Hemodialysis Access Flow Measurements Using Flow Dilution and In-Line Dialysance Are Not Interchangeable for Hemodialysis Access Flow Surveillance

Abstract

In Response: Our recent article1 compared hemodialysis access flow (QA) measurements using two methods: flow dilution (FD) and in-line dialysance (DD). The difference in the measured QA between the two methods was calculated, and the mean difference was reported as 29 ± 19 ml/min. Because the mean difference (rather than the difference itself) was provided, the standard error of the mean (SEM ∼19 ml/min) rather than standard deviation (SD ∼175 ml/min) was reported. Reporting the SEM follows the original Bland-Altman article2 because the intent here is to report the precision of estimating the mean. In addition, the SEM and SD are related by a simple formula, and hence, one can easily derive the SD (∼175 ml/min) from the SEM (∼19 ml/min). Some statisticians may prefer reporting the SD, and we understand that some readers may assume that the number reported might be the SD. However, by careful examination of our data (e.g., Figure 2), it would be apparent (given the data range shown in the figure) that the 19 ml/min cannot be the SD. We thank Dr. Krivitski (from Transonic Systems Inc., the manufacturer of the DD equipment) for pointing out that 19 ml/min cannot be the SD. The SD of 180 ml/min that he calculated is very close to the actual SD = 175 ml/min of our data. Because, it is common to show the mean ± 2 SD in Bland-Altman plots, we have added two dashed lines to our original Figure 2 showing these limits.Figure 2.: Bland-Altman analysis of flow dilution (FD) and in-line dialysance (DD). In the average, FD only slightly over estimates Q A compared with DD. The dashed lines show the mean and mean ± 2 SD of the difference (FD − DD).Comparison of our study1 results with an earlier similar study3 needs to be done with care. More specifically, the latter study reported a mean QA difference between the same two methods (using older models of the equipment) as 135 ± 229 ml/min. It is apparent from Figure 4 of Lacson et al.3 that 229 ml/min is the SD not SEM. Comparing QA differences between these two studies1,3 shows that the mean difference was 29 and 135 ml/min, respectively, and the SD was 175 and 229 ml/min, respectively. Although the SD values are comparable, the relatively smaller mean difference in our study1 suggests a possible improved agreement between DD and FD measurements. The mean difference represents a bias error that can be, relatively easily, reduced by calibration as earlier suggested.1 Comparison of these results suggests that the two studies1,3 are consistent and confirmatory. An important consequence of QA agreement/disagreement is the impact on imaging referral decisions. It can be seen that even with the larger QA bias reported,3 the QA differences resulted in nonsignificant differences in imaging referral decisions.3 Part of the referral differences is due to the QA bias in one technique vs. the other. The reduced QA bias found in our study1 suggests even less significant referral differences. Hence, it is our opinion that “FD and DD may be used interchangeably” as stated by Whittier et al.1 However, more studies are suggested to further address the referral issue.1,3 Regarding Dr. Krivitski's comment that we mistakenly “referenced a 1998 Gotch paper4 instead of Lacson et al. 2008,” we respectfully disagree because our study did not mention or refer to Gotch et al.4 In conclusion, there are no calculation or reference errors as Dr. Krivitski's letter suggested. Figure 2 was updated to reduce potential reader confusion about the value of SD (Figure 1). Although clear from context, we agree that it is better to clearly state that the numbers reported represent SEM not SD. Hansen A. Mansy William L. Whittier Richard H. Sandler Rush University Medical Center Chicago, IL