STATISTICAL ANALYSIS
All statistical analyses were performed using SAS version 9.2 (SAS Institute Inc.). Descriptive statistics such as means, standard deviations, medians, and 25% and 75% percentiles (Q1 and Q3, respectively) were provided for the continuous variables. Numbers and percentages were provided for the categorical variables. For statistical testing, Student’s t-tests were applied for the continuous variables and chi-square tests for the categorical variables. All the statistical tests were two-tailed. The sample sizes in this database study are remarkably large, such that differences that are not clinically important could still be statistically significant at the conventional alpha level of 0.05. Thus, we applied a more stringent requirement of the alpha level of 0.0001 to identify highly statistically significant results. The number and percentage of patients within each cohort with at least 1 instance of an adverse event of interest (those adverse events commonly expected for patients who receive IA injections for knee OA) were assessed. Times to TKR during the 36-month post-index period were analyzed and compared among different cohorts. Any patients who had not undergone TKR by the end of the post-index period were considered censored at 36 months. The Kaplan-Meier method was employed to model survival curves with time to TKR data, and log-rank tests were used to compare survival curves among different cohorts. A Cox proportional hazards model (PHM) was used to model the risk of TKR with a pre-specified set of covariates adjusted for baseline attributes, such as age, gender, comorbidities, and pre-index healthcare costs. Hazard ratios with 95% confidence intervals were used to examine the measures of event risk.
RESULTS
PATIENT CHARACTERISTICS
Applying study selection criteria to the claims database yielded 50,389 patients (Figure 1), providing an ample sample size for the statistical analysis. Only patients with evidence of knee OA and use of HA injections (the index medication of interest) were selected, regardless of whether they received TKR during the post-index period. The requirement for a knee OA diagnosis during the 12-month pre-index period resulted in the significant attrition of patients, with 584,956 patients being excluded. Among the 50,389 patients who received HA for treatment of knee OA, 36,260 (72.0%) received a single course of treatment, 8709 (17.3%) received 2 courses, 3179 (6.3%) received 3 courses, 1354 (2.7%) received 4 courses, and 887 (1.8%) received ≥5 courses of treatment.
Comparison of baseline characteristics among the 5 IA HA cohorts showed the fairly similar baseline characteristics of all cohorts (Table 1). Geographic region, physician specialty, and opioid use showed differences among the cohorts. Cohorts with ≥5 HA courses presented lower proportions of patients from Southern US states, patients seeing orthopedic surgeons, and patients using opioids than cohorts with fewer HA courses.
PROCEDURES OF INTEREST
An analysis of the procedures patients received after HA treatment initiation showed that higher numbers of HA treatment courses resulted in lower proportions of patients receiving TKR within 3 years after HA treatment initiation (Table 2). With an increasing number of HA treatment courses, the proportion of patients with TKR within 3 years post-index consistently decreased from 28.4% (for 1 HA course) to 5.0% (for ≥5 HA courses), with all differences being highly statistically significant (P < .0001). Similarly, partial knee replacement exhibited a similar trend, with the proportion of patients decreasing from 3.3% (for 1 HA course) to 0.8% (for ≥5 HA courses; P < .0001). Among the patients with TKR within 3 years post-index, increasing numbers of treatment courses correlated with increasing time to TKR, with a mean of 375.6 days (for 1 HA course) rising to a mean of 971.5 days (for ≥5 HA courses; P < .0001). On the other hand, patients with multiple courses of HA treatment were more likely to undergo radiologic examinations of the knee, arthrocenteses, and image-guided injections than patients with only a single course of HA treatment (P < .0001).
ADVERSE EVENTS
Arthralgia and joint pain in the knee were the most commonly recorded adverse events (Table 3). More courses of HA treatment were associated with higher rates of adverse events. Overall, the reported adverse events profile of repeated courses of HA treatment consisted of mostly common and mild adverse events and displayed no safety concern for patients with knee OA that was followed-up for 3 years. The causality of these adverse events directly related to HA injections vs a specific disease state cannot be determined from an administrative claims data set.
TIME TO TKR
Successive courses of HA led to high proportions of patients without TKR 3 years after HA treatment initiation. This result is evident in the Kaplan-Meier survival curves of time to TKR for different HA cohorts (Figure 2), with log-rank tests of multiple courses vs a single course of HA (P < .0001) showing highly statistically significance. Tabulation of proportions of patients without TKR by various time points showed that increasing numbers of HA treatment courses correlated with higher proportions of patients without TKR at almost all time points (Table 4); within 3 years post-index, 71.6% of patients in the 1 HA course cohort exhibited no TKR, whereas 95.0% of patients in ≥5 HA courses cohort presented no TKR. We also performed a multivariate Cox PHM (Table 5) to account for baseline characteristics of different HA cohorts with covariates when estimating the risks of receiving TKR. The results of the Cox PHM showed that multiple courses of HA treatment significantly decreased the risk of TKR (hazard ratio, 0.138 for ≥5 HA courses vs 1 HA course; P < .0001). Inspection of other highly significant covariates showed that being older, living in the Midwest region of the US (vs the Northeast), receiving pre-index corticosteroids, having an orthopedic surgeon as a treating physician (vs a general practitioner, a rheumatologist, or a physical medicine and rehabilitation specialist), experiencing hypertension or hyperlipidemia, and higher pre-index total healthcare costs were associated with an increased risk of TKR (all P < .0001). Vascular disease and high CCI scores were associated with a decreased risk of TKR (P < .0001).
Continue to: Discussion...