Professional Documents
Culture Documents
Camgfr V2: A New Model For Estimating The Glomerular Filtration Rate From Standardized or Non-Standardized Creatinine in Patients With Cancer
Camgfr V2: A New Model For Estimating The Glomerular Filtration Rate From Standardized or Non-Standardized Creatinine in Patients With Cancer
ABSTRACT
◥
Purpose: Management of patients with cancer, specifically Results: A total of 3,083 IDMS- and 4,612 non-IDMS–
carboplatin dosing, requires accurate knowledge of glomerular standardized creatinine measurements were obtained from 7,240
filtration rate (GFR). Direct measurement of GFR is resource
Introduction
The filtration function of the kidney is quantified as the glomerular
filtration rate (GFR). Knowledge of GFR informs clinical management
of many patients with cancer. For example, the dose of carboplatin, a
1
Cancer Research UK Cambridge Institute, University of Cambridge, Cambridge, well-established treatment for ovarian, breast, lung, and germ cell
England, United Kingdom. 2Cambridge University Hospitals NHS Foundation cancers, is calculated from GFR using the Calvert equation (1). GFR
Trust, Cambridge, England, United Kingdom. 3University of Cambridge, Cam- can be measured directly (mGFR) using the clearance of chemical
bridge, England, United Kingdom. 4Department of Oncology, Sahlgrenska
tracers, for example, chromium-51 EDTA (51Cr-EDTA; ref. 2). This
Academy at the University of Gothenburg, Gothenburg, Sweden. 5University
of Cambridge School of Clinical Medicine, Cambridge, England, United Kingdom.
approach is accurate and precise, but costly and not widely available. In
6
Cold Spring Harbor Laboratory, Cold Spring Harbor, New York. 7The Christie practice, GFR is therefore estimated (eGFR) by modeling of readily
NHS Foundation Trust, Manchester, England, United Kingdom. 8University of available clinical and biochemical data.
Manchester, Manchester, England, United Kingdom. 9Gynecologic Oncology Most eGFR models are based on the serum concentrations
Therapeutics, Kaiser Permanente, San Francisco, California. 10School of of creatinine, a metabolite of creatine, because it has robust
Medicine, University of St Andrews, St Andrews, Scotland, United Kingdom.
11
steady-state concentrations and is freely filtered in the glomerulus,
School of Mathematics and Statistics, University of St Andrews, St Andrews,
Scotland, United Kingdom. 12Columbia University, New York, New York.
with minimal active secretion (3). However, because there are
13
Northwell Health, New York, New York. several methods to measure creatinine, results can vary significantly
between centers for technical reasons (4). To reduce these differ-
Note: Supplementary data for this article are available at Clinical Cancer
Research Online (http://clincancerres.aacrjournals.org/). ences and thereby to harmonize clinical management decisions
based on eGFR, isotope dilution mass spectrometry (IDMS)-stan-
Corresponding Author: Tobias Janowitz, Cold Spring Harbor Laboratory, 1
dardized creatinine assays have been developed and are now widely,
Bungtown Road, Cold Spring Harbor, NY 11724. Phone: 516-367-8422; E-mail:
[email protected] but not universally, used (5, 6). Another source of creatinine
variability is due to human physiology: underlying disease
Clin Cancer Res 2021;27:1381–90
processes, relative muscle mass, and/or ethnicity influence the
doi: 10.1158/1078-0432.CCR-20-3201 relationship between the serum creatinine concentration and
2020 American Association for Cancer Research. GFR in different patient populations (7, 8).
AACRJournals.org | 1381
Williams et al.
1382 Clin Cancer Res; 27(5) March 1, 2021 CLINICAL CANCER RESEARCH
CamGFR v2 Estimates GFR in Patients with Cancer
Figure 1.
Schematic representation of data
acquisition, filtering, and study work-
flow. The partitioning into develop-
ment and validation datasets was
performed randomly. n, number
of GFR measurements (samples);
p, number of patients; nIDMS, number
of samples for which serum creati-
nine was measured with an IDMS-
standardized method; nEnzymatic,
number of samples for which serum
creatinine was measured with an
enzymatic method. The 7,695 sam-
ples in the “combined data filtered”
box include 4,983 from Cambridge,
2,056 from Manchester, and 656
from Gothenburg. The filter of
repeat measurements was the final
DuBois-DuBois equation; ref. 22) and divided by 1.73. In the case of the (AUC5) was calculated for all eGFR values generated using the Calvert
MDRD study equation, the IDMS- (23) and non-IDMS–adjusted (24) equation (1): dose (mg) ¼ AUC (mg/mL/minute) [GFR (ml/
versions were used for the respective data subsets. The median residual minute) þ 25]. For all models, we determined the proportion of
(mGFR eGFR) was used to assess bias, and the residual interquartile patients who would have received a dose with a percentage error
range (IQR) was used to assess precision. Accuracy is a combination of greater than 20% (P20) relative to what the dose would be if the
these two metrics and was estimated using the root-mean-squared measured GFR, rather than eGFR, was used for the calculation.
error (RMSE). Finally, we examined the clinical robustness of the A 95% confidence interval (CI) was calculated for each performance
estimation. A carboplatin dose for an AUC of 5 mg/mL/minute statistic using a bootstrap resampling procedure. Specifically, 2,000
Table 1. Patient characteristics by center and creatinine assay The core patient demographics, by center and creatinine assay
type. methodology, are presented in Table 1. Further demographic infor-
mation, including patients’ ethnicities and cancer diagnoses, is pre-
IDMS Non-IDMS sented in Supplementary Tables S1–S3. Supplementary Table S1
Median IQR Median IQR P value stratifies IDMS creatinine measurements further according to whether
Cambridge (n ¼ 4,983; p ¼ 4,557) they were performed using Jaffe or enzymatic methods. Unless oth-
Age (years) 58.9 21.5 60.6 20.1 0.00199 erwise stated, results from these two methods are grouped together for
BSA (m2) 1.84 0.331 1.84 0.325 0.287 all analyses of IDMS creatinine values. The numbers of patients with
Creatinine (mg/dL) 0.781 0.283 0.905 0.305 <0.0001 repeat GFR measurements are presented in Supplementary Table S4.
GFR (mL/minute) 82 37 82 40 0.332
Height (cm) 168 14.5 168 15 0.63 Comparisons of IDMS- and non-IDMS–standardized creatinine
Weight (kg) 74.5 23.2 73.6 22.9 0.0661 values
Manchester (n ¼ 2,056; p ¼ 2,056)
We first compared IDMS- and non-IDMS–standardized creatinine
Age (years) 63.1 17.1 65.2 18.5 0.0483
BSA (m2) 1.78 0.331 1.76 0.33 0.196
values within centers. This comparison was possible because Cam-
Creatinine (mg/dL) 0.758 0.311 0.939 0.249 <0.0001 bridge and Manchester changed from non-IDMS–standardized to
GFR (mL/minute) 79 39.5 74 38 0.0362 IDMS-standardized creatinine measurements during the sampling
1384 Clin Cancer Res; 27(5) March 1, 2021 CLINICAL CANCER RESEARCH
CamGFR v2 Estimates GFR in Patients with Cancer
significant differences in demographic parameters (Supplementary This survey of the possible models for GFR included other metho-
Table S5). We first refitted CamGFR v1 to the development dataset dologic approaches, namely stepwise regression or segmental regres-
including the additional variable of “creatinine type” (1 or 0 depending sion. However, performance metrics of these models were comparable
on the assay used), along with its interaction with the cubic log with those of CamGFR v2, and none of the other models had
(creatinine) terms. These features provided consistency with the significantly superior accuracy to the CamGFR v2 model across both
original model, while permitting independent adjustment of the IDMS and non-IDMS creatinine as assessed using permutation tests
coefficients for the cubic log(creatinine) terms in the IDMS- (ref. 26; Supplementary Materials and Methods; Supplementary Figs.
standardized data subset. The functional form of this model, hereby S2–S4; Supplementary Tables S7 and S8). To maintain consistency
termed “CamGFR v2,” is Equation A, and its non-IDMS–related with previous external validation work (10, 21), the CamGFR v2 model
coefficients did not differ significantly from their original counterparts was selected for further analysis.
in CamGFR v1 (ref. 21; Supplementary Table S6). As with the original CamGFR v1 (21), CamGFR v2 trained on the
pffiffiffiffiffiffiffiffiffi development set satisfied key assumptions of a linear model (Supple-
GFR ¼ b0 þ b1 Age þ b2 BSA þ b3 SexM þ b4 ScrIDMS
mentary Fig. S5), and thus permitted the generation of accurate
þb5 logðScrÞScrIDMS þb6 logðScrÞ2 ScrIDMS þb7 logðScrÞ3 ScrIDMS prediction intervals for eGFR.
þb8 logðScrÞScrnon-IDMS þ b9 logðScrÞ2 Scrnon-IDMS
þb10 logðScrÞ3 Scrnon-IDMS þ b11 AgeBSA þ b12 AgeSexM þ Performance of CamGFR v2 in the non-IDMS and IDMS
creatinine validation datasets
ðAÞ
We next used the non-IDMS and IDMS validation datasets to
b ¼ coefficients fitted by least-squares regression (see Supplementary compare the performance of the CamGFR v2 model with CamGFR
Table S5); Age ¼ age (years); BSA ¼ BSA (m2); SexM ¼ 1 for male, 0 for v1 and previously published creatinine-based models, namely the
female; Scr ¼ serum creatinine (mg/dL); ScrIDMS ¼ 1 if IDMS, 0 if non- CKD-EPI (29), Lund–Malm€ o revised (30), and Full Age Spectrum
IDMS; and Scrnon-IDMS ¼ 0 if IDMS, 1 if non-IDMS. The error term “” (FAS; ref. 31) models, as well as the MDRD study equation (23, 24).
is an independent, mean zero normally distributed random variable All of these models, with the exception of CamGFR v1, were
with a constant variance. developed using creatinine data calibrated to IDMS-standardized
measurement methodologies. In the case of MDRD study equation,
In addition to deriving new coefficients for CamGFR v2, alternate the non-IDMS–adjusted version (24) was used for non-IDMS
models with different predictors to CamGFR v2 were also explored. input data.
1386 Clin Cancer Res; 27(5) March 1, 2021 CLINICAL CANCER RESEARCH
CamGFR v2 Estimates GFR in Patients with Cancer
Figure 4.
Residual (measured GFR eGFR)
median, dose P20, and RMSE for the
CamGFR v1, CamGFR v2, CKD-EPI,
Lund–Malmo € (LM), and FAS models,
as well as the MDRD study equation, as
stratified by eGFR. Source data is pre-
sented in Supplementary Table S14.
for chemotherapy dosing is that its greatest gains in accuracy, relative C rather than creatinine into the CKD-EPI model leads to superior
to other models, are within the range of eGFRs commonly observed prediction of GFR, as well as of cardiovascular risk, end-stage renal
in patients with cancer (>60 mL/minute; Fig. 4; Supplementary disease, and all-cause mortality (39, 40). However, its usage is restricted
Table S14). It is also the only model other than CamGFR v1 that by cost, the lack of widespread standardized methodologies, and
generates accurate prediction intervals for each GFR estimate (Sup- altered physiology in the settings of thyroid dysfunction (41), inflam-
plementary Figs. S5 and S11). Applied to carboplatin, CamGFR v2 may mation (42), and cancer (43–48). Tumor size- and treatment
therefore represent a new standard of care. response–dependent fluctuations of cystatin C have been documented
A challenge inherent to modeling non-IDMS creatinine values is the in diverse cancer types (43–48) and represent significant obstacles to
significant variation in calibration between centers (Fig. 2; ref. 17). the use of cystatin C–informed eGFR in the management of patients
However, as adoption of IDMS standardization becomes more wide- with cancer and, in particular, the dosing of chemotherapy.
spread (6), this limitation will become less relevant to clinical practice. In summary, CamGFR v2 can incorporate either IDMS-
Currently, incorporation of either IDMS- or non-IDMS–standardized standardized or non-IDMS–standardized creatinine assays and is the
creatinine measurements best ensures robustness of CamGFR v2 to most accurate model for estimating GFR in patients with cancer. It is
technical variation between centers and internationally (Fig. 2; uniquely designed to assess the likelihood of clinically significant
refs. 5, 6, 17). Additional technical limitations to modeling the under- or overdosing, is most immediately relevant in the context of
relationship between creatinine and GFR come from the imprecision carboplatin, and should be examined and prospectively validated by
1388 Clin Cancer Res; 27(5) March 1, 2021 CLINICAL CANCER RESEARCH
CamGFR v2 Estimates GFR in Patients with Cancer
References
1. Calvert AH, Newell DR, Gumbrell LA, O’Reilly S, Burnell M, Boxall FE, et al. 22. Du Bois D, Du Bois EF. Clinical calorimetry: a formula to estimate the
Carboplatin dosage: prospective evaluation of a simple formula based on renal approximate surface area if height and weight be known. Arch Intern Med
function. J Clin Oncol 1989;7:1748–56. 1916;XVII:863–71.
2. Chantler C, Garnett ES, Parsons V, Veall N. Glomerular filtration rate mea- 23. Levey AS, Coresh J, Greene T, Marsh J, Stevens LA, Kusek JW, et al. Expressing
surement in man by the single injection methods using 51Cr-EDTA. Clin Sci the Modification of Diet in Renal Disease study equation for estimating
1969;37:169–80. glomerular filtration rate with standardized serum creatinine values.
3. Stevens LA, Coresh J, Greene T, Levey AS. Assessing kidney function - Clin Chem 2007;53:766–72.
measured and estimated glomerular filtration rate. N Engl J Med 2006;354: 24. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate
2473–83. method to estimate glomerular filtration rate from serum creatinine: a new
4. Coresh J, Astor BC, McQuillan G, Kusek J, Greene T, Van Lente F, et al. prediction equation. Ann Intern Med 1999;130:461–70.
Calibration and random variation of the serum creatinine assay as critical 25. Davison AC, Hinkley DV. Bootstrap methods and their application. Cambridge,
elements of using equations to estimate glomerular filtration rate. Am J Kidney UK: Cambridge University Press; 1997.
Dis 2002;39:920–9. 26. van der Voet H. Comparing the predictive accuracy of models using a simple
5. Myers GL, Miller WG, Coresh J, Fleming J, Greenberg N, Greene T, et al. randomization test. Chemom Intell Lab Syst 1994;25:313–23.
Recommendations for improving serum creatinine measurement: a report from 27. Chancharoenthana W, Wattanatorn S, Vadcharavivad S, Eiam-Ong S, Lee-
the Laboratory Working Group of the National Kidney Disease Education lahavanichkul A. Agreement and precision analyses of various estimated
Program. Clin Chem 2006;52:5–18. glomerular filtration rate formulae in cancer patients. Sci Rep 2019;9:19356.
6. National Institute of Diabetes and Digestive and Kidney Diseases. Creatinine 28. Garner AE, Barnfield MC, Waller ML, Hall GD, Bosomworth MP. Comparing
45. Bodnar L, Wcislo GB, Smoter M, Gasowska-Bodnar A, Stec R, Synowiec A, et al. 47. Kos J, Stabuc B, Cimerman N, Br€ unner N. Serum cystatin C, a new marker of
Cystatin C as a parameter of glomerular filtration rate in patients with ovarian glomerular filtration rate, is increased during malignant progression. Clin Chem
cancer. Kidney Blood Press Res 2010;33:360–7. 1998;44:2556–7.
46. Kume M, Yasui H, Yoshikawa Y, Horinouchi M, Higashiguchi K, Kobayashi Y, 48. Aydin F, Tezcan G, Gungor O, Cengiz AK, Hazar V, Akman S, et al. Can serum
et al. Transient elevation of serum cystatin C concentrations during perioperative cystatin C reflect the glomerular filtration rate accurately in pediatric patients
cisplatin-based chemotherapy in esophageal cancer patients. Cancer Chemother under chemotherapeutic treatment? A comparative study with Tc-99m DTPA
Pharmacol 2012;69:1537–44. two-plasma sample method. Nucl Med Commun 2010;31:301–6.
1390 Clin Cancer Res; 27(5) March 1, 2021 CLINICAL CANCER RESEARCH