A retrospective chart review was conducted in compliance with the Health Insurance Portability and Accountability Act (HIPAA) and with institutional review board (IRB) oversight. The study was deemed to be exempt from full IRB review. Charts from a single center were identified for patients that underwent a revision TKA after metal LTT sensitivity testing over a 3-year period from January 1, 2015, to December 31, 2017. Patients were referred for metal LTT sensitivity testing after other causes of a painful knee arthroplasty had been excluded (i.e., infection, malalignment, or instability). Patients underwent revision TKA to a hypoallergenic component or a standard component, based on the results of sensitivity testing. Patients testing positive for metal LTT sensitivity were revised to a hypoallergenic component, while those testing negative were revised for arthrofibrosis. Following revision, patients returned to the clinic at an interval of 6 weeks, 5 months, and 12 months for clinical assessment. Patients were excluded if they had incomplete pre-revision data sets. Results from metal LTT sensitivity testing, functional outcomes, pain scores, and satisfaction were analyzed.
Metal LTT sensitivity testing
Patients were referred for metal LTT sensitivity testing after all other causes of a painful knee arthroplasty had been excluded, including infection, malalignment, or instability via laboratory, radiographic, or mechanical testing, respectively. For sensitivity testing, venous blood was drawn in the out-patient clinic and sent to a third-party laboratory (Orthopedic Analysis, Chicago, IL). Lymphocytes were isolated and incubated for 5 days with a panel of metal allergens (Additional file 1: Figure S1). Reactivity was measured by quantifying lymphocyte proliferation using radioactive markers. Lymphocyte reactivity was normalized based on a known stimulant and reported on a stimulation index (SI). Each allergen received a numeric SI score and was categorized as non-reactive (SI < 2), mildly reactive (SI 2–4), reactive (SI 4–8), or highly reactive (SI > 8) (a sample report is provided in Additional file 2: Figure S2) . Reports for each anonymized patient were entered into a database. Maximum allergy SI for each patient was identified.
Patient function was assessed by physical exam and survey reporting. During the initial evaluation, range of motion (ROM) of the operative knee was measured in flexion, extension, and total ROM. Instability, measured by varus and valgus laxity, was also recorded. Repeat ROM measures were made at each post-operative visit. Patients were also provided a survey prior to revision and at each post-operative visit to assess their function. The primary outcome variable was walking distance, measured as unable to walk, indoors only, less than 5 blocks, 5–10 blocks, and unlimited. Walking distance categories were entered into the database as an ordinal variable.
Patients were surveyed prior to their revision regarding multiple characteristics of their TKA pain. They were asked how long the pain has persisted, the intensity of pain 6 months ago on a pain intensity numerical rating scale (PI-NRS) (scale 1–10; 10 being the worst), and current intensity of pain (scale 1–10; 10 being the worst). Change in pain was reported by calculating the percent change in PI-NRS for each patient between time points . Graphically, pain scores were reported as the mean of the PI-NRS over time for each sensitivity group. At each post-operative visit, patients were asked their current intensity of pain. They were also asked to characterize their current pain, measured as none, mild or occasional, mild or occasional stairs only, mild or occasional stairs and walking, moderate occasional, moderate continual, or severe. Pain characterization was converted to a numeric scale of 1–7 (1 = none to 7 = severe) and entered into the database as an ordinal variable.
At each post-operative visit, patients were asked if they were satisfied with the revision and if they would elect to have the surgery again. Satisfaction was measured as extremely satisfied, very satisfied, moderately satisfied, slightly satisfied, not at all satisfied, and it is too early to tell. Satisfaction was converted to a numeric scale of 1–5 (1 = extremely satisfied to 5 = not at all satisfied; it is too early to tell was excluded from analysis) and entered into the database as an ordinal variable. Patient’s willingness to have the surgery again was entered into the database as a binomial variable of yes = 1 or no = 0 (it is too early to tell was excluded from analysis).
The anonymized dataset was grouped by metal sensitivity as a binary independent variable (non-reactive or reactive) and as an ordinal variable (non-reactive, mildly reactive, reactive, or highly reactive). Data were tested for normal distribution and equal variances before analysis. Means and ranges or standard deviations (SD) were calculated for continuous variables (e.g., age). Categorical variables (e.g., sensitivity level) were expressed as number and percentage. For comparing non-reactive to reactive groups, data were analyzed by means of unpaired two-tailed t tests for continuous dependent variables and Mann-Whitney U tests for ordinal dependent variables. One-way analysis of variance (ANOVA) was used to test for differences between metal LTT sensitivity levels and pain or ROM; Bonferroni post hoc test were used when appropriate. Pearson’s chi-squared test was used to compare the frequency of willingness to repeat the surgery between the non-reactive or reactive groups. Paired t tests were applied for within-group comparisons between pre-revision pain and ROM and post-revision follow-ups. Post-operative pain and ROM were normalized as percentages of baseline (pre-operative) measurements and compared between groups using unpaired two-tailed t tests. Correlations between metal LTT sensitivity and percent change in pain or pre-revision pain were tested using Pearson’s product-moment correlation (for metal LTT sensitivity score) or Spearman’s rank correlation tests (for metal LTT sensitivity level). Data was aggregated and analyzed using open-source R . Statistical significance was set at p ≤ .05.