This was a pilot randomized clinical trial.
Inclusion and exclusion criteria
We recruited participants aged 50 years and older who are scheduled for primary TKA. Although the initial inclusion criteria was participants aged 65 years and older, due to poor enrollment, we changed the inclusion criteria in April 2016. We screened participants in the Preoperative Assessment Clinic at Wake Forest Baptist Medical Center. A trained study coordinator screened potential participants for baseline mobility status, using the mobility assessment tool-short form (MAT-sf). Those participants who score at or below the 50th percentile, based on our prior preoperative study data, were considered eligible (e.g. MAT scores ≤58 for men and ≤ 50 for women). Exclusion criteria included patients who were undergoing knee replacement for indications other than osteoarthritis, revision surgery or bilateral surgery, scheduled for emergency surgery; afraid of water or are not willing to undergo water exercise; having major deficits in hearing or vision; currently exercising more than 3 times a week; participating in another clinical trial; or cannot understand the questionnaires and directions due to cognitive impairment or language barriers. Following these tests, participants underwent a standard medical workup at the discretion of the PAC’s attending physician, including American Society of Anesthesiologists (ASA) physical status scoring, and Revised Cardiac Risk Index (RCRI) classification. For each participant, information regarding demographic characteristics, comorbidities, medications were gathered.
Blood pressure, pulse, height and weight were assessed. All enrolled participants were assessed by a trained study coordinator on 1) Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain, stiffness, and physical function , 2) physical function using the Short Physical Performance Battery (SPPB), 3) self-reported mobility using Mobility Assessment Tool-short form (MAT-sf) 4) depression using Geriatric Depression Scale Short Form (GDS-sf), and 5) cognitive function using Montreal Cognitive Assessment (MoCA). The WOMAC (score 0–96) is a widely used and validated tool to assess pain, stiffness, and physical function in patients with hip and/or knee osteoarthritis. It has three subscales including pain (5 items, score 0–20), stiffness (2 items, score 0–8) and physical function (17 items, score 0–68) . The SPPB consists of repeated chair stands, balance testing, and 4-m walking speed (total score 0–12). It is known to predict subsequent disability, institutionalization, and mortality . The GDS-sf (score 0–15) is a 15 item screening tool for depression for older adults and it has been validated in the preoperative setting. Score of 0–5 is considered as normal, while a score of > 5 suggests depression . The MAT-sf is a 10-item computer based assessment of mobility using animated video clips . The 10 items in the MAT-sf cover a broad range of functioning. The items include walking on level ground, a slow jog, walking outdoors on uneven terrain, walking up a ramp with and without using a handrail, stepping over hurdles, ascending and descending stairs with and without the use of a handrail, and climbing stairs while carrying bags. It has been validated against measures of physical function, including the Pepper Assessment Tool for Disability, the Short Physical Performance Battery, and 400-m walk test among a population of older community dwellers. The Montreal Cognitive Assessment (MoCA) is a widely validated 30 points test that measures several domains of cognitive function, including short-term memory recall tasks, visuospatial abilities, executive function, attention, concentration, working memory, language, and orientation . Baseline blood were drawn at the first visit for high-sensitivity-C-reactive protein (hs-CRP), tumor necrosis factor-alpha (TNF-α) and interleukin-6 (IL-6) and, stored at − 80 °C until the completion of the study and all samples were analyzed together.
Data to calculate Charlson comorbidity  was obtained by chart review and the score was calculated.
After baseline assessments, all the enrolled participants were randomly assigned to the intervention or control group, using a pre-generated randomization list. The AEI were conducted in warm water (a minimum of 90 °F) at the Wake Forest Baptist Medical Center Therapy Pool, three times a week, and for 60 min each session until the scheduled surgery (4–8 weeks). The duration of the intervention was determined based on the duration between the preoperative visit with orthopedic surgeon and the surgery date. To maximize the enrollment, we designed the study that would not interfere usual practice pattern. All the sessions were supervised by an aquatic therapist. The AEI was designed to improve 1) range of motion, 2) muscle strength, and 3) resistance, resulting in improvement in mobility in this low-mobility population. The protocol consists of a warm-up (10 min); joint range of motion for flexibility and strength (20 min); low intensity endurance such as walking to prevent chilling and maintain and/or improve cardiovascular fitness (20 min); and end with a cool down (10 min). Depending on the participant’s tolerance, resistance equipment, such as a noodle and/or neoprene ankle cuff were added to increase the exercise intensity. The participants’ exercise log was kept by the aquatic exercise director to record the compliance to the protocol. A trained study coordinator called the intervention group participants weekly to assist with problem solving, reinforce progress, and to encourage adherence. Control group participants received a brochure on perioperative nutrition in addition to the standard cares, such as optimization of their underlying medical problems and perioperative infection prevention measures.
Reevaluation before surgery
Participants in both groups were evaluated 1 week prior to their scheduled surgery. Blood pressure, pulse, and weight were assessed. Symptoms of osteoarthritis, self-reported mobility, depression, and cognition using the WOMAC, MAT-sf, GDS-sf and MoCA were assessed by a study coordinator. The SPPB was administered by two designated trained study coordinator who were not supervising the aquatic exercise and were masked to the randomization status. At this time, follow up labs were drawn to measure hs-CRP, TNF-α, and IL-6.
Reevaluation after surgery
Participants in both groups were asked to return 4–6 weeks after the scheduled surgery. The follow up time was chosen so the visit can coincide the patient’s visit with their postoperative visit with their surgeon. Blood pressure, pulse, and weight were assessed. Osteoarthritis symptoms, self-reported mobility, depression and cognition, using WOMAC, MAT-sf, GDS-sf and MoCA were assessed by a study coordinator. Physical function was assessed using SPPB by two designated trained study coordinators who were not supervising the aquatic exercise and were masked to the randomization status. Follow up labs were drawn to measure hs-CRP, TNF-α, and IL-6. The follow up duration was decided based on the primary outcome of interest − 30 day National Surgical Quality Improvement Project (NSQIP) defined morbidity and mortality as well as hospital length of stay, ICU length of stay, delirium, and institutionalization.
The primary outcome of interest was 30 day National Surgical Quality Improvement Project (NSQIP) defined morbidity, including surgical site infection, wound disruption, pneumonia, unplanned intubation, pulmonary embolism, ventilator > 48 h, acute renal failure, urinary tract infection, stroke, coma, peripheral nerve injury, cardiac arrest, myocardial infarction, requirement of transfusion, deep vein thrombosis, sepsis and mortality , hospital length of stay, ICU length of stay, delirium, and institutionalization. The secondary outcomes were osteoarthritis symptoms using WOMAC, self- reported mobility measured by MAT-sf, physical function, measured by SPPB, depression, measured by GDS, cognition, measured by MoCA, and inflammatory profiles, including hs-CRP, TNF-α, and IL-6.
The power calculation for detecting the intervention effect on the unfavorable outcomes was done using the two proportion test. Assuming 20 participants in each group (in this study, the sample size is 20 in the intervention group and 23 in the standard care group), a two-sided test and a significance level of 0.05, we have 80% power to detect an odds ratio of 6.6 and 9 when the proportions of unfavorable outcome are 10 and 30% in the intervention group, respectively. We have limited power in this analysis. If we could increase the sample size to 100 per group in future studies, then the detectable odds ratios become 3.0 and 2.3, respectively. Additionally, we note that enrolling 20 per group provides 80% power to detect mean differences (via analysis of covariance) of 1.04, 0.88, 0.68, 0.39 standard deviation in outcomes between the two intervention groups, assuming a correlation coefficient between the outcome measure at baseline and at follow-up equals to 0.3, 0.5, 0.7, and 0.9, respectively, given a two-sided test and a significance level of 0.05. We have more power to detect the differences when the correlation coefficient is larger than 0.5. If we could increase sample size to 100 per group, the detectable differences are 0.47, 0.39, 0.31, 0.18 standard deviation in outcomes between the two intervention groups, assuming a correlation coefficient of 0.3, 0.5, 0.7, and 0.9, respectively. In general, the effect sizes are reasonably small when sample size is 100 per group for future trials.
All statistical analyses were performed using SAS software, version 9.4 (SAS Institute, Cary, NC). Sample means and standard deviations were computed for the continuous baseline characteristics, and counts and proportions were calculated for the discrete baseline characteristics according to intervention groups. For non-normally distributed characteristics, medians, 1st quartile, and 3rd quartile were also calculated. Logistic regressions were used to examine the associations between intervention groups and unfavorable outcomes, including postoperative complications, longer hospital stay of over 2 days, disposition to nursing home/rehab after adjusting for anesthesia time. Odds ratios (ORs) and their 95% confidence intervals (CIs) were presented. In order to best approximate the conditional normality assumption, the distributions of functional measurements were checked and transformed if needed. Differences in mean values of each functional measurement between intervention groups were estimated using repeated measures analysis of covariance with baseline outcome measure, intervention assignment, visit, and an intervention by visit interaction included in the model. Hypothesis tests for intervention effects at the follow-up visits were performed using contrasts. Overall comparisons between groups across follow-up visits were obtained using a contrast to compare average intervention effects across both follow-up visits. Raw mean and standard deviation were presented at baseline, adjusted least squares means and standard error were presented at post-intervention and post-operative visits.