In Reply:
We thank Drs. Gurunathan and Harrison et al. for their interest in our publication. To address Dr. Gurunathan's comments, we would like to point out that in making our statements we did not purely adopt or quote the findings presented by Sliva et al. ,1but we critically reviewed and interpreted them in the context of our study. Although the total number of patients in the study by Sliva et al. was 332, only 267—as reported in our paper—had their procedure performed during the same hospitalization (n = 241 staggered [i.e. , 4–7 days apart] and n = 26 sequential [i.e. , during the same anesthetic]). Because our study focused only on patients whose procedures were performed during the same hospitalization, we correctly identified this subgroup of interest (n = 241 + 26 = 267).2
Importantly, major complications occurred in four patients in the staggered group, whereas none occurred in the sequential group. Although the numbers in the study may not be sufficient to show statistical significance, major adverse events in the perioperative period are of great clinical concern. This is the reason why mortality was chosen as the primary outcome in our analysis. The importance of mortality and major complications is appropriately made evident by Dr. Gurunathan's comment, regarding their highest incidence in the staged bilateral knee arthroplasty patients, despite not reaching statistical significance as well. This issue gets to the heart of the problem when studying low-incidence outcomes, such as mortality, in studies with limited numbers because often authors conclude that the procedures can be considered safe based on underpowered results failing to show statistically significant differences between groups. With perioperative mortality being the primary outcome in our study, we tried to overcome the problem of small sample size by using the largest all-payer database available in the United States. Although our interpretation regarding the study of Sliva et al. may have not been in line with the authors' conclusion, who based their statements of safety on the occurrence of overwhelmingly minor complications, we believe that our independent interpretation of their findings regarding mortality and major complications is correct. We do not dispute, however, that by being more precise in our presentation, we could have avoided this miscommunication.
The sentence should read: “...in a study including 267 patients who underwent bilateral knee arthroplasty during the same hospitalization, Sliva et al. found that bilateral procedures performed 4–7 days apart were associated with higher incidence of mortality and major morbidity when compared with simultaneously performed procedures. No statistical difference could be shown however, likely because of low numbers.”
Dr. Harrison et al. posed questions regarding the validity of the Nationwide Inpatient Sample and its ability to produce nationally representative data for total knee arthroplasty procedures. We would like to refer the interested reader to the publication “Introduction to the Healthcare Cost and Utilization Project Nationwide Inpatient Sample” published by the Agency for Healthcare Research and Quality*for general background information on this database.
To answer their specific questions:
The total number of entries for hospitalizations for the years between 1998 and 2006 was 68,836,152. This means that of all hospitalizations, 0.97% were associated with primary knee replacement. One of the stated goals of the Nationwide Inpatient Sample is to provide data that allow for national estimates, which confirm confidence in this data source as shown by its wide use in the medical research field when seeking to provide nationally representative data. Further, the frequencies for a specific time frame published and derived from another nationally representative database—the National Hospital Discharge Survey—are very similar, providing another source of validation.3
Although differences in complications (in this case device-related) between unilateral and bilateral knee arthroplasty were found, we can only restate that no causal relationships can be established from these data, and thus, possible explanations for the findings have to remain speculative.
As explained in the article, databases of this kind are limited by the amount of variables they collect. As such, detailed information on laterality, patient choice, causality in decision-making processes, and procedures performed during different hospitalizations are not available. Thus, the very good points made by Harrison et al. regarding such cofounders cannot be addressed further in this study.
The total number of deaths was 73 (0.26%) in the simultaneous bilateral, 21 (0.29%) in the staged bilateral, and 845 (0.14%) in the unilateral group. The weighted national estimates for in-hospital mortality based on these entries were n = 354, n = 107, and n = 4,121, respectively.
As with any study, the results and conclusions have to be interpreted in the context of its design. Thus, definitions of bilateral knee arthroplasty and unilateral total knee arthroplasty as presented in the methodology have to be considered.
†Hospital for Special Surgery, New York, New York. memtsoudiss@hss.edu
