An electronic technetium-99m-diethylenetriaminepentaacetic acid glomerular filtration rate spreadsheet with novel embedded quality assurance features.

BACKGROUND: Critical clinical decisions are made on the basis of the glomerular filtration rate (GFR) measured using technetium-99m-diethylenetriaminepentaacetic acid (DTPA) administration, followed by multiple time-point plasma sampling. As GFR studies rely on few data points and produce a single result, they are prone to technical errors that may remain inconspicuous.
OBJECTIVE: We describe a data analysis worksheet that provides real-time quality control (QC) indicators and evaluate our initial clinical experience.
METHODS: Two hundred and forty-six consecutive GFR studies carried out at our clinics were included. Our protocol used plasma samples at 2, 3, and 4 h after injection of technetium-99m-DTPA. Duplicate plasma samples, background samples, and aliquots of an activity dilution standard were counted. Times were logged for injection and dilution standard preparation, blood sampling, and counting. Data were entered into a custom GFR analysis spreadsheet that flagged QC in real time at warning and error levels, including QC of the expected ratio between dilution standard counts-activity ratio (CARs) measurements, which was newly introduced to our clinic. The prevalence of QC events was analyzed in three phases: baseline, training, and evaluation (n=31, 69, and 146, respectively).
RESULTS: From the baseline and training phases (n=100), CAR reference values were determined for each of two sites. In the absence of the CAR QC indicator, errors were present in 5/31 (16%) examinations, but with QC indication decreased to 7/146 (5%) (P<0.05), suggesting that the real-time QC information guided the technologists to ensure proper standard preparation and sample handling, as intended. Improvements in other QC measures were also noted, resulting in an overall error rate reduction from 23 to 8%.
CONCLUSION: Real-time analysis of redundant information as a component of the GFR worksheet ensures quality results, but training of technologists and interpreting physicians is essential for optimal utilization of these QC indicators.

Introduction

Glomerular filtration rate (GFR) is a well-established metric of renal function 1 used to guide vital clinical decisions 2–5. Although several methods are available for quantifying renal function 6,7, serial plasma sampling following the administration of technetium-99m (99mTc)-diethylenetriaminepentaacetic acid (DTPA) has been shown to be a practical and accurate technique 8,9. This GFR method can also be combined with gamma-camera imaging to derive differential function indices, thereby enabling evaluation of individual kidneys 10.

Administration of a known amount of DTPA intravenously is followed by blood sampling at serial time points. Plasma is separated from blood and the concentration of activity is measured and compared with the initially administered dosage. The rate at which DTPA is cleared from the plasma, or GFR, can then be calculated. A ‘corrected GFR’ value is also customarily reported by normalizing the patient’s body-surface area (BSA) to that of a standard-sized individual 4.

Although methods of obtaining precise GFR measurements have utilized as many as 10 blood samples to capture multi-exponential clearance, a more feasible ‘slope-intercept’ method incorporates only two to four blood samples starting at two hours following injection. The slope-intercept method has been shown to characterize the terminal DTPA clearance phase with sufficient accuracy and precision to be used in clinical practice, although additional corrections have been proposed to further improve accuracy 11–13. Simpler single-blood-sample methods have also been described 14–17, and a recent review has shown that the Fleming method 17 can achieve equivalent performance to the slope-intercept method 18, especially in pediatrics 19. However, single-blood-sample GFR may be inaccurate below 30 ml/min/1.73 m2 18,19, making it ill-suited for a general population consisting of renal failure cases, and it has not yet been endorsed ubiquitously 8.

Although these DTPA clearance examinations have been characterized and validated in specialized laboratories, the reliability of any individual test needs to be ensured when ported to routine clinical practice. It is of concern that tests that rely on discrete measurements, such as the DTPA GFR method, are subject to technical errors or blunders that corrupt accuracy, and that the results may not clearly indicating the presence of these errors. In a recent study, McMeekin et al. 20 evaluated the effectiveness of previously published quality control (QC) methods for the detection of clinically significant errors. They concluded that QC methods based on expected volume of distribution and agreement between different model results (e.g. single-sample vs. slope-intercept) were not useful error predictors. The correlation between multiple time blood samples was a moderately useful QC indicator, especially at high GFR values, but insufficient on its own. They emphasize that there is no substitution for careful measurements and their recording, along with the use of well-established physics-based QC.

Duplicate sampling and measurement can be useful for detecting random errors and blunders by exploiting physics-based relationships between data; these strategies have been utilized previously to improve the quality of GFR examinations 12,14,20. In this paper, we have extended this concept by leveraging other known interdependencies in the measured data that are used to flag inherent discrepancies indicative of probable technical errors. An integrated approach, encompassing all exam variables including dilution and measurement of standard, has, to our knowledge, not been described previously. A further novelty of our technique, intended to improve the ease and accuracy of performing this examination in our routine clinical practice, is that we have embedded the QC flags into the technologist’s electronic spreadsheet, so that QC checks are incorporated automatically into the clinical workflow with minimum added effort. In this communication, we describe our GFR measurement protocol, related QC methods, and describe our experience during introduction and subsequent routine use of the method in a busy clinical environment.

Methods

This work encompasses all GFR studies carried out between 13 May 2015 and 13 February 2018, at either of two affiliated clinics (Table 1). Studies carried out before 1 September 2015 were used to establish baseline QC performance metrics. Studies carried out between 1 September 2015 and 15 March 2016 (inclusive) were used to phase in the finalized GFR calculation spreadsheet and train the staff on the QC indicators. Training included the following: (a) updating of the study protocol (including intended meaning of QC indicators and appropriate staff response actions), (b) presentation of the revised protocol to technologists at the in-service meeting at each site, (c) presentation of the revised protocol at departmental grand rounds, (d) e-mail communication to all technologists and physicians, and (e) one-on-one consultation of staff members with the department physicist. Studies carried out after 15 March 2016 were used to evaluate the performance of the method with QC fully implemented. In this paper, the three periods are referred to as the baseline, training, and evaluation phases, respectively.

Table 1

Scintillator counters and parameters

As a clinical quality assurance improvement initiative, this project was exempt from requiring research ethics board approval, as confirmed by our local research ethics board.

Clinical workflow

All aspects of the studies, including both in-vivo and in-vitro laboratory work, were carried out by one of the certified nuclear medicine technologists on duty on the day of the exam, as per regular work assignment. The clinical workflow is highlighted in Fig. 1 and follows guidelines by the British Nuclear Medicine Society 8. The times of all radiopharmaceutical administration, blood draws, and counting of activity were initially hand-recorded from a single clock on a printed worksheet in real time; data were transcribed subsequently into the electronic spreadsheet as detailed below.

Fig. 1

GFR workflow. “dupl.” indicates that samples were taken in two duplicates. Imaging is an optional component of this workflow. Solid arrows indicate physical manipulations, dashed arrows indicate data flow and gray lines indicate timed events. QC, quality control; 99mTc-DTPA, technetium-99m-diethylenetriaminepentaacetic acid.

Glomerular filtration rate protocol

Dose calibrators and well counters at each of our clinical sites (Table 1) are subjected to routine daily QC to ensure consistency with respect to reference sources. Patients were reviewed for counter indications including previous administration of radiopharmaceuticals, expanded body space (e.g. ascites, oedema), and hydration therapy. Patient height and weight were measured and recorded at the start of the exam. Patients were instructed to drink 500 ml of water and to void before commencement of the study. Activity in the injection syringe was assayed in the dose calibrator using the lowest possible range setting both before and after administration of radiopharmaceutical to account for residual activity in the syringe and IV access setup. Radiopharmaceutical injection setup was a butterfly IV access with a four-way stopcock enabling a 10 ml 0.9% saline flush immediately after radiopharmaceutical injection. Imaging of the injection site at ∼45 min after injection was included to rule out extravasation. In our practice, the clear majority of studies follow IV administration of ∼370 MBq (10 mCi) of 99mTc-DTPA and incorporate imaging over the subsequent 45 min. A minority of patients were studied without imaging and were administered a reduced ∼37 MBq (1.0 mCi) dosage of radiopharmaceutical.

In all cases, venous blood samples for the calculation of GFR were withdrawn at 2, 3, and 4 h after injection using a preinserted, saline-flushed, angiocatheter (20 or 22 G) in the opposite arm used for radiopharmaceutical injection. Each withdrawal consisted of a saline flush of the catheter, draw of ∼3 ml blood to be discarded, draw of ∼6 ml blood sample into an evacuated heparinized vial, and postdraw catheter flush with 10–20 ml saline. All vials were labeled immediately with patient identifiers and time of blood draw.

Dilution standard

To relate the injected activity measured in the dose calibrator to the sample activity measured in the well-counter, a dilution standard was prepared for cross-calibration. Initially, a 500 ml volumetric flask was filled with ∼400 ml of tap water. A syringe of ∼12 MBq of 99mTc-pertechnetate was measured in the dose calibrator both before and after discharge into the flask. The flask was topped up with tap water to the 500 ml meniscus line, capped, and agitated rigorously. 100 µl aliquots of the dilution standard were pipetted in duplicate into well-counter vials after a second, vigorous mixing.

Sample preparation and counting

At the completion of the study, the venous blood samples were centrifuged at 2000g for 10 min or longer as needed. 100 µl aliquots of plasma were pipetted in duplicate from each of the venous blood samples into well-counter vials. Two additional blank vials were used to measure the background counts. All 10 vials were then counted according to the settings delineated in Table 1.

Data entry and analysis

During the course of each patient’s study, all data, including time, activity, and demographics, were recorded on a temporary worksheet. Following counting of the 10 samples, all data were transcribed from the temporary worksheet into an electronic spreadsheet that was implemented in Microsoft Excel 2010/2016 (Microsoft, Redmond, Washington). Details of the embedded QC checks and calculation of GFR are described below (Fig. 2 and Table 2).

Fig. 2

Example spreadsheet for the calculation of glomerular filtration rate (GFR), body surface area-corrected GFR, and quality control checks. Shaded boxes indicate input data. The last two columns consist of a visual indicator of quality control (checkmark, x, or exclamation marks) and an explanation thereof. BSA, body surface area GFR, glomerular filtration rate.

Table 2

Glomerular filtration rate spreadsheet data input, dependent calculations, and related quality control checks

Glomerular filtration rate calculation

All dose calibrator and well-counter measurements were decay corrected to the time of DTPA injection on the basis of the recorded times and the 6.01 h physical half-life of 99mTc. Methods of GFR calculation have been described previously 4 and are summarized here briefly as they relate to our QC process. A mono-exponential washout function [N(t)=N0e−] was fitted to the plasma count data, N(t), where N0 represents the extrapolated plasma counts at the time of injection (t=0) and λ represents the rate of DTPA washout from the plasma (min−1). GFR was then calculated using the following equation:where A is the activity administered to the patient (in Bq), Astd is the activity of the dilution standard (in Bq), and Nstd is the count of the 100 µl dilution sample. The dilution standard serves to relate the N0 sample counts into an activity concentration (Bq/ml). GFR corrected for standardized BSA 21 was calculated as GFR×1.73 m2/BSA using the Du Bois and Du Bois equation 22:

Quality control testing

QC embedded into the spreadsheet was designed to evaluate all entered data (Table 2) using predefined threshold values to indicate three outcome levels: pass, warning, and error (Table 3). Warnings directed staff to re-evaluate their procedure, but the values were considered tolerable after review; a study with an error indication was considered not acceptable under any circumstance. Warning and error levels were selected to be sensitive and specific to QC issues with intended ∼1 : 20 and ∼1 : 1000 false-positive rates of alarm, respectively. Warning levels were therefore more stringent than error levels. Values falling within the low and high warning thresholds passed QC. Four specific analyses were carried out:

Table 3

Glomerular filtration rate spreadsheet quality control checks, nominal value, and warning and error limits

(1) Repeat count agreement

Significant differences between paired sample background counts (N1 and N2) were tested using a z-score assuming a Poisson distribution (i.e. variance equal to the mean, ) of counts:

Similarly, for standard and plasma sample pairs, a z-score incorporating σP=1% pipetting variance (measured in-house by a novice user using an analytical balance as a reference), was used:

To account for four independent tests (standard and three plasma), a Bonferroni correction factor of 3 was applied to reduce the overall study false-positive rate.

(2) Sequential plasma sample agreement

Agreement of the mean counts of the three consecutive plasma samples with mono-exponential washout was evaluated using the Pearson correlation (r) goodness-of-fit parameter.

(3) Counts–activity ratio (CAR) testing

The CAR (counts/kBq) was calculated as the ratio between dilution standard mean number of sample counts [Nstd=(Nstd1+Nstd2)/2] and the net activity used to prepare the standard (kBq), Astd, after decay correction. This formulation is not to be confused with the efficiency of the well counter (counts/s/Bq) as CAR is intentionally designed to be influenced (and therefore reflect) by multiple potential sources of operator error and instrumentation faults associated with the well counter, dose calibrator, pipette, and volume flask. Because standard dilution volume (500 ml), pipetting volume (100 µl), and counting time were used (Table 1), a simple ratio could be used to calculate CAR when using decay-corrected values:

Deviation of CAR from the reference truth, µ, was reported as a percent error.

The reference CAR, µ, and SD, SD, were determined for each site independently using the composite baseline and training (i.e. pre-evaluation) data and are listed in Table 1. To account for outlier values, µ was determined as the median CAR, C, and SD was estimated using SD=interquartile range/1.35. The evaluation phase data were used to validate our estimate of µ and SD.

(4) Range checking

To eliminate errors of data entry, several miscellaneous checks were performed to ensure that the recorded data were within expected ranges, including patient BSA, injected activity according to protocol, residual activities in syringes, and blood withdrawal timing. In addition, timing of patient injection, blood withdrawals, and sample counting were tested for proper sequence.

Quality control indicators

QC test limits incorporated into the GFR calculation spreadsheet are summarized in Table 3. During the evaluation phase, a QC summary indicator was added beside the final GFR metrics (as shown in Fig. 2) that reported the most severe status of all the above-described tests with the following accompanying wording:

Pass – ‘QC test passed’.

Warning – ‘QC test at tolerance – Results may be inaccurate’.

Error – ‘QC test failed – Results are likely inaccurate’.

On the basis of the direct QA feedback embedded in the spreadsheet, when warnings or errors were encountered, the technologists were prompted to recheck their data, often resulting in repeated pipetting or self-correction of other errors, without consulting with the physicist. In these cases, the interim spreadsheet was replaced by the subsequent correction. In other cases, especially when errors were refractory, the technologist staff consulted with the department physicist in real time, which always resulted in successful remediation of the examination. As a general rule, the physicist did not review study results in real time during study accrual unless consulted.

Retrospective review of quality control

At the conclusion of the evaluation phase, all GFR studies carried out in our department were reviewed by the department physicist to assess changes in quality. The results presented here therefore represent an evaluation of routine technologist-driven and physician-driven performance. Incidents of warning and error indicators were recollected from the clinically reported GFR spreadsheets. Rates of warning and error indications were tabulated retrospectively for each study phase. Incidences of interim errors, which were self-corrected, were not captured.

Statistical analysis

Event rates were compared using a two-proportion z-test. Except for CAR, continuous variables were summarized as mean±SD, with precision expressed as the standard error (SE=SD/mean×100%). Comparisons between the means of continuous variables were performed using the Student t-test, and variances were compared using the f-test. P values less than 0.05 were considered statistically significant. Data analysis was carried out using Matlab 2017b (MathWorks, Natick, Massachusetts, USA).

Results

Two hundred and forty-six GFR exams were performed in 207 patients; 39 patients underwent two examinations. The first 31 studies were carried out during the baseline phase, 69 studies were carried out during the training phase, and the remaining 146 studies were carried out during the evaluation phase. Studies were roughly evenly split between the two performance sites (110 vs. 136 studies).

Patient demographics and GFR measurements are summarized in Table 4. All patients were adults (>18 years old). For all 246 studies, GFR measured 99.6±38.4 (9.6–209) ml/min [mean±SD [minimum–maximum)], with BSA-corrected GFR of 88.5±31.7 (10.1–218) ml/min/1.73 m2. There were no significant differences in the means between the three phases of the study (P>0.24) and the referral patterns also did not change; thus, we assumed that the patient population remained similar. 92% of the studies incorporated imaging, whereas 8% did not. Injected activities were 367±89 (233–442) MBq for studies with imaging and 89±111 (20–392) MBq without imaging. The activity concentration of the dilution standard measured 16.0±5.9 (1.5–47.6) kBq/ml at the time of sample counting.

Table 4

Patient demographics

Counter–activity ratios

Reference CAR and precisions, as reported in Table 1, were determined from the pre-evaluation (baseline and training) phase data. Standard error estimates (precision) between the two sites varied significantly even after excluding outlier values (P=0.013); hence, we decided to use ±10 and ±15% deviations in CAR measurements for warning and error levels, respectively, instead of normal distribution-derived z-score limits. CAR deviations from reference values are shown as percent error over the course of the study in Fig. 3.

Fig. 3

Count–activity ratio (CAR) errors over the course of the study. The vertical lines indicate the boundaries of the three study phases and the horizontal lines represent the warning and error indicator threshold levels.

In the final evaluation phase of the study, CARs were 9.84±0.20 and 15.06±0.49 counts/kBq for the two respective sites. There was no statistically significant change in average or precision (median or SD) values between the pre-evaluation phases and the evaluation phase for either site. Precision remained superior at site 1 than at site 2 (2.01 vs. 3.24%, respectively, P<0.001).

Quality indicator rates

The resulting warning and error rates at each phase of the study are reported in Table 5. For most indicators (adherence to protocoled blood sampling times, agreement between duplicate counting samples, and curve quality of fit), the incidences of errors and warnings were too small to test for significant change between study phases. CAR error rates were higher in the pre-evaluation phases than in the evaluation phase (14 vs. 5%, P=0.01), as were the combined warning and error rates (18 vs. 5%, P=0.002), showing improved adherence to quality standards during the evaluation phase. The summary QC indicator similarly indicated higher error rates in the pre-evaluation phases than in the evaluation phase (17 vs. 8%, P=0.02). Rates of combined error and warning also decreased between these corresponding phases (33 vs. 22%, P<0.05).

Table 5

Quality control test results from the clinical glomerular filtration rate worksheet review

Discussion

This work describes implementation of a rigorous QC program for nuclear GFR measurements studies, and evaluates its adoption in a routine clinical practice. The GFR protocol involves several manual steps that are prone to human error that may not be apparent in the final results, which has motivated our attempt to incorporate automated quality testing into the clinical worksheet. These tests are therefore intended to identify potentially inaccurate data, prompting real-time review and correction, with minimal incremental workload for the staff. Our experience has shown that our GFR protocol and QC efforts can be implemented seamlessly and effectively in a busy clinical setting.

A key novelty of our methodology is testing of the dilution standard against an expected count-rate dependent on the well-counter and dose-calibrator efficiencies, which should remain constant over the life of the devices 8. The incremental work required by the technologists to perform these QC measures is negligible, primarily requiring recording of the time of sample counting.

Retrospective review of data from the baseline phase of this study showed that in the absence of a QC indicator, CAR errors were common (16% of studies), justifying the need for more rigorous QC. With QC feedback, in the final evaluation phase of our study, errors that would have otherwise gone unnoticed were identified and corrected before interpretation of the GFR measures, as indicated by lower error rates (5%). CAR errors were driven by dilution standard preparation and/or pipetting errors – both of which proportionately affect GFR results.

Our experience has shown that training of technologists and interpreting physicians was required to facilitate understanding of QC indicators and how they should be utilized. On the basis of these lessons, we iteratively adjusted the spreadsheet to provide more intuitive feedback, with the final version including a summary QC indicator. Change management requires ongoing education and development of a quality-oriented culture. Stringent QC is paramount for accurate and precise results, and improves clinician confidence.

Data redundancy

QC testing is possible through redundant information such as duplicate samples and utilization of the goodness-of-fit (Pearson r) parameter of the mono-exponential washout model. CAR leverages redundancy between the activity of the dilution standard and the counts of its samples by taking advantage of fixed counter and dose calibrator efficiencies, and protocoled pipetting and standard volumes. In many GFR protocols, the dilution standard traditionally serves a dual purpose: (a) converting counts into activity concentration, (b) performing decay correction between the time of patient injection and the time of sample counting. By assuming that the counter and dose calibrator efficiencies are stable over time and applying accurate time-based decay correction of the radiopharmaceutical activity, it is possible to carry out an accurate GFR study in the absence of a standard altogether. We opted to continue the use of a dilution standard, and maintain redundancy, to enable quality testing that encompasses pipetting and counter accuracy.

Counts–activity ratio error limits

Originally, we sought to determine CAR warning and error limits on the basis of empirical measurements of the natural variability from the pre-evaluation phases of the study and assuming a Gaussian distribution. On the basis of these results (Table 1), the respective CAR error and warning limits (associated with 5 and 0.1% false-positive rates) would have been 5.0 and 8.2% for site 1 and 7.3 and 12.1% for site 2. In practice, we encountered much higher warning and error rates, prompting us to use arbitrarily selected 10 and 15% warnings and errors limits that we believed were acceptable within the medical context 20. We speculate that the main sources of CAR errors were variations in pipetting techniques amongst operators and/or standard preparation as the instrumentation (i.e. dose calibrator, counter, pipets, volumetric flask) precisions far exceed these measures. Nevertheless, we are hopeful that with ongoing efforts, the magnitude of the CAR errors will continue to decrease, enabling us to apply more stringent error limits in the future.

Additional quality control measures

Sanity checks on entered and calculated data can be effective safeguards for detecting errors such as use of erroneous units or transcription errors. In our spreadsheet, in addition to the tests listed in Table 3, we also ensured that patient identification numbers contained eight numerical digits and that time stamps followed the sequence shown in Fig. 1.

Other physiologic parameters could be considered for QC testing, including volume of distribution, extra-cellular volume, and time to 50% washout (T1/2) 4,23. We did not include these error measures as variabilities, even in normal populations of potential donors, were relatively high compared with the physics-based quality metrics described in this work, and certainly were outside of ‘normal’ values in patients with renal disease. QC on the basis of physiologic metrics is difficult in a heterogenous clinic such of ours, consisting of patient populations ranging from healthy, potential kidney donors to patients with severe renal dysfunction. As an example, errors between volume of distribution estimated using BSA and Brochner–Mortensen corrected GFR and BSA alone have been proposed for QC 23, but had 24.0% variability, which is too large to be of clinical utility as a QC measure 14. This conclusion is consistent with those of McMeekin et al. 20.

The British Nuclear Medicine Society guidelines recommend that sample counts exceed 10 000 counts 8. In this work, 40/246 exams had one or more sample that did not conform to this requirement. Of these, 18 tests did not include imaging and therefore had a lower injected activity of ∼37 MBq. Of the remaining 22 exams, 20 exams had duplicate counts that summed to more than 10 000 counts. The two remaining cases had sample counting performed on the following day (using 10 min/sample) because of technical issues. In one case, duplicate sample counts agreed to within the 95% confidence interval. The single remaining case indicated proper injected activity (374 MBq) and no evidence of subcutaneous injection (by imaging of the injection site), but did have a warning for low agreement between duplicate 3 h plasma counts – nevertheless, correlation between time samples was high (r=0.998), indicating a good model fit. The use of a broader energy window in the well-counter settings (Table 1) would serve to increase count statistics, but would necessitate derivation of new reference CAR values.

The British Nuclear Medicine Society guidelines 8 recommend r=0.985 as a lower limit for the correlation coefficient between multiple sample count rates. We applied this value as a warning threshold and used r=0.95 as the error threshold, recognizing that lower r values are expected in low GFR cases, in which the gradient of the fit is low 20. Future work could evaluate the use of a GFR-dependent lower limit for r or another metric for consistency between multiple samples.

Persistent errors

Although we sought to eliminate all errors in the GFR exams performed by our clinic, some remaining errors did persist even into the evaluation phase (Fig. 3 and Table 5). Five of the seven studies with indications of CAR errors were in the first months of the evaluation phase, following which the prevalence of this error decreased markedly. One of the remaining two errors was associated with an intermittent counter system hardware issue that triggered servicing of the device, showing the utility of the CAR test in detecting equipment errors. The last CAR error (−61%) was not handled in a timely manner and therefore no corrective action was undertaken – an error margin was included in the patient report with recommendation for a repeat study.

In a single study an error was indicated for late withdrawal of a blood sample which resulted from tardy return of the patient to the department. This study also had a low correlation between the plasma sample counts error indicated (r=0.911). The physicist was consulted on this case in real time and it was determined that there were no other technical issues with the study and that the low correlation was associated with the patient having very low plasma clearance (GFR=10 ml/min), consistent with a clinical history of chronic kidney disease. An additional blood sample was recommended for validation, but the patient could not comply with this request.

Of all the examinations in the evaluation phase, 22% had one or more warnings, which was an improvement over the pre-evaluation phases (33%). Roughly 30% of the warnings were associated with blood withdrawal timing deviating from the protocol. Because the actual time of the blood draw is to a large degree accounted by the curve fitting, this type of deviation from the protocol may not result in GFR inaccuracy 18, unless the error was because of an error in time keeping 8,20. Through ongoing quality initiatives, we will endeavor to further reduce these and other warning rates.

Our clinic is staffed by over 25 technologists, eight physicians, and multiple trainees, requiring training of over 35 individuals across two practice locations. Most staff are shift workers, further complicating face-to-face communication with all individuals. Furthermore, because of the relatively small volume of GFR studies in our clinic, it is likely that this study is carried out infrequently by some individuals. Thus, some individuals have had limited opportunity for hands-on experience with our new QC measures. One potential means to achieve more rapid improvement in quality is to improve the proficiency of select individuals in the group by restricting the performance of GFR studies to a subset of technologists and physicians. Improved communication and training techniques may also have been able to achieve better compliance with quality assurance directives, and are the subject of future efforts.

Translation to other sites

Translation of our QC methods to other institutions requires the configuration of site-specific parameters, including CAR, size of dilution volume, pipetting volume, and protocolled injection activity, not to mention warning and error thresholds. An editable GFR calculation spreadsheet is included in Supplementary Material 1 (Supplemental digital content 1, ), and includes a settings page where these parameters may be configured.

Measurement variability is expected because of physical, instrumentation, and human factors and can be estimated from prospective, ad-hoc measurement experiments (e.g. measurement of activity, dilution volume, pipetting volume, and counting statistics). Furthermore, these evaluations should also incorporate multiple technologists to account for interoperator variability. An alternative approach, such as that described in this work for the CAR metric, is to analyze retrospective data that inherently reflect measurement variability resulting from all the above-mentioned sources.

Alternative protocols

Our clinical protocol consists of three blood samples at 2, 3, and 4 h after a radiopharmaceutical injection. Alternative protocols consisting of different numbers of blood samples, different sampling times, and different modeling techniques have been proposed previously 8,18. The QC measures described in this work could be adapted easily by a qualified physicist for alternative protocols.

Limitations

A limitation of this work is the relatively small number of GFR exams, which is because of the infrequent performance of this exam at our institution (∼100 studies/year). The use of previous data to increase the exam pool was not possible as the protocol used for earlier studies did not include recording of exact times of sample counting, precluding CAR calculations. Nevertheless, appropriate statistical testing could identify significant differences that support our conclusions.

Conclusion

We have developed, validated, and reported on a rigorous quality assurance workflow for GFR studies using the 99mTc-DTPA methodology and based on a dedicated calculator spreadsheet that is readily implementable in a routine clinical practice. This real-time QC reduced the frequency and magnitude of errors, which increased confidence in GFR measurements. Efforts to provide education and develop a culture of quality were required to ensure its optimal use.

Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal's website, www.nuclearmedicinecomm.com.