Welcoming the robot therapist

As written, the New England Journal of Medicine article (Robot-Assisted Therapy for Long-Term Upper-Limb Impairment after Stroke) claims quite modest results for robot therapy. Between the lines, though, lie powerful hints of the impending robot revolution. (The article is free and I encourage you to read it.)Researchers recruited Veterans Administration (VA) patients who'd suffered strokes an average of 5 years earlier. Volunteers were randomized to 12 weeks of robot-assisted therapy, intensive comparison therapy, or usual care (medical management and some rehabilitation services available to all patients).  The primary outcome was change in motor function at 12 weeks as measured by the Fugl-Meyer score. Patients were also evaluated after 6, 24 and 36 weeks.Results showed that robot-assisted therapy did not significantly improve motor function at 12 weeks compared to usual care or intensive therapy, although there was significant improvement on the Stroke Impact Scale (SIS). At 36 weeks robot-assisted therapy and intensive therapy performed about the same, and both were superior to usual care.So what's the big deal? Actually a few things:

• Intensive therapy --whether performed by a robot or a human-- demonstrated its superiority to usual care. The intensive therapy delivered by therapists in this trial was so intensive: >1000 movements in a session compared to 45 for typical stroke treatment --that it's really only available under controlled conditions with a researcher standing over the therapist with a stopwatch. If you want this kind of intensive therapy for yourself you'd better hire someone with a whip (or bring John Henry back from the dead and take the hammer out of his hand).
• A close read of Table 2 (Changes in Primary and Secondary Outcomes at 12 Weeks) strongly suggests better results for the robots than the authors are willing to claim explicitly. The left side of the table measures robot-assisted therapy vs. usual care. The right columns measure robot-assisted therapy vs. intensive comparison therapy. The performance of the robot group is much better on the right side of the table than the left. If we compare the robot numbers on the right with the usual care numbers on the left, the comparisons look much better for the robots, including a 5 point improvement on the Fugl-Meyer score (which is extremely impressive for someone 5 years after a stroke).So why is the table set out this way? It appears that patients were only enrolled in usual care for the first 16 months of recruitment, whereas the recruitment period for robot-assisted and intensive therapy continued for 24 months. I don't understand why the usual care recruitment was stopped --perhaps the study was on a tight budget? In any case it's likely that the therapists operating the robots followed a learning curve during the study and were better able to operate the robots in the second part of the study than the first, which is the portion  of the study analyzed on the left side of the table. The authors state that the robot manufacturer (Interactive Motion Technologies) had "no role in the study," which could mean the therapists had to figure out the robot for themselves and would not have been as good at first.The fact that this information is presented in the same table appears to be a tacit acknowledgment that the researchers also believe in the comparison I am suggesting.
• Labor savings were large. One-hour sessions required only 15 minutes of contact with the therapist for the robot versus 60 minutes for the intensive comparison therapy. That has important implications for those worried about workforce shortages over the next decades.
• Overall costs were equivalent in all 3 treatment groups (including usual care) at about $15K despite the ~$10K cost of using the robot. That means the robot group patients used only one-third of the medical resources of the usual care group ($5K versus $15K) while making significant improvements in health status. That's incredibly good news and must have been a big surprise to the researchers.

An accompanying editorial (Brain Repair after Stroke) acknowledges the accomplishments of this study and emphasizes the broader implications.

In the bigger picture, the potential for robotic therapy after stroke remains enormous. Robotic devices can provide therapy in different functional modes, a point that was not examined by Lo et al. Robots work in a consistent and precise manner and over long periods without fatigue. They can modulate timing, content, and intensity of training in reproducible ways, with a reduced need for human oversight. Robotic devices can also measure the performance of patients during therapy. In addition, robot-based therapy can interface with computers in brain-stimulation treatment or to provide simultaneous cognitive training.

You can also check out a video of the robot in action.