The research is a cross-sectional study of factors relevant to DES incorporating a double-masked randomised controlled cross-over experiment using the WRRT. This cross-over phase is an investigation of the immediate effect of convex (low plus) lenses.

Participants were recruited from local employers, newspaper advertisements, posters (e.g., at a local university), leaflet drops, and by “snowball effect” from participants spreading the word to colleagues. The inclusion and exclusion criteria are given in Table 1.

The research followed the tenets of the Declaration of Helsinki and was approved by the Institute of Optometry Research Ethics Committee.

Potential participants who consented to participate completed the CVS-Q and mSANDE questionnaires and those who fulfilled entry criteria 1–3 (Table 1), were asked to attend for an appointment when the tests in Table 2 were carried out. Participants were asked to estimate the typical number of hours per day they used digital devices.

During the eye examination, in the unlikely event that any pathology was detected that requires medical attention, the participant was referred according to local protocols and excluded from the research study.

At the end of the eye examination, participants underwent an investigation of the immediate effect of convex lenses. Participants were informed that the purpose of this part of the research is to compare different types of reading lenses. This comprised two parts; the first part involved testing speed of reading and the second part investigated subjective preference. Throughout this testing, participants wore the subjective refractive findings of the research optometrist (maximum accepted plus for distance vision), in addition to the “adds” specified below.

A workstation was set up to simulate the participant’s workplace. This consisted of a laptop computer, placed at the distance the participant selected as matching their work monitor viewing distance and direction of gaze. Participants then undertook the WRRT (standard paper version) at this distance and direction of gaze, after a practice run, under four different conditions.45,49,52,56 The four conditions consisted of 4 pairs of trial lenses of powers plano (control condition), +0.50, +0.75, and +1.25 (not presented in this order). The lenses were standard bi-convex full-aperture optometric trial lenses inserted in a typical optometric trial frame, which also included a correction for the participant’s refractive error (subjective findings at the research appointment). The identity of the interventions was not known to either the researcher or the participant. Each intervention was randomly assigned a letter (A–D), and the order of the interventions was ABCDDCBA. The researcher was positioned out of sight but within earshot (to record results) of the participant, so the process was double-masked.

The second part involved the participants viewing the WRRT text through the four conditions, as above, and at the same working distance. They were asked to decide whether each pair of lenses (identity still masked) made their perception of text better or worse, and by how much on a scale of 0–5 where 0 indicates no difference.

The placebo control condition was a plano lens with a visible anti-reflection coating (the experimental interventions were all uncoated trial lenses). The rationale behind using a visible anti-reflection coating as a placebo is as follows. When the lenses for the plus lens conditions are introduced, participants may experience a slight magnification effect that could psychologically generate a placebo effect, even if the intervention is ineffective from an optical perspective. Plano control lenses would not create such an effect, and so participants may realise these were an inert control, compared with low plus interventions. If so, the placebo effect resulting from the magnification might be misinterpreted as a treatment effect. Therefore, we sought to use a control condition with a feature that would, from casual inspection, have a property that may be expected to generate a minimal placebo effect. An early generation anti-reflection coating was used that provides a “coloured bloom” reflection from the front of the lens. Care was taken not to use a modern anti-reflection coating whose properties have been developed to filter out blue light to potentially benefit computer users. This was because the goal of the control intervention was to act as an inert placebo control.

The DEWS (I) study criteria57 were used for the assessment and grading of dry eye (Table 3). Each participant was given a score by calculating the sum of each dry eye severity level reached (grades 1–4; the columns in Table 3) for the 9 items investigated (the rows in Table 3). Since there are 9 symptoms and signs (the rows) and four levels of each of these (the columns), the maximum score is 36. Data for the first row, relating to discomfort, originate from the mSANDE questionnaire.

In view of the sample size of >100, Kolmogorov–Smirnov tests were used to test the normality of continuous variables, and parametric and non-parametric analyses were used as appropriate. When correlations were investigated, if one or more of the variables had a non-parametric distribution, the Kendall tau correlation was used. Proportions and cross-tabulations (contingency tables) were analysed with Pearson’s chi-squared test.

The effect of different lens options on WRRT was investigated using a repeated measures ANOVA, when appropriate corrected for lack of sphericity. A significant correlation was investigated further with post hoc analyses.

Primary analyses were planned analyses of the key variables to address the research aim. Secondary analyses were exploratory in nature, for example to explore the relationship between the key variables.

A total of 107 participants were seen in the research, of whom 36% were male. The mean age was 31y (SD 5.25; range 20–40y).

No participants were excluded for criteria 4–6 in Table 1 and only one was excluded for criterion 7. The refractive data are summarised in Table 4. Other optometric characteristics will be described in a separate publication, which will also consider the impact of different criteria for prescribing optometric interventions on the proportion of participants who would have been prescribed optometric interventions. The dry eye questionnaire (mSANDE) and DEWS data are considered here, and the relationship between these data and results of CVS-Q, WRRT, and subjective preference for convex lenses.

The CVS-Q data differ significantly from a normal distribution (Kolmogorov-Smirnov test, p = 0.012). The median is 16 (IQR 12, range 6–36). The mSANDE data also differ significantly from a normal distribution (Kolmogorov-Smirnov test, p = 0.009). The median is 53 (IQR 31, range 1–90).

The data on number of hours digital device use per day differ significantly from a normal distribution (Kolmogorov-Smirnov test, p < 0.0001). The frequency distribution was inspected, and the non-normality arises because a disproportionately high number of participants (nearly 40%) spend between 7.5 and 8.5 h a day on the computer. Clearly, this artefact results from a typical working day. The median is 8 h a day (IQR 3, range 1–13).

The process of asking participants which of the lens options they preferred when viewing text on a computer display was introduced after the start of the research. Of the 84 participants for whom these data were available, 57 (68%) were able to choose one option as preferable, 17 chose two options as jointly preferable, 7 chose 3 equally as best and two could not choose. These data are summarised in Fig. 1. To determine the frequency of lens choice and to allow for participants who chose more than one lens, a ‘weighted’ system was used where a score of 1.00 was given when only one lens was chosen, 0.50 when 2 lenses were equally chosen, a score of 0.33 for 3, and a score of 0.25 when 4 lenses were chosen as equally preferred. Using this method, it was found that 20% preferred the control lens; 25%, +0.50; 47%, +0.75; and 8%, +1.25.

Fig. 1.

Venn diagram of participants’ subjective lens choice. The size of the circles is proportionate to the number of participants in each category. Note, not shown are the 2 participants who could not decide.

(0.05MB).

Participants were divided into four age categories: 20–25y (N = 8); 26–30y (N = 26); 31–35y (N = 15); and 36–40y (N = 8). These categories only include those who chose one lens power as their preferred (68% of the total who performed the subjective preference test) and the remainder chose two or more lenses as their preferred lens. The relationship between lens choice and age is shown in Table 5.

Chi-square analysis of the data in Table 5 reveals that there is a significant relationship between subjective lens choice and age (Pearson Chi-Square, p < 0.05; Phi and Cramer’s V symmetric measures both also gave p < 0.05). This is illustrated in Fig. 2, revealing that a disproportionate number of the age group 26–30y chose the +0.75D lens option.

Fig. 2.

Graph of subjective preference in age-groups for different lens options for the 57 participants who chose one option as preferred.

(0.05MB).

The WRRT data for each of the four conditions does not significantly depart from a normal distribution (Kolmogorov–Smirnov test, p = 0.20). The minimum and maximum WRRT for any participant in any condition were 97 and 264 wpm. WRRT data are summarised in Fig. 3.

Fig. 3.

Bar chart illustrating the mean and standard errors of WRRT data for each condition.

(0.04MB).

A repeated measures ANOVA was carried out to determine the effect of different add options on WRRT performance. Using a Greenhouse-Geisser correction for the lack of sphericity (Mauchly’s test of sphericity, p = 0.044), the mean scores of WRRT for the different conditions were statistically significantly different (F = 16.531 p < 0.0001). Post hoc analyses, adjusted for multiple comparisons, revealed that the +0.50 (p < 0.001) and +0.75 (p < 0.0001) conditions were significantly different from the control condition, but the +1.25 condition did not differ significantly from the control condition (p = 1.00). The +0.50 differed from the both the +0.75 and +1.25 lens (p < 0.05), whilst the +0.75 format also differed from the +1.25 condition (p < 0.0001).

The frequency for which lens condition gave the best reading rate was analysed. Of the 107 participants, 12% read fastest with the control lens, 19% with +0.50, 51% with +0.75 and 18% with +1.25. These figures take into account when subjects read equally well under more than one condition using the same method as described above for subjective choice (5 participants read equally fast with more than one lens; see Fig. 4).

Fig. 4.

Venn diagram of frequency with which lens participants performed best on the WRRT. Note; not shown is 1 person who read equally under all 4 conditions.

(0.04MB).

The WRRT data were also analysed by comparing the proportion of participants who read faster with any plus power compared to the control lens (participants did not know which was the control intervention), using previously published criteria for significant performance improvement with the WRRT (see Discussion). Of the 107 participants, 24% read more than 10% faster with convex lenses and 8% read more than 15% faster with adds than control.

The influence of age was investigated using the four age categories described above. Since there were very few ties for the fastest lens condition (5 participants out of 107 read equally fast with more than one lens condition), only the data for participants who read fastest with one lens (N = 102) were analysed. This is illustrated in Fig. 5 and Chi-square analysis reveals that there is a significant relationship between the lens that gave the fastest reading speed and age (Pearson Chi-Square <0.001, Phi and Cramer’s V symmetric measures both p < 0.001). As with subjective preference, the graph clearly reveals that there is a disproportionate number of participants who perform best with the +0.75D lens in the 26–30 age group.

Fig. 5.

Graph illustrating relationship between lens producing best WRRT performance (excluding ties) and age group.

(0.05MB).

Participants’ subjective choice for preferred lens was compared with the lens that gave them best performance with the WRRT. Data on subjective preference were available for 84 of the 107 participants. First, if one only considers the 57 participants who were able to decide on one lens as preferred, there was a tendency for the ranking of lens preference to correlate with WRRT performance (Kendall correlation 0.711, p = 0.052). The subjectively preferred lens agreed with the lens to give best WRRT performance in 28% of cases. Second, including the participants who ranked more than one lens as their equal best lens, this agreed with the lens that gave best WRRT performance in 49% of cases and the Kendall correlation was 0.356 (p = 0.072).

The combined DEWS score differs significantly from a normal distribution (Kolmogorov-Smirnov test, p < 0.0001). The median is 12 (IQR 2.00, range 9.00–17.00).

In view of the non-parametric nature of some variables, correlations were tested using the non-parametric Kendall tau correlation, which also allows the inclusion of subjectively preferred lens and the add power that gave best performance at WRRT. The correlations are illustrated in Table 6. Multiple correlations in a correlation matrix of this type need to be viewed with caution and only significance at the level of p < 0.001 is highlighted (in bold) in the table.

The contribution of dry eye to DES was further explored by contrasting the participants with most marked dry eye with those exhibiting least symptoms and signs of dry eye, based on the DEWS scores (including mSANDE symptoms; Table 3). The initial plan was to divide the participants into tertiles, but this was problematic because the range of DEWS scores is narrow (9–17), with 20% of participants having an identical score of 13. After careful consideration of Table 3 and the distribution of scores, the participants were divided into three groups: negligible dry eye (score 0–12, N = 67); borderline dry eye (score 13, N = 21); and significant dry eye (score >13, N = 19). The borderline group were not considered further. The key findings in this research were contrasted between the participants in the negligible dry eye and significant dry eye groups. The results of these analyses are reproduced in Table 7.

The participants’subjective lens choice was inspected in the two new subgroups to see if either subgroup departed from the trend observed in the whole sample and illustrated in Fig. 1. Both subgroups showed a similar trend to Fig. 1, with +0.75D the most popular choice and +0.50D the second most popular choice. Similarly, the WRRT mean performance with each lens (Fig. 3) was replotted in the two new subgroups. These revealed a similar trend to that in Fig. 3, with no diminishing of the pre-eminence of +0.75 in either sub-group.

The finding that approximately two thirds of our population with DES were female is in agreement with the literature.1,5 The CVS-Q data highlight the symptomatic nature of our sample, who used digital devices for a median of 8 h a day.

Considering subjective preference, the most popular add is +0.75, preferred by almost half of participants, with one quarter preferring +0.50 and only 8% preferring +1.25. Fig. 1 shows that approximately half of those whose preference was for +1.25 placed this joint first choice with +0.75. It is interesting that a consideration of the relationship between age and subjective lens preference (Fig. 2) did not reveal any tendency for older participants to choose higher power adds.

The WRRT data confirmed the subjective preference, with half the participants performing best with +0.75 (Fig. 4). Analysis by repeated measures ANOVA revealed that performance with both the +0.50 and +0.75 options was significantly better than with the control lens, but performance with the +1.25 option was not significantly better than with the control lenses (Fig. 3). As with subjective preference, the data do not indicate a tendency for older participants to perform better with higher power convex lenses (Fig. 5), although it should be noted that the age range is pre-presbyopic.

Although the correlation between the add power that was preferred and the add power that gave best performance was higher than one would expect from chance, there were clearly some participants whose subjective preference differed from the add that gave them best performance. This is not surprising, since the factors that would influence subjective preference (including perceived comfort) may differ from those affecting performance.

The SANDE questionnaire is typically used to monitor change,38,58,59 although the correlation matrix supports previous research60 indicating a role in detecting dry eye. Indeed, considering the well-known challenge in dry eye practice of clinical tests only poorly predicting symptoms,61 it is encouraging that the present study obtained a significant correlation (Table 6) between the mSANDE questionnaire result and the DEWS signs (rtau = 597, p <0.001; note, the discomfort score was excluded from the DEWS score for this correlation).

The significant correlation between the CVS-Q and mSANDE questionnaires (rtau = 0.469, p < 0.001) replicates previous work highlighting the importance of dry eye in the aetiology of DES. In 2003 Sheedy and colleagues identified two different causes of asthenopia, external symptom factor (dry eye) and internal symptom factor (refractive, accommodative, and binocular anomalies).7 This classification has subsequently been applied to DES.2,5 It seems likely that, within a cohort of patients with DES, there will be some whose symptoms are wholly attributable to external factors (dry eye), some with symptoms wholly attributable to internal factors, and some with a combination of internal and external aetiology. The low correlations between most of the variables in Table 6 supports a multifactorial aetiology for DES. Therefore, it should not be expected that convex lenses will help all people with DES.

This issue was further explored with the subgroup analysis, comparing participants whose DEWS scores indicate negligible dry eye with those whose scores indicate significant dry eye. Unsurprisingly, the subgroup with significant dry eye had statistically significantly worse CVS-Q scores. However, in key variables relating to the effect of low adds, the two subgroups did not differ significantly (Table 7). These results support previous findings that dry eye is in some cases a contributory factor to DES, but indicate that it is independent of the benefit from low plus lenses.

An interesting finding is the statistically significant tendency for the add that gives best performance and that is subjectively preferred to be +0.75. It is perhaps not surprising that almost twice as many participants preferred +0.75 (47%) to +0.50 (25%), since both powers are similar in effect, but +0.75 would be likely to give more magnification and possibly clarity than +0.50. It is perhaps more surprising that only 8% of participants preferred +1.25. This is unlikely to be a chance finding because it is strongly supported by the WRRT data. With the WRRT test, care was taken for the procedure to be double masked, for the lens options to be presented randomly rather than in order of ascending lens power, and for the sequence of testing (ABCDDCBA) to be counterbalanced to control for practice or fatigue effects.

Fig. 2 and Fig. 5 reveal that participant age does not explain the plus addition that is preferred or that gives best performance. A detailed analysis of accommodative and binocular vision variables will be presented in a future publication that investigates the influence of different approaches to classifying and diagnosing binocular vision and accommodative anomalies. Suffice it to say here the pre-eminence of +0.75 is not explained by accommodative amplitude or lag, nor by any consistent pattern in the results of binocular vision tests. As noted in the preceding section, dry eye does not account for the pre-eminence of +0.75. It seems likely that two factors explain the supremacy of +0.75 over +1.25. First, the pre-presbyopic age range of the participants, 20–40 years, is likely to be relevant. We speculate that if participants had been included who were aged in the early presbyope category (e.g., aged 40–45 years), an age effect may have been apparent with +1.25 more likely to help this age group.

Second, the method that was used for the subjective preference and for the WRRT test was to view text on a computer placed at a distance that the participant selected as matching their typical viewing distance for digital devices. This viewing distance was measured (but was not recorded) and ranged from approximately 40−110 cm. For the participants where the monitor was further than 80 cm, the lens power of +1.25 would have been too strong. Also, proximal accommodation is likely to play a role, so even for some participants for whom the text was placed at a distance close to 80 cm, the combined effect of +1.25 lenses and proximal accommodation may have caused them to be over-plussed with this lens.

Our recruitment concentrated on office workers, with some university students also included. Most of these people predominantly used desktop computers, with a significant minority using laptops and very few predominantly using tablets or smartphones for work. We speculate that if our sample had included more people whose use of digital devices predominantly involved tablets or smart phones, which tend to be held closer,42 we would have found more participants preferring the +1.25 option.

The Wilkins Rate of Reading Test (WRRT)49 is designed to assess the impact of visual factors on reading and requires only very basic reading skills, so performance is minimally affected by language skills. The WRRT has been used in a variety of research studies.45,51–56 Early research with this test used a “within subjects” improvement in performance of 5% as a meaningful improvement.49 However, although in research using repeated measures statistics an improvement as small as 2.9% can be statistically significant,49 Kriss and Evans questioned the clinical significance of a 5% criterion and suggested a 10% criterion.62 More recently, a 15% criterion has emerged in research on coloured filters for visual stress as indicating an improvement that is likely to exceed any due to random variation.20 This has been applied retrospectively to earlier research19 and has been used in recent proposed diagnostic criteria for sensory visual stress.63 In the present research, we use both the 10% and 15% criteria. The fact that approximately one quarter of our population with DES read more than 10% faster with low adds indicates a performance benefit in some cases.

Previous research has used the WRRT to investigate the effect of low refractive corrections on reading performance.45 This study showed that correcting small hypermetropic refractive errors, up to +2.00D, in pre-presbyopes does not significantly improve performance at the WRRT. This study differed from the present research in that size lenses were used as the control intervention, although this is unlikely to be a factor because the size lenses were shown to be inert.52 A more likely explanation of the absence of a treatment effect of low powered convex lenses in that research in contrast to the presence of an effect in the present research, is that in the previous study participants were not selected as having symptoms whereas in the present research all participants had symptoms of DES.

A strength of the research is the large sample size of over 100 individuals. Also, the individuals came from a variety of workplaces and, unlike some previous research on DES, were not predominantly drawn from a young student population.

It is possible that a cycloplegic refraction may have detected latent hypermetropia in a few cases that was not detected by the non-cycloplegic refraction used in this research. However, this seems unlikely for a population aged 20–40y. Additionally, the research sought to produce results that are transferable to routine optometric practice, where cycloplegia is typically not used in the optometric assessment of pre-presbyopic adults.

Participants viewed the WRRT through standard full-aperture trial lenses in an optometric trial frame and it is conceivable that aberrations or prismatic effects could have influenced the results or impacted on binocular vision. This seems unlikely because the trial frame was horizontally and vertically centred to align with the visual axis. The size of the WRRT paragraphs (7 × 3.5 cm) does not require eye movements that are far off-axis. Additionally, most participants had low refractive errors (Table 4). The standard paper version of the WRRT was used because it was not possible to use the type of monitor each patient habitually viewed.

In the WRRT the participant is only required to read the text for a maximum of one minute at a time. The test designer implied that the visually intensive nature of the WRRT allows it to detect a beneficial effect that would otherwise require longer test periods,64 and this seems to be the case for investigating sensory visual stress, where the symptoms typically only occur after reading normal sized text for longer periods.65 However, a limitation of the present research is that there may be long-term benefits from low adds in DES that are not identified in this study. Also, for adult skilled readers it is possible that there is a “ceiling effect” with the test, which could mean that small variations in motivation/attention have a disproportionate effect on performance compared with the interventions under investigation. However, the significantly faster performance with +0.50 and +0.75 additions indicates that such a confounder did not greatly impair the usefulness of the test.

The main goal of this paper is to investigate whether, in a broad cross-section of symptomatic pre-presbyopic adult computer users, low adds are helpful, in addition to full distance refractive error correction. Our finding that in many cases low adds are helpful, raises questions about prescribing criteria and the prevalence in our sample of accommodative and binocular vision anomalies. These questions are not straightforward to address because of the lack of agreement on optometric prescribing criteria66–68 and will be the subject of a further publication.