al. (2015) cannot be generalised to patients with advanced glaucoma, who often have low visual acuity, because only patients with a best‑corrected visual acuity of 10/20 or better were included. In this study, 3 Icare ONE measurements were taken, whereas only 1 Icare PRO and GAT measurement was taken. This may have introduced reporting bias.

Three studies (Dahlmann‑Noor et al. 2013; Avitabile et al. 2010; Marini et al. 2011) found that the difference between GAT and Icare increased with increasing IOP values, which may limit the clinical utility of this device in the detection and diagnosis of ocular hypertension and glaucoma. The effect of CCT on Icare readings reported in the Marini et al. (2011) paper was higher than those generally seen in clinical practice for GAT.

Dahlmann‑Noor et al. (2013) examined patient preference; however, they did so only after the last measurement. Asking for the patient preference after the final IOP test (as opposed to being asked to rate each individual test and comparing responses, for example) could result in recall bias. In addition, this study was limited because it had missing data, therefore the amount of available data varied for the reproducibility and repeatability of IOP measurements using Icare.

None of the authors of the publications reported any financial incentive or conflicts of interest.