Improvement in activities of daily living after visual training in patients with homonymous visual field defects using Goal Attainment Scaling

2.1 Patients and study design

This study was part of a larger project approved by the ethical committee CMO Arnhem-Nijmegen in conjunction with the 1964 Declaration of Helsinki.

Forty stroke patients were included following written informed consent. Inclusion criteria were: homonymous visual field defect (confirmed by Goldmann perimetry during intake) due to post chiasmatic stroke, time since lesion at least 10 months (chronic phase of stroke), between 18–75 years of age and the ability to undergo (f)MRI scanning. Patients were excluded if they had other visual anomalies that could not be corrected for (such as macular degeneration, cataract, glaucoma) or visual neglect. Patients were excluded if they showed signs of visual neglect assessed by the line bisection task (systematic deviation towards the ipsilesional visual field).

Patients were trained following a randomized controlled crossover design. Each patient trained successively two predefined regions of the visual field at home during two rounds of training. At least one of these regions was located in the visual field defect. In the first three cohorts one of the training periods was directed to the visual field defect and the other training period was directed to the intact hemifield. Following three cohorts of ten patients we modified the training procedure, because in some patients the control training (directed to the intact hemifield) reversed the field improvement of the preceding test training. Therefore, the last cohort of ten patients followed a training paradigm with different training regions. In the last cohort of ten patients, the two training rounds were offered to one of two complementary halves of their defect (directed training) while the other half was not trained (and received undirected training). Note both training paradigms provided a directed and an undirected training round of the entire field defect.

For each training round the stimuli were only shown within the training region of interest (ROI). Therefore, the region of the visual field that was targeted in that round is evaluated as ‘directed training’ and the other part of the visual field that was not targeted during that round was evaluated as ‘undirected training’. In each region we compared the effects of both training rounds, one of which was thus ‘directed’ and the other ‘undirected’ (see Fig. 1). The patients trained 1 hour a day, 5 days a week during 8 weeks to complete at least 40 hours of training per training round.

OPEN IN VIEWER

Fig.1

Study design. Two predefined regions in the visual field were trained successively (training 1 and training 2). The rectangles shown in bold represent the targeted region (evaluated for a ‘directed’ training effect) and the diagonal textures represent the untargeted region (evaluated for an ‘undirected’ training effect).

Prior to the inclusion of the first patient, a training scheme for all patients was created using Matlab (Mathworks Inc.). For each cohort of 10 patients’ numbers (J01–J10; J11–J20; J21–J30; J31–J40), the order of training rounds was randomly assigned to a patient number (Table 1). The patients were included in order of registration on our website.

In all patients, training started 2–6 months after intake if all inclusion criteria were met. Both GAS and QoL questionnaires were completed prior to the first training round and evaluated after the 2nd training round.

For perimetry of the visual field defect we applied both Humphrey (blinded) and Goldmann perimetry (not blinded, because of insufficient staffing). Baseline perimetry occurred just prior to training and was repeated following the first round of visual training (at least 40 hours) and again following the second round of training (at least 40 hours) (Fig. 1). The directed and undirected training were evaluated for effectiveness by the difference between two succeeding perimetry assessments.

A detailed description of the training at home and the custom made training unit can be found in Elshout et al. (2016) (Elshout, van Asten, Hoyng, Bergsma & van den Berg, 2016). The patients’ head was stabilized with a chin and headrest and the viewing distance was fixed at 40 cm. Patients performed a visual discrimination task at many different locations within the targeted part of the visual field. The patient maintained gaze binocularly on a ring (diameter = 0.5 degrees) at the center of the screen. During stimulus presentation (7 s), patients shifted attention covertly (i.e. without shifting the eye from the fixation point) towards the training location and responded using the keyboard. At the training location, a point was shown displaced in a clockwise (CW) or counterclockwise (CCW) direction relative to a line extending from the fixation point like the hand of a clock. Alternatively, the stimulus consisted of a patch of dots rotating in a CW or CCW direction about the patch centre. The patient’s forced-choice response (CW or CCW) was indicated by arrow keys on the keyboard.

All training stimuli were scaled following the cortical magnification factor (CMF) (Cowey & Rolls, 1974) to provide equivalent cortical stimulation at all eccentricities.

2.3 Assessment of changes in visual field after training

We used two complementary clinical perimetry methods. Humphrey perimetry samples at fixed locations visual detection thresholds. Goldmann perimetry precisely localizes the blind field’s border. Monocular perimetry was performed of the eye opposite to the affected hemifield. We applied the sita-fast 30-2 program of the Humphrey Field Analyzer II in a standardized way. During Goldmann perimetry we used a white IV size stimulus with maximum luminance (4e = 1000 apostilb (asb) (≈ 318 cd/m2)) against a white background with a luminance of 31.5 asb. Fixation was continuously monitored via the spyglass by the experimenter and checked at random occasions using blind spot localization (Heijl & Krakau, 1975) that is also used during Humphrey perimetry.

Mean sensitivity of Humphrey perimetry was calculated for each trained and untrained region. Thus, we were able to investigate the change in sensitivity through directed training (i.e. change in sensitivity of the trained visual field) and undirected training (i.e. change in sensitivity of the untrained visual field). We incorporate all locations measured during the 30-2 program of the Humphrey perimetry. The mean sensitivity changes of the locations that fall into the trained region is referred as change through directed training. The mean sensitivity change of all other complementary locations is referred as change through undirected training.

Furthermore, we created isopters based on the Goldmann measurement and expressed the change of the visual field border in Equivalent Cortical Surface Gain (ECSG) (Bergsma et al., 2012; Elshout et al., 2016). Performance on many visual tasks follows the CMF, meaning that high CMF (central vision) and low CMF (visual periphery) corresponds to high and low performance respectively. ECSG transforms the shift of the field border’s eccentricity, into an equivalent mm cortex shift, using the human CMF function. Thus, we quantified the notion that more functional recovery occurs when 1deg of central vision is recovered than 1deg of peripheral vision.

2.4 Assessment of changes in activities of daily living and quality of life after training

We used three different validated questionnaires for evaluation of QoL: the 36-item short-form Medical Outcomes Study (MOS-SF-36), the Stroke-Specific Quality of Life scale (SSQoL) and the Utrecht Scale for Evaluation of Rehabilitation-Participation (USER-P).

The MOS-SF-36 is a general health questionnaire for assessment of care outcomes (Ware & Sherbourne, 1992). It is composed of 36 items, organized into eight domains. For each domain the item scores are coded, summed and transformed into a scale ranging from 0–100, where 100 is the best possible rating. We used the scoring method referred as RAND 36-Item Health Survey 1.0. Each domain will be evaluated separately.

The SSQoL is a questionnaire specially designed for stroke clinical trials (Williams, Weinberger, Harris, Clark & Biller, 1999). It consists of 49 items each scored on a 5-point scale. Items are averaged to obtain a mean score, and a higher score (where 5 is the best possible rating) indicates a higher level of QoL.

The USER-P questionnaire is an instrument to rate participation (Post et al., 2012). It consists 32 items and measures three aspects of participation: 1) frequency of behaviour, 2) restriction in participation and 3) satisfaction with participation. We calculated sum scores for the Frequency, Restrictions and Satisfaction scales and converted each sum score to a score on a scale ranging from 0 to 100. Higher scores indicate better levels of participation (higher frequency of activities, less restrictions, greater satisfaction).

In addition to these general QoL questionnaires we used a personalized approach (GAS) to assess improvements of ADL. Each patient formulated three SMART (Specific, Measurable, Attainable, Realistic and Timely) goals prior to training which were checked by a therapist experienced in applying GAS. The therapists were provided a visual field map depicting the field defect and the sector of the visual field that might improve through training. In this way the therapist could assess whether the formulated goal by the subject was realistic or not. For instance, a patient with a visual field defect close to the center could never attain a goal that requires peripheral vision. Preferably, the formulated goals are quantitative in some aspect so that the therapist can confirm the GAS level by an objective measurement (e.g. number of columns seen in excel sheet, number of times the patient stumble into objects, time needed to read article in newspaper). The therapists were blinded to the perimetric effects. By using a 6-point Likert-scale it can be evaluated to what extent the goals are attained after rehabilitation. Each patient started with a baseline performance assessment that was scored as –2. The achievement after training could be worse (–3), the same (–2), somewhat better but goal not achieved (–1), goal achieved (0), better than goal (+1) or much better than goal (+2). The overall GAS T-scores were calculated by applying the formula of Kiresuk and Sherman (Kiresuk & Sherman, 1968) with an expected correlation of the goal scales of 0.3:

GAS = 50 + 10 ∑ ( W i X i ) / ( 0.7 ∑ W i 2 + 0.3 ( ∑ W i ) 2 )(1)

Where W_i = the weight assigned to the i^th goal (if equal weights, W_i = 1) and X_i = the score achieved at the i^th goal. We considered each goal as equally relevant and thus assigned 1 to W_i.

To compare the pre- and post-training scores for the questionnaires and GAS we used a Wilcoxon paired samples test.

For Goldmann and Humphrey perimetry we used a one sampled t-test with a hypothesized mean of 0 to test for an overall training effect. For Goldmann perimetry we used a paired sampled t-test to test for an effect of intervention compared to no intervention. Regression analyses were used to study linear relationships between effect measures. We considered a p < 0.05 as statistically significant. The False Discovery Rate (FDR) was used to control for multiple comparisons. SPSS V.20.0 (SPSS Inc, Chicago, Illinois, USA) was used for analyses.

3.1 Training effect assessed by perimetry

Humphrey perimetry revealed an overall sensitivity increase of 0.79 (SEM±0.20) dB after both training rounds (whole field analysis). Goldmann perimetry showed an average ECSG of 5.82 (SEM±0.94) mm after both training rounds (Fig. 2). Thus, both Humphrey and Goldmann showed a positive improvement of the visual field that significantly (t(27) = 3.961, p < 0.005; t(31) = 6.178, p < 0.005, respectively) differed from the null hypothesis of no training effect (i.e. hypothesized mean of 0). For the Goldmann perimetry we corroborated that assumption because the defect border did not change between intake and pre-training border (ECSG = 0.27±0.58) and we confirmed that ECSG was significantly larger after training than after no intervention (t(29) = 5.061; p < 0.005).

OPEN IN VIEWER

Fig.2

Visual field change. Training effect for Humphrey perimetry reported in sensitivity increase (dB) in the whole field after both training rounds. Goldmann perimetry describes the border shift of the defect contour (ESCG) after both training rounds. Clearly, the visual field was improved following the complete training.

The mean scores of both the SS-QoL and MOS_SF-36 questionnaires improved (Table 2; p = 0.009 and p = 0.024 respectively). Also the restriction (p = 0.039) and satisfaction (p = 0.002) scales of the USER-P significantly improved after intervention. Satisfaction was the only significant subscale after correction for multiple comparisons.

Two out of eight subscales of the MOS-SF-36 health survey significantly improved after training: ‘Role limitations due to physical health’ (p = 0.022), and ‘Emotional well-being’ (p = 0.002). All other items were not significant (‘physical functioning’: p = 0.165 z = –1.39; ‘Role limitation due to emotional problems’: p = 0.943, z = –0.07; ‘Energy/fatigue’: p = 0.062, z = –1.86; ‘Social functioning’: p = 0.125, z = –1.54; ‘Pain’: p = 0.937, z = –0.08 and ‘General health’: p = 0.843, z = –0.20). After correction for multiple comparisons the only significant subscale was ‘Emotional well-being’. No significant correlations were found between field recovery (expressed in Goldmann ECSG or Humphrey dB) and QoL improvement assessed by these three questionnaires (p > 0.359).

After both training rounds, the therapist assessed the progress on the goals that patients had formulated. In 7 patients this involved only 2 goals; three or more were evaluated in the other (26) patients. Of all evaluated patients, 73% improved on the GAS T-score (9 of the 33 subjects showed no improvement) and 52% of all patients attained at least one of their formulated goals (GAS score of ‘0’ for a particular goal after evaluation). We found a significant improvement in T-score after both training rounds (Fig. 3A; Z = –4.294, p < 0.005). Visual field improvement and GAS improvement varied considerably among patients. Patients that showed no improvement on the GAS (n = 9) had a low averaged field gain (Goldmann: M = 3.13 mm, SEM±0.97 mm; Humphrey: M = 0.13 dB, SEM±0.29 dB). In contrast, patients with improvement (n = 24) had a relatively high averaged field gain (Goldmann: M = 6.03 mm, SEM±1.17 mm; Humphrey: M = 0.85 dB, SEM±0.28 dB). We thus investigated whether the individual outcome of the GAS is related to the magnitude of the training effect. Indeed, the improvement on the GAS was linearly related to the reduction of the field defect following both training rounds combined as measured with Goldmann perimetry (Fig. 3B; p = 0.042; r = 0.381). No such effect was found for Humphrey perimetry (Fig. 3C; p = 0.335; r = 0.189).

OPEN IN VIEWER

Fig.3

Goal Attainment Scaling. (A) Median GAS T-score prior to treatment and after both training rounds. N = 33. (B) Regression between Goldmann field reduction and improvement on GAS (N = 29). The filled circles represent the data of the last cohort (N = 5). (C) Regression between Humphrey sensitivity increases in the affected hemifield and improvement on GAS (N = 28). The filled circles represent the data of the last cohort (N = 5).

Our finding that a directed training results in higher field gain compared to an undirected training (Elshout et al., 2016), made us wonder if we could address separate contributions of directed and undirected training to GAS. Thus, we performed a multiple linear regression to find the dependency of the GAS score on field improvement by directed and undirected training components for either Goldmann or Humphrey perimetry. The Goldmann data showed an overall trend (F(2,26) = 3.170, p = 0.059, r = 0.443). Patients’ predicted GAS score is equal to 9.5 + 0.9(directed training)+1.7(undirected training), where both directed and undirected training are coded as Goldmann ECSG. Remarkably, only undirected training significantly contributed to the GAS improvement (p = 0.111, beta = 0.332; p = 0.021, beta = 0.496 respectively).

Humphrey perimetry also revealed an overall trend (F(2,24) = 3.119, p = 0.063; r = 0.454), but only the directed training contributed significantly to this relationship (p = 0.025, beta = 0.441) not the undirected training (p = 0.768, beta = –0.055). Patients’ GAS score is predicted by 10.18 + 5.39(directed training) – 0.9(undirected training), where both directed and undirected training are coded as mean Humphrey sensitivity increase in dB.

Together these data suggest that the increase of the visual field and the improvement of visual sensitivity during both directed and undirected training may contribute to the probability that patients achieve their set goals.