Effect of training on inspiratory load compensation in weaned and unweaned mechanically ventilated ICU patients.

BACKGROUND While inspiratory muscle weakness is common in prolonged mechanical ventilation, inspiratory muscle strength training (IMST) can facilitate strengthening and ventilator weaning. However, the inspiratory load compensation (ILC) responses to threshold loads are not well characterized in patients. We retrospectively compared ILC responses according to the clinical outcomes of IMST (ie, maximum inspiratory pressure [PImax], weaning outcome), in difficult-to-wean ICU patients. METHODS Sixteen tracheostomized subjects (10 weaned, 6 unweaned) from a previous clinical trial underwent IMST 5 days/week, at the highest tolerated load, in conjunction with daily, progressive spontaneous breathing trials. PImax and ILC with a 10 cm H2O load were compared in the subjects before and after IMST. Changes in ILC performance were further characterized (5, 10, 15 cm H2O loads) in the trained subjects who weaned. RESULTS Demographics, respiratory mechanics, and initial PImax (52 ± 26 cm H2O vs 42 ± 13 cm H2O) did not significantly differ between the groups. Upon enrollment, PImax significantly correlated with flow ILC responses with the 10 cm H2O load (r = 0.64, P = .008). After IMST, PImax significantly increased in the entire sample (P = .03). Both before and after IMST, subjects who weaned generated greater flow and volume ILC than subjects who failed to wean. Additionally, ILC flow, tidal volume, and duty cycle increased upon ventilator weaning, at loads of 5, 10, and 15 cm H2O. CONCLUSIONS Flow ILC at a threshold load of 10 cm H2O in ventilated, tracheostomized subjects positively correlated with PImax. Although PImax improved in both groups, the flow and volume ILC responses of the weaned subjects were more robust, both before and after IMST. The results suggest that ILC response is different in weaned and unweaned subjects, reflecting dynamic inspiratory muscular efforts that could be influential in weaning.