Moderator's View

Low-protein Diet in Chronic Kidney Disease

Effectiveness, Efficacy and Precision Nutritional Treatments in Nephrology

Carmine Zoccali; Francesca Mallamaci


Nephrol Dial Transplant. 2018;33(3):387-391. 

In This Article

Effectiveness Versus Efficacy of Low-protein Diets in CKD

To make a balanced judgement of the arguments of the two contenders, we should consider separately intention-to-treat and per-protocol analyses of clinical trials performed so far. Intention-to-treat analysis is the primary, solidly established, analytical approach of clinical trials data. This analysis, respecting in full the randomization independent of actual adherence to treatments, gives a global estimate of the effectiveness of interventions being tested. Per-protocol analysis is a secondary analysis restricted to patients who complete the treatment where they were originally allocated. Even though prone to bias, per-protocol analyses constitute a useful exercise providing an estimate of the efficacy of treatment when adherence can be maintained. In this regard, uncritical reliance on the intention-to-treat principle may not be directly relevant for guiding decisions in clinical situations with different adherence patterns,[12] which is exactly the case with low-protein diet. If detailed clinical information is gathered during trials about adherence, modern statistical techniques allow proper censoring at the time when patients stop adherence to therapy for clinical reasons. Adjustment for confounding attributable to incomplete adherence, i.e. for pre-randomization and post-randomization prognostic factors that predict adherence to treatment, can be done by inverse probability weighting or by other generalized methods.[12]

Of course, such an approach cannot guarantee that adjustment for these prognostic factors will eliminate all bias from the traditional per-protocol analyses. Indeed, per-protocol analyses of randomized trials are inherently observational analyses, i.e. analyses where the possibility of potential confounding by unmeasured factors remains. Such a limitation notwithstanding, these analyses provide precious, patient-centered information that may be useful in clinical practice. As discussed, intention-to-treat analyses of low-protein diet in the MDRD study were substantially inconclusive. However, standard per-protocol analyses of the MDRD trial supported a beneficial effect of low-protein diet. Indeed, each 0.2 g/kg/day lower achieved protein intake associated with a 1.15 mL/min/year slower GFR decline (~30% of the mean GFR decline) and a 49% reduction in the risk of the composite outcome kidney failure or death.[13]

The modern approach to per-protocol analyses was unavailable 20 years ago and therefore it would be of utmost interest to see whether application of the new statistical techniques confirm traditional per-protocol analyses of the same trial. But the fact remains that standard per-protocol analyses of the MDRD trial and of other trials[14,15] support the contention that low-protein diet affords nephroprotection in selected patients. Thus the real issue at stake here is not whether properly applied low-protein diet can slow renal disease progression in selected cases, but rather the actual proportion of patients in whom long-term compliance can be safely achieved and the resource investment needed to maximize compliance and safety.

In the MDRD study, clinical care was very accurate, with monthly visits by physicians and dietitians. Nonetheless, the degree of compliance to the low-protein regimen was adequate only in 29% of patients.[16] By the same token, in the multicentre Northern Italian Cooperative Study[17] dietary compliance was similar or less than in the MDRD study. It is likely that in the resource-limited real-life clinical setting compliance is less than in the ideal setting up of a clinical trial. Furthermore, an issue remarked on by both contenders, the risk of malnutrition cannot be overlooked. In a resource-intensive study like MDRD, the low-protein diet group developed warning signals of malnutrition, including lower energy intake, body weight and percent body fat, biceps and triceps skinfold thickness and creatinine excretion.[18] From this it is clear that the application of a low-protein diet in CKD requires a more solid and larger evidentiary basis for being recommended for the majority of CKD patients. Per-protocol analyses in clinical trials coherently suggest potentially significant renoprotection, but the issue of compliance, safety and intensity of clinical surveillance demands new research efforts building on the knowledge developed so far.