The Potential Effect of Using the Cockcroft-Gault Method on Tenofovir-Associated Renal Impairment Reports and on Clinical Decisions Regarding Tenofovir Use in Individual Patients: Implications for the Future

Introduction: In Namibia, the Cockcroft-Gault (C-G) method is recommended for monitoring renal function in HIV patients receiving Tenofovir Disoproxil Fumarate (TDF)-containing Combination Antiretroviral Therapy (cART). However, there are concerns with the potential over-reporting of TDF-associated renal impairment. Methods: Retrospective study comparing the renal function of patients receiving 2nd line cART with either C-G or Chronic Kidney Disease-Epidemiology (CKD-EPI) methods. Results: 71 patients were included. The majority (62%) received TDF-containing 1st line ART. All received 2ndline cART containing TDF/Lamivudine (3TC)/Zidovudine (AZT) and LPV/r. Before switching to 2nd-line cART, 40.8% and 8.5% had abnormal eGFR according to C-G and CKD-EPI methods respectively. During 2nd-line cART, 47.9% and 7% of patients had abnormal eGFR by C-G and CKD-EPI methods, respectively, and 4.1% and 2.8% respectively experienced a decline in eGFR. There was a significant lack of agreement between the two methods. Conclusion: The C-G method has the potential to report more cases of TDF-associated renal impairment. Consequently, national guidelines in Namibia and other pertinent countries should be reviewed if this is the recommended method for monitoring renal function.


Background
Antiretroviral (ARV) medicines typically undergo fast-track approval due to the recognised need to improve the quality-of-life and prolong the lives of people infected with the Human Immuno-De ciency Virus (HIV) [1,2]. Sub-Saharan Africa, being the most a ected region of the world with the HIV pandemic, has gained tangible bene ts from the use of combination antiretroviral therapy (cART) [3][4][5]. Despite these tangible bene ts, ARV medicines are associated with adverse drug reactions, the impact, incidence, and prevalence of which is not completely measured at the time these medicines are approved for clinical use especially if patients have greater co-morbidities than those enrolled into Phase III trials [6]. Nevertheless, the establishment of pharmacovigilance centres like the erapeutics Information and Pharmacovigilance Centre (TIPC) in Namibia exposes the safety concerns pertaining to these ARV medicines. is is important in Africa given the appreciable genetic di erences between patients in Africa and the western countries where many studies are currently undertaken [7,8].
Earlier in 2016, we published a paper on the e ect of changing from rst-to second-line cART on renal function [9] studying the e ects at a leading health facility in Namibia, from which all patients who were receiving second-line cART (n=71) were included in this study. We used the Cockcro -Gault (C-G) formula to estimate CrCl [10], because at the time of the study the C-G method was the recommended method in Namibia for assessing renal function in all patients receiving Tenofovir Disoproxil Fumarate (TDF)-containing cART in Namibia [9,11]. TDF is a pro-drug for Tenofovir (TFV) the active ingredient [9], which is known to cause renal impairment by, but not limited to, interfering with the proximal tubular function [12,13]. As such, in Namibia, any TDF dosage adjustments or withdrawals were generally prompted by the estimated CrCl via the C-G formula.
In regards to the monitoring of renal function, the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) method, which approximates to the Gold-Standard, is known to be better than the C-G method [14]. However as mentioned, the C-G method is the one currently recommended for monitoring renal function in Namibia  [9,11]. However, we do not know the extent to which the C-G method could a ect the number of reports of TDF-associated renal impairment. Furthermore, we do not know the e ect of the result of renal function assessed by the C-G method can have on subsequent clinical decisions such as a reduction in the dose of TDF or withdrawal of TDF.
Awareness of the superiority of the CKD-EPI method over the C-G method [14] prompted us to use the patients in our earlier study to rstly assess the relationship between the C-G and the CKD-EPI eGFR results. Secondly, ascertain the extent of any di erences between the C-G and CKD-EPI methods in the proportion of patients considered to have experienced renal impairment. Lastly, we set out to discover if there was a di erence between the C-G and CKD-EPI eGFR results in the number and/or proportion of potential clinical decisions on TDF s dose or TDF s withdrawal.
As a result, guide the authorities in Namibia on whether to continue with recommending the C-G method or change to the CKD-EPI method. is may also have implications for other African resource limited countries if the C-G formula is also being recommended for monitoring renal function in patients who are infected with the HIV and are receiving a cART regimen containing TDF.
Objectives e rst objective of this study was to assess if there was agreement between the eGFR results computed by the C-G and CKD-EPI methods in our study population. e second objective was to identify eGFR results by either the C-G and CKD-EPI methods that met the de nition of decline in renal function as per the criteria used in this study, which was also previously used by Kalemeera et al. [9]. ese would qualify to be reported as TDF-associated renal impairment to the pharmacovigilance centre. e reduction or withdrawal of TDF.

Methods
is was retrospective study assessing the potential e ects that the C-G method may have on ADR reports and clinical decisions pertaining to TDF, and comparing these with the CKD-EPI method.
is was based on patients receiving second-line cART in out-patients at the Katutura Intermediate Hospital [9]. In the previous study, the C-G method was used to calculate CrCl, which in this study was converted to eGFR using the patients body surface area ( e C-G equation that was used is documented in Box 1).

Dichotomous Grouping of renal function
Any eGFR below 90 ml/min/1.73 m was categorised as abnormal. eGFR was calculated using the C-G and CKD-EPI formulae. For C-G the formula was: [(140-Age) X Weight]/SeCr (X0.85 for females), while for CKD-EPI the general formula was:C X [(SeCr/0.7) e ] X [(0.993) Age ], where C varied according to ethnicity, and e varied according to the SeCr measurements and sex.

Further categorisation of renal function
The eGFR was further categorised into stages I to IV, where stage I was ≥ 90; stage II was 60-<90; stage III was 30-<60; and stage IV was <30, ml/min/1.73 m 2 .

Declined renal function
Renal function was considered to have declined if the last eGFR during secondline cART was ≥ 25% less than the eGFR computed during first-line cART. In addition, when the SeCr was noted to have increased by ≥ 25%, during secondline cART, renal function was considered to have declined.

Improved renal function
Renal function was considered to have improved if the last eGFR during second-line cART was ≥ 25% more than the eGFR during first-line cART. When the SeCr was noted to have decreased by ≥ 25%, during second-line cART, the renal function was considered to have improved, even though the change in eGFR was <25% Box 1: De nitions of renal function.
We also computed another set of eGFR values using the CKD-EPI method ( e CKD-EPI equations that were used are documented in Box 1). e computations of eGFR were based on the last recorded SeCr measurements before patients were switched to second-line cART and the last SeCr measurements during second-line cART. We used the criteria documented in Table 1 to place the patients in the requisite category of renal function. We used the criteria documented in Box 1 to determine if the patient had experienced a decline or improvement in renal function.
To assess the relationship between eGFR values computed by C-G and CKD-EPI methods, we used Cohen s Kappa; Pearson correlation; linear-regression coe cients; and Bland-Altman plots. We used Odds Ratios, Pearson Chi-Square test, Fisher s Exact Test, and Binomial Distribution to assess the di erence between the eGFR values computed by C-G and CKD-EPI methods on the potential numbers of reports of TDF-associated renal impairment adverse reactions. Lastly, to assess the potential e ect of the eGFR computed by the C-G method on TDF s dose or withdrawal in comparison with the eGFR computed by the CKD-EPI method, we used Odds Ratio. We set the con dence level at 95%, and statistical signi cance at a p-value <0.05.

Patient characteristics
A total of 71 patients were included in this study. Of these 57.7% (n=41) were female. Table 2 contains details of the mean weight and age before and during second-line cART. 62% (n=44) of the patients received TDF/lamivudine (3TC) and EFV or NVP, and one patient received TDF/FTC/EFV. All patients were receiving a second-line regimen that consisted of TDF/Lamivudine (3TC)/Zidovudine (AZT) and LPV/r. Upon evidence of treatment failure, patients were switched to second-line cART. At the time of data collection, these patients had spent an average of 5.2 ± 2.7 years on rst-line cART, and an average of 1.8 ± 0.5 years on second-line cART ( Table 2).

Renal function before second-line therapy
Before switching to second-line ART, the C-G and CKD-EPI methods showed that 40.8% (n=29) and 8.5% of the patients respectively had abnormal renal function (Table 3). During second line cART, the C-G method indicated that more patients (n=10, 14.1%) experienced a decline in renal function, than shown by the CKD-EPI method (n=2, 2.8%). Similarly, the number of patients who experienced an increase in eGFR according to the C-G method (n=6) was slightly lower than those who experienced the same according to the CKD-EPI method (n=8). Moreover, the eGFR calculations during second-line cART yielded similar results to those we observed to have occurred during rst-line cART (Table 3).  Table 3: Number (Percentage) of patients in di erent stages of renal function as categorised by using the C-G-or the CKD-EPI formulae, before second-line cART.

Assessment of agreement between C-G and CKD-EPI ndings
Based on Cohen s kappa, the results show a lack of agreement between C-G and CKD-EPI eGFR computations during rst-line cART (kappa=0.224, p=0.003) and during second-line cART (kappa=0.110, p=0.042) ( Table 3). Pearson correlation analysis showed that the eGFR for the two methods had a strong relationship (r=0.803; p=0.000) during rst-line cART, and there were similar ndings during second-line cART (r=0.749; p=0.000), (Figure 1 e correlation scatter plots ( Figure 1) show that the eGFR measured by C-G and CKD-EPI correlated strongly. However, the Bland-Altman plots based on the mean bias (standard deviation) during rst-line cART: 17.4 (11.4), and during second-line cART: 21.9 (15.6), exposed the absence of agreement between the two tests ( Figure  2: Panel C and Panel D). Linear regression analysis con rmed the lack of agreement between the two methods: beta-coe cient of -0.191 (p=0.001), and -0.336 (p=0.001).

Potential numbers of TDF-associated renal impairment reports
We computed the di erence in eGFR between the last measurement during rst-line cART and the last measurement during second-line cART, for both C-G and CKD-EPI. We reviewed the magnitude of the di erence in eGFR and determined which result met the criteria for decline in renal function. Our ndings were that the chance of identifying TDF-associated ADRs with C-G was ve times as much as using the CKD-EPI methodology. is nding was signi cant with the Odds Ratio (Cl 95%; OR=5 (1.06 23.6); p-value=0.04), and also with other tests (Table 4).

Potential numbers of TDF dose reductions or withdrawals
We assessed each patient s record of eGFR computed by CG and CKD-EPI to determine if the computed eGFR had the potential to recommend a dosage reduction in TDF or its withdrawal. We found that the chance of reducing TDF s dose or withdrawing TDF when C-G is used was three times that of the CKD-EPI methodology (Cl 95%; OR=3.0 (0.12 74.9); p-value=0.05). Since the CKD-EPI results had no patient with an eGFR requiring TDF s dosage reduction, the p-value was only computable for the Odds Ratio (Table 4). the gold-standard. ey found that the CKD-EPI and MDRD methods out-performed the C-G method. e C-G method was associated with lower eGFR computations. e authors also found that the CKD-EPI eGFR closely approximated the gold-standard [15,16]. In their study, Matsushita et al. found that CKD-EPI, which out-performs MDRD, led to the reclassi cation of many patients from lower to higher renal function categories [17]. Since our study was retrospective, and the gold standard is generally not used in clinical practice in Namibia, and because CKD-EPI out-performs MDRD, which out-performs the C-G method [17,18], we compared the C-G based estimations of GFR with those computed by the CKD-EPI method. Since the superiority of CKD-EPI over C-G is an established fact, we did not implement this study to a rm or disprove the CKD-EPI s superiority. However, we set out to highlight the potential impact the lesser method, is the C-G method, may have on the number of reports of TDF-associated renal impairment in Namibia, and its potential to a ect clinical decisions such as TDF s dose reduction or its withdrawal in patients purportedly experiencing renal impairment.

Discussion
As expected, the C-G method yielded lesser eGFR values than the CKD-EPI method. According to Pearson Correlation analysis, the eGFR results computed by C-G and CKD-EPI had a strong positive correlation. However, according to the Bland-Altman plots ( Figure 2) and linear regression analyses, the eGFR results of C-G and CKD-EPI were not close to agreeing. Since the CKD-EPI based method is re ective of the true renal function of the patient, the related eGFR results showed that only a few patients experienced a decline in renal function, contrary to the C-G based estimations, which suggested more patients were experiencing renal impairment. Our ndings are similar to Botev et al. study in which the C-G method wrongfully categorised 60% of the study patients into lower renal function categories [19].
Our ndings show that the C-G method has implications on TDF s safety because this method yielded a greater proportion of patients with lower stages of renal function before and a er switching (Table 3). Furthermore, the C-G method detected more events of decline in renal function (14.0%, n=10) than the CKD-EPI method (2.8%, n=2). Interestingly, our ndings indicate that the C-G method has not had implications on TDF s use to any extent di erent from the CKD-EPI method. is is based on the fact that the majority of patients who were declared to have experienced a decline in renal function via the C-G method had an eGFR that was above 50 ml/min, below which dosage adjustments are required [20]. Furthermore, if we assume that eGFR had stabilised at the estimated levels, the C-G based method ceases to pose any challenges of wrongfully categorising patients into lower renal function stages. However, the latter could not be con rmed. In regards to TDF s withdrawal or dosage-interval adjustments due to low eGFR computations, our results show no signi cant di erence between CKD-EPI and C-G (OR=3; p=0.05). However, since the C-G method underestimates eGFR, a signi cant proportion of patients are likely to miss-out on its use. (NB: e Namibia cART guidelines recommend that before initiating TDFbased rst-line cART, the CrCl should be calculated. ey also state that initiation of TDF-containing cART should be avoided in cART naïve patients with CrCls below 60 ml/min). Despite the absence of signi cance in regards to switching or dose reduction in the study patients, there remains a concern that the use of the C-G method may place patients experiencing actual declines of their renal function in categories of lower renal function status, as observed by Botev et al. [19]. Pharmacovigilance centres need to be aware of this fact especially in countries where the C-G method is still used for assessing renal function.
rough the passive reports received by the erapeutics Information and Pharmacovigilance Centre (TIPC) in Namibia, some work by TIPC has been published [21][22][23]. Amongst these adverse reaction reports, are reports of renal impairment associated with TDF [21], which were based on reduced CrCl computed by the C-G method. Since the C-G method is likely to raise false alarms of renal impairment, it is likely that some of the reports of TDF-associated renal impairment were not re ective of the actual renal function of the patient. ese ndings may also be true for other African countries. Consequently, the use of C-G method to assess renal function in patients receiving TDF-containing cART is likely to wrongfully increase the number of reports of TDF-associated renal impairment. Subsequently, the health system s concern for cardiovascular disease in these patients may be escalated. Additionally, the low eGFR computed by the C-G method can in uence clinical decisions such as prolongation of TDF s dosage intervals. As a result, subjecting the patients to sub-optimal concentrations of TDF (and 3TC) thereby creating an environment for the emergence of resistance against cART regimens. Lastly, the low eGFR computed by the C-G method can result in TDF s withdrawal.
e addition of the Urine-Dipstick to CrCl as strategy to detect TDF-associated renal impairment is a recent development in cART guidelines [24,25]. e use of the urine dipstick in this regard is based on the fact that the normal nephron would reabsorb all the ltered glucose except in hyperglycaemic patients, who at the detection of glycosuria have plasma glucose levels above the normal threshold for glucose reabsorption: 180 mg/dl [26]. Consequently, the cART guidelines rest on the understanding of the pathophysiology of TDFassociated proximal tubulopathy, during which the reabsorption of glucose is likely interrupted resulting in glycosuria in euglycaemic patients [13,27].
e addition of the urine dipstick is promising to partly or fully negate the possible mis-categorisation of renal impairment by the C-G method in patients receiving TDF-containing cART.
is is because the low eGFR in the presence of proximal tubulopathy likely co-occurs with glycosuria and albuminuria. However, we have not found any literature to date that discusses the co-variance of eGFR and urine-glucose or urine-proteins in patients receiving TDF-containing cART. In addition, the sensitivity of the urine dipstick in this patient category needs to be evaluated.
is should be the subject of future research. In any event, there are concerns with the C-G method in Namibia and other countries [28,29].

Limitations
e major limitation to our study is that there was no gold standard to compare the C-G-and CKD-EPI-eGFR computations with. It may be that C-G-based assessments identi ed actual renal impairment cases. We are also aware that we only assessed 71 patients. is is because we built on a previous study conducted in Namibia to address current concerns with the C-G method. However, in view of the ndings, we believe our study results are valid, providing guidance to key stakeholder groups in Namibia and throughout Africa. We will be following this up in future studies.

Conclusion
e C-G method is likely to increase the number of reports of renal impairment compared with the CKD-EPI method.
e increased incidence in renal impairment reports is of concern to the pharmacovigilance centre, whose critical role is to safe-guard the population against drug related adverse reactions. e disproportionate increase in TDF-associated renal impairment can instigate policy changes in its use. Our study has also shown that the wrong eGFR computations by the C-G method may not in uence clinical decisions at the individual patient level, at least not within 2.6 years of second-line cART. It is apparent that within 2.6 years of second line cART, the C-G method did not lead to any clinical decision pertaining to TDF s use such as prolongation of TDF s dosage interval in patients with normal renal function and the withdrawal of TDF from the cART regimen from such patients. e repercussions of such decisions would include the emergence of resistance, new infections with a resistant virus, the replacement of TDF with zidovudine-a less safe ARV medicine than TDF, and lastly increased expenditure on second-line cART. However, these were not observed. e results of the Urine Dipstick may prove to be bene cial in terms of con rmatory renal function tests. However, the evidence of this is currently lacking. Since TDF-associated nephropathy can be progressive, it is advisable that better methods of renal function assessment than the C-G method, such as the CKD-EPI method, should be considered alongside the use of urine-dipsticks, as this may prevent administration of low doses of TDF instigated by falsely low GFR estimates computed by the C-G method. ese are implications for the future in Namibia.

Key Message
e strong positive correlation between eGFR values by C-G and CKD-EPI methods does not necessarily mean that the clinical interpretations of renal function from both tests are the same. It is known that the C-G based method is associated with lower eGFR values than the Gold-Standard, unlike the CKD-EPI method.
In regards to pharmacovigilance, the C-G method of assessment of renal function in patients receiving TDF-containing cART is likely to increase the number of reports of TDF-associated renal impairment, which is not a true re ection of actual renal function. e addition of the urine dipstick may or may not be bene cial in identi cation of clinical tubulopathy, and thus needs to be evaluated.
When the C-G method is used to assess renal function, the result of reduced eGFR may lead to dose adjustment of TDF, subjecting the patient to sub-therapeutic levels and consequently to the emergence of resistance against TDF.
ere is need to utilise better methods to assess renal function in patients receiving TDF-containing cART, such as CKD-EPI as this method more closely re ects the Gold-Standard results.