A randomised exploratory investigation of the effects of Attention vs Working Memory Training on cognitive performance and everyday functioning following stroke

Difficulties with attention are common following stroke and are associated with poor outcome. Home-based online cognitive training may have to the potential to provide an efficient and effective way to improve attentional functions in such patients. Little work has been carried out to assess the efficacy of this approach in stroke patients, and the lack of studies with active control conditions and rigorous evaluations of cognitive functioning pre and post training means understanding is limited as to whether and how such interventions may be effective. Here we compare the effects of 20 days of active cognitive training using either novel Selective Attention Training (SAT) or commercial Working Memory Training (WMT) programme, versus a waitlist control group, on a wide range of attentional and working memory tasks, as well as on self-reported everyday functioning. We demonstrate separable effects of each of the active training conditions, with SAT leading to improvements in both spatial and non-spatial aspects of attention and WMT leading to improvements only on very closely related working memory tasks. Both training groups reported improvements in everyday functioning, which were associated with improvements in attentional functions, suggesting that improving attention may be of particular importance in maximising functional recovery in this patient group.

Prior to their first assessment session participants were randomised into one of three 197 conditions: WMT, SAT, or WL. At this time, the WL patients were further randomised to 198 WMT or SAT to be completed at the end of the WL period. The entire randomisation 199 sequence was completed by PP prior to the recruitment of the first participant and then 200 subject numbers (and their corresponding conditions) were allocated by PP in a sequential 201 order to participants as they became available without any prior knowledge of the individual. 202 After completing their initial assessment in their own homes, participants in training 203 conditions were shown how to log in to the relevant websites and navigate through the tasks. 204 They were asked to try and complete the training each weekday for the next 4 weeks (20 205 sessions). Participants were encouraged to get in touch with the project team if they 206 experienced any difficulties accessing the tasks. In addition a weekly phone call was   . This task required participants to verbally report the identities of as many letters 235 of a pre-specified target colour (either black or white) as they could from arrays of briefly 236 presented targets and non-targets, whilst maintaining central fixation. Each trial followed 237 essentially the same pattern. An initial red fixation cross flashed on and off a grey 238 background at a rate of between 150 and 230 ms four times. An array of letters was then 239 presented along with the fixation cross for 150ms before being replaced by the fixation cross three to the right, of fixation. From these conditions, 3 separable attentional parameters 250 (closely related to those defined in TVA, but using simplified formulae) were defined: 251  Absolute spatial bias; the relative extent to which performance is preserved on a 252 particular side of space in the presence of competing target information on the other 253 side of space. To examine this, we compared relative reduction in performance 254 between the 3T and 6T conditions for items presented on the left versus right sides of 255 space, using the following formula:

256
Absolute spatial bias=ABS 0.5-( ( 6 3 ⁄ ) Where pcorr is the proportion of targets correctly identified in that condition.     The training batteries were internet based and completed in participants' own homes.

327
Following an initial induction they were completed without assistance from the research 328 team. The batteries shared some essential core features, namely: that they were adaptive and 329 therefore designed to keep patients working at their maximal ability, and that trial by trial Visual Search. In each trial an abstract shape was presented on the screen for a few seconds.

356
It was then replaced by an array of objects (S2 Figure) and the participant was asked to judge 357 whether any exactly matched the original shape. Difficulty was manipulated by increasing the 358 similarity of the objects to the target along dimensions of shape, size, colour and texture. The  Participants played the task for 3 minutes in each training session.

378
Button Sorting. On each trial of this set-shifting task a shape was presented upon which the 379 participant was asked to make a speeded judgment based on a rule also presented on the screen (S5 Figure). If the rule was 'shape' the participant had to indicate whether the shape 381 most closely resembled a circle or square by clicking on an arrow pointing to one of two 382 reference shapes (circle and square) that were coloured red and yellow (the color was 383 irrelevant to the 'shape' rule). If the rule changed to 'color' the participant had now to click 384 on the arrow pointing to the correct color and ignore the shape of the reference. Difficulty confirm that all ongoing and related trials for these interventions are prospectively registered.
Complete data were analysed from twenty of the twenty-three participants. Of those whose 404 were omitted, two were removed having suffered subsequent neurological events between 405 initial assessment and final assessment and one had to drop out owing to family 406 circumstances. One-way ANOVAs were carried out to see whether the 3 groups differed on 407 background measures. No significant differences between the groups were observed for age  Thus SAT alone appeared to have a beneficial impact on spatial awareness. As our SAT was focussed on improving selective attention it may be expected that any 470 attentional affects on awareness may stem from improvements in top-down control (α').

474
The K' capacity measure might have been expected to have been influenced by both WMT  What predicts reductions in disability? A key question is whether self-reported improvements 543 were related to objective changes in cognitive function. A regression using 'pre-test core 544 symptoms', 'change in absolute bias', 'change in K', 'change in Dot Matrix performance' 545 and 'change in Variability' as predictors of pre-post test change in core symptoms indicated 546 that these variables explained 63% of the variance (R 2 =0.63, F(5, 28)= 7.74, p<0.001). As 547 may have been expected, 'pre-test core symptom score' was a significant predictor (β=0.44, stimulating activities may help to improve alertness [48]. It is possible, therefore, that 669 providing a daily structure within which patients were helped to focus on a cognitively  The success of any intervention is dependent not only upon the potential for improvement 686 following treatment, but also upon how practical and tolerable it is for patients. Here patients' 687 ability to cope with navigating to websites, logging in etc. was good and attitudes to both 688 interventions were generally positive, with a good proportion of patients feeling it was 689 worthwhile continuing after the study. In accordance with this, despite the time commitment 690 of the study, drop-out rates were very low. A caveat is that this sample was recruited from a 691 panel of individuals who have already indicated that they are motivated to take part in 692 research. It remains to be seen whether such good compliance would be seen in an unselected 693 population of stroke patients.
The results so far indicate some specific effects of the two types of training and some 696 generally positive effects from both compared to WL. The specific training effects are well 697 controlled in terms of exposure to training, interaction with the experimenter and the 698 knowledge of being engaged in training hypothesised to be helpful. However, interpretation 699 of the more general effects, is limited by reduced stimulation in the WL and potential 700 expectancy effects. To a degree this is offset by the finding that reductions in spatial bias and 701 improved K' variability over the course of the study predicted changes in self-reported 702 disability, suggesting that improvements in attentional functioning could be key to reducing 703 disability. Of course the reverse causality also remains a possibility. An active and plausible 704 control condition hypothesised not to be beneficial is required to clarify these issues.

705
It is generally accepted that the majority of spontaneous recovery occurs within the first six  In summary, our study provides evidence that cognitive training is feasible in stroke patients, 711 and can lead to both specific improvements in cognitive functions and more general 712 reductions in self-reported disability. Further work is required to examine whether such 713 effects can be replicated in a larger sample.