Can a simple risk score predict chronic kidney disease among people living with HIV?

Development and validation of a risk score for chronic kidney disease in HIV infection using prospective cohort data from the D:A:D study.

Mocroft A, Lundgren JD, Ross M, Law M, Reiss P, Kirk O, Smith C, Wentworth D, Neuhaus J, Fux CA, Moranne O, Morlat P, Johnson MA, Ryom L, D:A:D study group, the Royal Free Hospital Clinic Cohort, and the INSIGHT, SMART, and ESPRIT study groups. PLoS Med. 2015 Mar 31;12(3):e1001809. doi: 10.1371/journal.pmed.1001809. eCollection 2015.

Background: Chronic kidney disease (CKD) is a major health issue for HIV-positive individuals, associated with increased morbidity and mortality. Development and implementation of a risk score model for CKD would allow comparison of the risks and benefits of adding potentially nephrotoxic antiretrovirals to a treatment regimen and would identify those at greatest risk of CKD. The aims of this study were to develop a simple, externally validated, and widely applicable long-term risk score model for CKD in HIV-positive individuals that can guide decision making in clinical practice.

Methods and findings: A total of 17 954 HIV-positive individuals from the Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) study with ≥3 estimated glomerular filtration rate (eGFR) values after 1 January 2004 were included. Baseline was defined as the first eGFR >60 ml/min/1.73 m2 after 1 January 2004; individuals with exposure to tenofovir, atazanavir, atazanavir/ritonavir, lopinavir/ritonavir, other boosted protease inhibitors before baseline were excluded. CKD was defined as confirmed (>3 mo apart) eGFR ≤60 ml/min/1.73 m2. Poisson regression was used to develop a risk score, externally validated on two independent cohorts. In the D:A:D study, 641 individuals developed CKD during 103 185 person-years of follow-up (PYFU; incidence 6.2/1000 PYFU, 95% CI 5.7-6.7; median follow-up 6.1 y, range 0.3-9.1 y). Older age, intravenous drug use, hepatitis C coinfection, lower baseline eGFR, female gender, lower CD4 count nadir, hypertension, diabetes, and cardiovascular disease (CVD) predicted CKD. The adjusted incidence rate ratios of these nine categorical variables were scaled and summed to create the risk score. The median risk score at baseline was -2 (interquartile range -4 to 2). There was a 1:393 chance of developing CKD in the next 5 y in the low risk group (risk score <0, 33 events), rising to 1:47 and 1:6 in the medium (risk score 0-4, 103 events) and high risk groups (risk score ≥5, 505 events), respectively. Number needed to harm (NNTH) at 5 y when starting unboosted atazanavir or lopinavir/ritonavir among those with a low risk score was 1702 (95% CI 1166-3367); NNTH was 202 (95% CI 159-278) and 21 (95% CI 19-23), respectively, for those with a medium and high risk score. NNTH was 739 (95% CI 506-1462), 88 (95% CI 69-121), and 9 (95% CI 8-10) for those with a low, medium, and high risk score, respectively, starting tenofovir, atazanavir/ritonavir, or another boosted protease inhibitor. The Royal Free Hospital Clinic Cohort included 2548 individuals, of whom 94 individuals developed CKD (3.7%) during 18 376 PYFU (median follow-up 7.4 y, range 0.3-12.7 y). Of 2013 individuals included from the SMART/ESPRIT control arms, 32 individuals developed CKD (1.6%) during 8452 PYFU (median follow-up 4.1 y, range 0.6-8.1 y). External validation showed that the risk score predicted well in these cohorts. Limitations of this study included limited data on race and no information on proteinuria.

Conclusions: Both traditional and HIV-related risk factors were predictive of CKD. These factors were used to develop a risk score for CKD in HIV infection, externally validated, that has direct clinical relevance for patients and clinicians to weigh the benefits of certain antiretrovirals against the risk of CKD and to identify those at greatest risk of CKD.

Abstract [1]  Full-text [free] access [2]

Editor’s notes: The nephrotoxicity of antiretroviral drugs, particularly tenofovir, is of concern, particularly where there is limited access to laboratory monitoring of kidney function. The development of kidney impairment among people with HIV is associated with poor outcomes, and in low resource settings where dialysis is not available this can be catastrophic.

This study, like previous work, attempts to address this problem by developing a risk score for the development of chronic kidney disease (CKD). The strength of this study is the availability of data for over 17 000 men and women living with HIV enrolled in cohort studies for many years, and in over 40 countries globally. The resulting risk score uses nine simple clinical variables which predict CKD both overall, and after starting potentially nephrotoxic antiretrovirals. A short risk score, not including cardiovascular risk factors, which may be more suitable for low resource settings, shows almost as good a prediction of CKD.

So will this risk score become widely used in clinical decision making? For high income countries this tool may be useful to identify people where strategies to prevent cardiovascular and renal disease are best focussed. It may also be useful to identify people at high risk of developing CKD for whom use of tenofovir may be unacceptable, especially when monitoring of kidney function is limited. However, few of the enrolled people were from low and middle income countries, and there was limited information on the race of participants. Therefore, the risk score may need to be validated in low resource settings before it can be widely used. Whether the use of the tool would help to improve clinical outcomes where kidney function is frequently monitored is unclear.

Meanwhile, a new drug formulation, tenofovir alafenamide (TAF), is currently in clinical trials. This appears to be associated with less renal toxicity, and to be safe and well tolerated among adults with decreased kidney function. If future trial results support this evidence, and tenofovir alafenamide becomes widely available, concern about drug nephrotoxicity may become a less pressing clinical issue.

Africa [10], Asia [11], Europe [12], Latin America [13], Northern America [14], Oceania [15]
Argentina [16], Australia [17], Austria [18], Belarus [19], Belgium [20], Bosnia and Herzegovina [21], Brazil [22], Bulgaria [23], Canada [24], Croatia [25], Czech Republic [26], Denmark [27], Estonia [28], Finland [29], France [30], Germany [31], Greece [32], Hungary [33], Ireland [34], Israel [35], Italy [36], Japan [37], Latvia [38], Lithuania [39], Luxembourg [40], Morocco [41], Netherlands [42], Norway [43], Poland [44], Portugal [45], Russia [46], Serbia and Montenegro [47], Singapore [48], Slovakia [49], Slovenia [50], Spain [51], Sweden [52], Switzerland [53], Thailand [54], Ukraine [55], United Kingdom of Great Britain and Northern Ireland [56], United States of America [57]
  • [58]