Renal and Systemic Effects of Calorie Restriction in Patients With Type 2 Diabetes With Abdominal Obesity: A Randomized Controlled Trial

Obesity, especially if centrally located (1), and diabetes (2) are both associated with renal dysfunction sustained by glomerular hyperfiltration (3,4), a risk factor for accelerated loss of renal function and onset and progression of nephropathy (5). Thus, glomerular hyperfiltration might be one of the possible pathogenic links between obesity and chronic kidney disease (CKD) (6,7). Finding that bariatric surgery ameliorates glomerular hyperfiltration associated with severe obesity (8) suggests that weight loss, in addition to ameliorating a series of cardiovascular risk factors, might also affect the onset and progression of CKD (5,8). This invasive procedure is, however, necessarily restricted to a selected population at very high risk of obesity-related complications. Thus, calorie restriction (CR) remains the principal method for inducing weight loss (9). However, no trial so far has formally tested the role of CR and weight loss on glomerular filtration, in particular by directly measuring the glomerular filtration rate (GFR) in subjects with glomerular hyperfiltration and abdominal obesity (10).

Thus, we evaluated whether and to what extent measured GFR (11) could be affected by CR in the context of a controlled, randomized clinical trial “Caloric REstriction in Subjects with abdominal Obesity and Type-2 diabetes at increased risk” (C.RE.S.O).

This academic, single-center, parallel-group, prospective, randomized, open-label, blinded end point (PROBE) trial was conducted at the Clinical Research Center (CRC) for Rare Diseases of the IRCCS - Istituto di Ricerche Farmacologiche Mario Negri. Participants were identified among patients referred to the outpatient clinics of the CRC and the diabetology units of Bergamo, Treviglio-Caravaggio, Romano di Lombardia, and Seriate Hospitals, all in Italy. Participants were individuals with type 2 diabetes (American Diabetes Association criteria) aged >18 years old with abdominal obesity defined as waist circumference of >94 cm in men and >80 cm in women (12), serum creatinine <1.2 mg/dL, and urinary albumin excretion (UAE) <20 μg/min in overnight urine collections. They had a stable body weight and calorie intake, and a stable diet with a standardized content in micro- and macronutrients and salt, according to guidelines (13), and no systematic changes in blood pressure (BP), glucose, and lipid-lowering medications during the previous 6 months. We excluded patients with primary, immune-mediated, or ischemic kidney disease; urinary tract obstruction or infection; concomitant therapy with renin-angiotensin system (RAS) inhibitors, steroids, or nonsteroidal anti-inflammatory agents; heart failure; uncontrolled diabetes; hypo- or hypernatremia from any cause; previous bariatric surgery; depression; alcohol and drug abuse; pregnancy; ineffective contraception; perimenopausal age; cancer or chronic disease that might jeopardize study completion; primary endocrinological diseases; poor compliance; or were unable to provide informed consent. The study conformed to the principles of the EU Clinical Trials Directive (2001/20/EC), Good Clinical Practice, and the Declaration of Helsinki. The ethics committee of the local health agency in Bergamo, Italy, approved the study. All patients provided written informed consent. Data were recorded in dedicated case record forms and then entered into the database at the CRC. The study was reported according to Consolidated Standards of Reporting Trials guidelines.

Abdominal circumference was measured at the end of a normal expiration at the level of the iliac crest. Body weight was measured in duplicate in the morning after a 12-h fast with the subject wearing a hospital gown and no shoes. The BMI was calculated using the standard formula. Office BP was measured with an oscillometric device (HEM-705CP; Omron, Tokyo, Japan) with the patient sitting after 15 min of rest. The average of three measurements, 2 minutes apart, was recorded. Blood for laboratory assessments was sampled the morning after overnight fasting. UAE was measured in three consecutive overnight urine collections, and the median was recorded.

GFR was measured by the plasma clearance of unlabeled iohexol (11) after a single, intravenous injection of 5 mL iohexol solution (647 mg/mL Omnipaque 300; Nycomed Amersham Sorin, Milan, Italy). Participants with a GFR >120 mL/min (i.e., a GFR exceeding the upper limit of normal range for measured GFR) were defined as hyperfiltering, and those with a GFR ≤120 mL/min as nonhyperfiltering (5,14). The GFR was not normalized for the body surface area to avoid the confounding effect of changes in body surface area associated with diet-induced changes in body weight (15,16), and absolute GFR values were considered for the analyses. On the following day, the total-body glucose disposal rate (GDR) was assessed with hyperinsulinemic-euglycemic clamp (17).

Participants were randomly assigned (1:1) to 25% CR or to continue on their already prescribed SD by a computer-generated list of random permuted blocks prepared by a statistician (Giovanni Antonio Giuliano) of the CRC, who was not involved in the analyses. All data assessors were masked to treatment allocation.

Intervention in the SD aimed to reinforce compliance with the recommended diet. Patients in the CR arm were provided with personalized dietary guidelines to decrease their daily calorie intake by 25%. The nutrient composition recommended with both CR and SD interventions was flexible to accommodate individual preferences but was designed to provide 45–50% of energy from carbohydrates, 30–35% from fat, and 15–20% from proteins, supply 100% of the daily recommended micronutrient intake, >20 g/day of fiber, and <300 mg/day of cholesterol. Patients were encouraged to consume moderate and low glycemic index and nutrient-dense foods (18). No particular lifestyle modification was introduced. Dietary prescriptions were based on energy intake at baseline, estimated by the subjects’ resting metabolic rate (RMR) using the Mifflin predictive equation (19), and results from the Physical Activity Recall (PAR) and Total Daily Energy Expenditure (TDEE) questionnaire (20). CR corresponded to a 25% calorie decrease from estimated total daily energy intake.

Patients allocated to the CR intervention were given a prescription of total calories to consume daily and dietary plans based on exchange systems that deliver a fixed amount of calories per food portion. Weight loss goals were set together with the patients, and to facilitate adherence, patient-dietitian contact (in person, by telephone, or e-mail) was provided throughout the study period once a week during the first 3 months and once every 2 to 3 weeks during the remaining 3 months. In case the patient-dietitian contact did not prove enough for maintaining dietary compliance, behavioral intervention strategies, such as stimulus control (avoiding triggers that prompt eating), social support (assistance from family members and friends in modifying lifestyle behaviors), cognitive restructuring (thinking in a positive manner), problem-solving skills (systematic method of analyzing problems and identifying possible solutions), and relapse prevention (methods to help recovery from episodes of overeating or weight regain), were provided.

Patients were instructed to keep daily records of their weight and weekly fasting glucose measurements. One week before each trimonthly follow-up visit, participants completed a 7-day food diary using household measures. Diaries were analyzed by means of the dietary analysis software package MètaDieta, Version 1.0.2, 2009 (METEDA S.r.l., San Benedetto del Tronto, AP, Italy) and used to assess compliance in the allocated study group. The dietary software uses official national food composition databases such as the INRAN (Istituto Nazionale di Ricerca per gli Alimenti) and the IEO (Istituto Europeo di Oncologia).

Clinical and laboratory parameters, including serum urea levels taken as an indirect marker of dietary protein intake, evaluated at baseline were reevaluated at 3 and 6 months after randomization, with the exception of GFR and GDR, which were reevaluated at 6 months only (final visit). At each visit, adverse events were recorded, and physical and laboratory parameters were assessed for safety.

Blood and urine samples were collected after subjects had fasted overnight and were centrally analyzed at the CRC for Rare Diseases. Routine laboratory parameters were measured by spectrophotometry (UniCel Synchron Clinical System DXC800; Beckman Coulter s.r.l., Milan, Italy). Glycated hemoglobin (HbA_1c) values were expressed by using mmol/mol units according to the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) and were then converted to percentage values according to the National Glycohemoglobin Standardization Program (NGSP) by using the online HbA_1c converter at http://www.ngsp.org/convert1.asp.

Serum insulin and angiotensin II concentrations were measured by chemifluorescence (Access 2; Beckman Coulter, Inc.) and the enzyme immunoassay kit (Angiotensin II SPIE-IA; Bertin Pharma, Montigny le Bretonneux, France), respectively, and hs-CRP, apolipoprotein A, apolipoprotein B, and urinary albumin by rate nephelometry (Immage; Beckman Coulter, Inc.).

The primary end point was the change in GFR at the 6-month follow-up versus baseline. Other outcomes included changes in GDR (coprimary outcome), BP, heart rate (HR), blood glucose, HbA_1c, serum lipid, plasma renin activity, C-reactive protein (CRP), and safety variables, including vital signs, clinical laboratory tests, and adverse events.

Sample size was estimated for the main prespecified outcome variable assuming a two-group t test (two-sided) of the difference between CR and SD. On the basis of GFR data available at the database of the CRC at the time of study planning, we assumed a baseline mean ± SD GFR of 111 ± 19.0 mL/min. We predicted a 15% GFR reduction from 111 to 94.35 mL/min with CR and no change with SD. On the basis of these assumptions, a sample size of 29 evaluable participants per group would give the trial 90% power to detect as statistically significant (α = 0.05 two-tailed test) the expected difference in GFR change between the two treatment groups. To account for a 20% dropout rate, we planned to include 36 participants per group.

All statistical analyses were conducted by modified intention to treat, using SAS 9.1 and Stata 12 software. Changes in GFR and all other between-group effects were assessed by ANCOVA, adjusted for baseline measures. Within-group comparisons were assessed by paired t tests, repeated-measures ANOVA, or the McNemar test. Correlations were tested with the Pearson r correlation coefficient. Multiple regression models were used to investigate the association between baseline independent covariates and GFR changes. We considered age, sex, and those baseline covariates that, in simple regression models, were associated with GFR change at an α = 0.10 level of significance (two-tailed). In the case of correlated covariates, variable selection was guided by clinical criteria. To test the relationships between changes in different considered parameters and the concomitant 6-month changes in GFR, we first identified which one among the considered anthropometric, clinical, and metabolic variables and serum lipids had a strongest correlation with the outcome. We then entered changes in these variables along with changes in mean BP (taken as a surrogate of both systolic and diastolic BP) into a multivariable model considering GFR changes at 6 months as the outcome variable. Data are expressed as mean ± SD, median (interquartile range), or number (%), unless otherwise specified. All P values were two-sided.

Of the 149 screened patients, 75 did not fulfill the selection criteria or declined to participate. From September 2009 to May 2012, 36 of the 74 included patients were randomized to CR and 38 to SD. Two participants withdrew from the CR arm at treatment months 1 and 5 because of noncompliance. One participant on the SD was excluded at month 3 because of a protocol violation (initiation of RAS inhibition therapy during hospitalization because of atrial fibrillation), and another subject withdrew consent at month 3 for personal reasons. Thus, 34 participants on CR and 36 on SD completed the study and were available for final analyses (Fig. 1).

All 74 included patients were Caucasian, 56 (75.7%) were male, and 11 (14.9%) were current smokers. Age averaged 59.8 ± 7.1 years. At baseline, 34 participants (45.9%) were overweight (BMI 25–29.9 kg/m²) and 33 (44.6%) were obese (BMI ≥30 kg/m²), with a mean BMI of 29.8 ± 3.8 kg/m², and a waist circumference of 103.1 ± 10.3 cm. The GFR averaged 107.9 ± 20.0 mL/min, and 20 (27.0%) patients were hyperfiltering. BP, blood glucose, and serum lipids were relatively well controlled. Results for other laboratory parameters were unremarkable. Sociodemographic (Supplementary Table 1) and anthropometric, clinical, and laboratory parameters (Table 1); calorie intake, energy consumption, and diet composition (Table 2); and distribution of concomitant medications (Table 3) at inclusion were similar between the groups, with the exception of some excess of patients on statins in the SD group. Ten patients per group were hyperfiltering (Table 1). Independently of treatment allocation, the GFR at baseline correlated with body weight, BMI, serum angiotensin II levels, the LDL-to-HDL ratio, and UAE (Supplementary Table 2).

The GFR significantly decreased by 7.6 ± 11.7% (P = 0.0006) with CR, whereas the 2.7 ± 11.1% reduction observed with SD was not significant (P = 0.172). GFR changes versus baseline were significantly different between groups (P = 0.0472) (Table 1 and Fig. 2, top panel). Within the hyperfiltering group, the GFR significantly decreased by 11.7 ± 9.9% (P = 0.005) with CR, whereas the 5.3 ± 9.0% reduction observed with SD failed to achieve statistical significance (P = 0.095). In the nonhyperfiltering group, the GFR decreased by 3.8 ± 8.1% (P = 0.032) with CR but did not change appreciably with SD (Table 1 and Fig. 2, middle and bottom panel). GFR reduction tended to be larger in patients with a BMI ≥30 kg/m² than in those with a smaller BMI. No significant change was observed with SD in both BMI groups (Table 1).

UAE decreased significantly from 5.1 ± 2.7 to 4.4 ± 2.4 μg/min (P = 0.0248) in the CR group but did not change appreciably in the SD group (Table 1 and Supplementary Fig. 1, bottom panel).

Body weight decreased by 4.7 ± 5.5 kg (5.2 ± 5.8%) in the CR group (P < 0.0001) and by only 0.6 ± 1.6 kg (0.7 ± 1.9%, P = 0.031) in the SD group (Table 1 and Supplementary Fig. 2, top panel). These changes were significantly different between groups (P = 0.0001). BMI consistently decreased by 1.6 ± 1.9 kg/m² (5.2 ± 5.8%, P < 0.0001) and waist circumference by 5.9 ± 4.7 cm (5.8 ± 4.5%, P < 0.0001) with CR and by only 0.2 ± 0.6 kg/m² (0.7 ± 1.9%, P = 0.034) and 1.6 ± 3.5 cm (1.5 ± 3.4%, P = 0.009), respectively, with SD. Changes were significantly different between groups (P < 0.0001 for both parameters) (Table 1 and Supplementary Fig. 2, middle and bottom panel).

Systolic (P = 0.0003), diastolic (P < 0.0001), and mean (P < 0.0001) BP consistently decreased with CR and only marginally decreased with SD. Interestingly, HR also decreased significantly with CR (P = 0.0003) but did not change appreciably with SD. Changes between groups were significantly (P < 0.05) different for all considered parameters (Table 1 and Supplementary Fig. 3).

Blood glucose (P = 0.0001) and HbA_1c (P = 0.0001) levels both significantly decreased with CR, and on SD the opposite trend was observed, which was significant for HbA_1c (P = 0.039). Changes in both parameters were significantly different (P = 0.0004 and P < 0.0001, respectively) between the groups (Table 1 and Fig. 3, top and middle panel). These changes were associated with a significant increase in GDR with CR (P = 0.0007). GDR was stable with SD, and GDR changes were significantly different between the two treatment groups (P = 0.0075) (Table 1 and Fig. 3, bottom panel). In patients without long-acting insulin therapy, insulin levels were similar between treatment groups and did not change appreciably during the observation period.

Serum HDL levels increased (P = 0.043) and LDL levels decreased (P = 0.027) with CR. The opposite was observed in SD. Thus the LDL-to-HDL ratio significantly decreased with CR (P < 0.01) and tended to increase in SD. Changes in this parameter were significantly different between the groups (P = 0.023) (Table 1). Other considered parameters did not change appreciably within and between the groups (Table 1).

Changes in BP, metabolic control, and serum lipids were not explained by changes in concomitant treatment because the distribution of different BP and lipid-lowering medications in the two groups did not change appreciably during the study and because the proportion of patients on oral hypoglycemic agents similarly increased in both groups (Table 3).

Aspartate transaminase (P = 0.0001) and alanine aminotransferase (P = 0.0001) levels decreased with CR and tended to increase with SD. Changes between the groups were significantly different (P = 0.0067 and P = 0.0089, respectively (Table 1). Creatine phosphokinase did not change appreciably within and between the groups. hs-CRP levels significantly decreased (P = 0.0075) on CR and did not change appreciably with SD, whereas serum angiotensin II levels tended to decrease with CR and to increase with SD. Changes in both variables were significantly different between groups (P = 0.0164 and P = 0.0421, respectively) (Table 1 and Supplementary Fig. 1, top and middle panel).

According to the 7-day food diaries, mean energy intake decreased by 14.95 ± 17.8% (P < 0.0001) with CR and by 5.35 ± 15.7% (P = 0.049) with SD. These changes were significantly different between the groups (P = 0.0061), whereas RMR, MET, and TDEE did not change appreciably within and between the groups (Table 2). The reduction in calorie intake achieved in the CR, compared with the SD group, was largely explained by a reduced intake of carbohydrates and alcohol, whereas the total intake of proteins was similar between the groups as documented by data obtained by dietary diaries evaluation, including data on phosphate intake (Table 2), and by serum urea values that were very similar between the treatment groups and did not change appreciably throughout the entire study period (Table 1). The dietary intake of monounsaturated fatty acids, saturated fats, animal proteins, and fat decreased, whereas the intake of total fiber, polyunsaturated fatty acids, and vegetable fat increased with CR compared with SD (Table 2). Subjects in the CR group introduced significantly more iron, magnesium, phosphorus, potassium, vitamin C, riboflavin, folate, and β-carotene than those in the SD group, whereas the intake of other dietary micronutrients was similar between the groups. In particular, sodium intake was very much the same at inclusion and decreased similarly in the two groups during the study (Table 2).

GFR reduction significantly correlated with body weight (P = 0.048), BMI (P = 0.017), and the GFR (P = 0.008) at inclusion. At multiple regression analyses, considering the variables listed in Table 1, which at simple regression analyses were associated with the outcome at a significance level of P < 0.10, GFR reduction was predicted by CR (P = 0.045) and baseline GFR (P = 0.004).

GFR reduction also significantly correlated with a reduction in daily calorie intake, body weight, BMI, waist circumference, systolic, diastolic, and mean BP, blood glucose, serum triglyceride levels, and an increase in GDR (Table 4). The correlation between changes in GFR and body weight was significant in the study group as a whole (r = 0.409, P = 0.0007) and in patients with CR (r = 0.438, P = 0.0095) considered separately but not in those with SD (r = 0.271, P = 0.133). Multivariable regression analyses showed the reduction in mean BP was the strongest predictor of GFR reduction. The association of weight reduction with GFR reduction was borderline significant, whereas changes in blood glucose and serum triglycerides had no predictive value (similar findings were observed when diastolic BP was entered into the model instead of mean BP) (Table 4). Independently of treatment allocation, 1 mmHg of mean BP reduction and 1 kg of weight loss were associated with a mean GFR reduction of 0.45 and 0.60 mL/min, respectively.

Only two serious adverse events occurred, both in the SD group. Overall, nonserious adverse events were generally mild and transient in nature and were similarly distributed between the groups. Viral and respiratory tract infections were slightly more frequently reported in the SD group, whereas musculoskeletal events tended to be more frequent with CR (Table 5). No event, however, was considered to be treatment related by the investigators.

In this PROBE clinical trial in patients with type 2 diabetes and abdominal obesity, the 6-month CR significantly decreased GFR compared with SD, an effect that was largely driven by GFR reduction in patients with a higher GFR to start with and which was associated with a reduction in waist circumference, body weight, BMI, systolic and diastolic BP, blood glucose, serum LDL-to-HDL cholesterol levels, and amelioration of insulin sensitivity, as assessed by hyperinsulinemic-euglycemic clamps in all patients. Of interest, every 1 kg of weight loss was associated with ∼0.6 mL/min GFR reduction. CR and SD were both tolerated well, and no adverse effects possibly related to inadequate or unbalanced nutrient supply were observed throughout the study. Treatment effect was unlikely explained by changes in factors independent of CR that can affect glomerular hemodynamics, such as protein and sodium intake, which was very similar between the treatment groups. Moreover, baseline patient characteristics and distribution of concomitant medications at inclusion and during the study were also similar between groups. Thus, study results appear to reflect a genuine effect of CR on glomerular filtration.

These findings could have clinical implications, because persistent hyperfiltration predicts a faster GFR decline and an excess risk of progression to micro- or macroalbuminuria in patients with type 1 (2) or type 2 diabetes (5,21), whereas amelioration of hyperfiltration is associated with a slower GFR decline in the long-term and nephroprotection (5). We previously found that in a large cohort of patients quite similar to the C.RE.S.O. cohort, a larger GFR reduction at 6 months strongly and independently predicted a slower GFR decline in the long-term (5). In particular, a 7.6% short-term GFR reduction similar to that achieved by CR predicted a mean long-term GFR decline of 0.08 (SEM 013) mL/min/1.73m² per month, whereas a 2.7% reduction similar to that observed in patients on SD predicted a long-term decline of 0.36 (SEM 0.07) mL/min/1.73m² per month. If the above findings are generalized to our C.RE.S.O. cohort, we can speculate that CR might reduce the rate of long-term GFR decline by approximately four- to fivefold compared with SD. This renoprotective effect might translate into a rate of renal function loss similar to that observed in healthy adults with aging (22). Interestingly, the benefit of CR on glomerular dysfunction was more consistent and clinically relevant in those patients with the highest GFR at baseline. Thus, the renoprotective effect of CR is expected to be larger right in those patients who, because of hyperfiltration, are at increased risk of accelerated renal function loss (5).

Finding that a large part of the effect of CR on kidney function appeared to be explained by the reduction in BP and, to a lesser extent, by weight reduction is consistent with the hypothesis that an early rise in GFR in obesity is largely mediated by sodium retention (23). Increased renal sodium reabsorption, which appears to be mediated by activation of the renin-angiotensin-aldosterone system (RAAS) and sympathetic system and by altered intrarenal physical forces, may eventually result in volume expansion and increased BP (24). Moreover, increased proximal tubular reabsorption may reduce sodium chloride delivery to the macula densa and cause, via deactivation of tubuloglomerular feedback, reductions in afferent arteriolar resistance and increases in glomerular perfusion and filtration (8,23,25). Thus, we speculate that CR might reduce the GFR by reducing the sodium pool and therefore reducing BP and kidney perfusion. This effect could be mediated by decreased RAAS and sympathetic activity, as suggested by the reduction in angiotensin II levels and heart rate we observed with CR compared with SD. Enhanced responsiveness to natriuretic peptides, which may even precede CR-induced weight loss, might also play a role (26). Moreover, by reducing tubular sodium reabsorption, CR might enhance sodium chloride delivery to the macula densa, restore preglomerular resistances, and therefore, limit glomerular hyperperfusion and consequent hyperfiltration. These findings, however, must be interpreted with caution because of the post hoc and observational nature of the analyses and because mechanisms mediating the effects of CR on kidney function should be investigated in prospective pathophysiology studies.

CR was also associated with other clinically relevant functional and metabolic effects:

1. Amelioration of metabolic, BP, and lipid control: These changes were associated with a significant increase in GDR with CR compared with SD, an effect that indicated amelioration of insulin sensitivity. This functional effect most likely explained blood glucose and HbA_1c reduction in the active treatment arm and probably could have at least partly contributed to BP reduction and amelioration of dyslipidemia in this subgroup. However, this effect could not be explained by changes in energy consumption and concomitant medications, which were similar between the groups. Independent of involved mechanisms, amelioration of the functional and metabolic parameters mentioned above can be seen in the context of an overall amelioration of metabolic syndrome and are expected to translate into a clinically relevant reduction in long-term cardiovascular risk. Interestingly, the increase in GDR was also independently associated with GFR reduction, a finding that is consistent with the hypothesis that insulin resistance may also have a role in the pathogenesis of glomerular hyperfiltration (27).

2. Reduction in sympathetic tone and RAAS activity: The significant decrease in HR and serum angiotensin II levels observed with CR compared with SD might have clinical relevance. Indeed, a high resting HR has long been independently associated with an increased risk of all-cause mortality and cardiovascular complications in type 2 diabetes (28), as well as in the general population (29), and more recently, with new onset and worsening of retinopathy and nephropathy (30), findings that are most likely explained by the increase in BP and sympathetic activity associated with overweight and obesity (31). Consistently, long-term CR reduced the HR and improved its variability in overweight but otherwise healthy adults, an effect associated with reduced sympathetic activity and a concomitant increase in parasympathetic nervous system tone (32). RAAS activation is another cardiovascular risk factor that has also been involved in the pathogenesis of glomerular hyperfiltration and progression of renal disease in experimental and human diabetes (33,34). Actually, the initial state of hyperfiltration associated with excessive adiposity, especially if centrally located, is largely sustained by raised systemic and renal production of angiotensin II (35), which may promote systemic and local chronic inflammation, reactive oxygen species formation, lipogenesis, and hypertension (35,36), with progressive renal dysfunction and structural damage (37). We consistently found that 6 months of CR significantly reduced CRP, a systemic marker of inflammation and an independent cardiovascular risk factor (38). The small reduction in albuminuria we observed with CR might also have clinical relevance because albuminuria has been identified as an independent and continuous risk factor for renal and cardiovascular disease, even in the normal albuminuric range (39).

3. Decrease in liver aminotransferase levels: This effect was most likely explained not only by reduced alcohol intake but also by weight loss and improved insulin resistance achieved by CR. Elevated liver enzymes are a risk of progressive nonvirus-related nonalcoholic fatty liver disease in type 2 diabetes and associate strongly with increased HbA_1c, insulin resistance, and obesity. Reduced liver aminotransferase levels through CR could preserve liver function and reduce steatohepatitis because primary prevention requires weight loss, improved glucose control, and metabolic syndrome amelioration (40).

The number of participants was estimated a priori on the basis of the expected treatment effect, which made it possible to adequately power analyses despite the relatively small sample size. Moreover this was a pilot, exploratory, and technically challenging study, and renal function, insulin sensitivity, and albuminuria were measured by gold standard techniques, which, by reducing the risk of random data fluctuations, increased the statistical power of the study analyses. Protein intake was not monitored by measuring urinary urea excretion because 24-h urine collections were not available. However, finding that serum urea levels were stable over time and comparable between treatment groups at all visits, along with data from dietary diaries evaluation, confirmed that protein intake was comparable between the groups and stable over time, and reasonably, could not explain the GFR changes observed with calorie restriction. This conclusion is corroborated by finding that a reduction in dietary proteins of at least 50–60% is needed to obtain an appreciable change in GFR and that protein intake in both C.RE.S.O. groups was more than double that reported in previous studies of patients with diabetes with low-protein diets (41,42).

Despite the highly labor-intensive design, the study had a high retention rate of enrolled participants and good adherence to the study interventions, as shown by the successful weight and waist circumference reduction achieved by CR. These findings confirm that compliance to dietary recommendations is an achievable goal, provided that dietitians and doctors are strongly motivated and are devoted enough to also transmit their motivations to more disinclined patients. A “trial effect” most likely explained why some weight loss was also observed in controls on the SD during the treatment period, an effect that most likely diminished between-group differences in at least some of the considered outcome variables. Finding that the treatment effect was captured despite this limiting factor provided additional evidence of the robustness of the results. However, whether these results can also be generalized to obese patients without diabetes must be investigated.

One major strength was the PROBE design, which allowed blinded analyses of outcome variables despite the open design, and at the same time, minimized costs and closely reflected standard clinical practice, which should make the results more easily applicable in routine medical care (43). Changes in considered variables over the follow-up period were consistent and uniformly confirmed the potential beneficial effect of CR on a series of renal and cardiovascular risk factors. However, whether this short-term effect will/can translate into long-term nephro- and cardioprotection in this clinical context needs to be addressed in longer and appropriately powered trials.

The CR-induced negative energy balance results in substantial improvements of several major risk factors for the initiation and progression of CKD in patients with diabetes with abdominal obesity and no evidence of renal involvement. In particular, CR and weight loss, along with amelioration of insulin resistance and other functional and metabolic abnormalities, achieved a significant short-term reduction in the GFR that conceivably reflected amelioration of glomerular hyperfiltration and that resembled the reduction observed after an invasive procedure such as bariatric surgery (8). Long-term randomized clinical trials are needed to assess whether CR may achieve clinically relevant protection against progressive renal function loss and development of nephropathy in the long-term as well as reduce overall patient cardiovascular risk.

Acknowledgments. The authors are indebted to Flavio Gaspari, who supervised the laboratory analyses; Jorge Arturo Reyes Loaeza, Claudia Patricia Ferrer Siles, Karen Courville, Patricia Espindola, Silvia Prandini, Veruscka Lecchi, and Svitlana Yakymchuk, who took care of the study participants; Giovanni Antonio Giuliano, who generated the list of random permuted blocks; Nadia Rubis and Giulia Gherardi, who, respectively, supervised the monitoring of the study and the activities of the day hospital of the CRC; Paola Boccardo, who took care of the ethical and regulatory aspects of the trial; Norberto Perico and the staff of the CRC, who contributed to the conduction of the study (all from IRCCS - Istituto di Ricerche Farmacologiche Mario Negri, Centro di Ricerche Cliniche per le Malattie Rare “Aldo e Cele Daccò”, Bergamo, Italy); and Antonio Bossi (Azienda Socio Sanitaria Territoriale [ASST] Ospedali di Treviglio-Caravaggio and Romano di Lombardia), Ruggero Mangili (ASST Ospedale Bolognini di Seriate), Roberto Trevisan (ASST Ospedale Papa Giovanni XXIII of Bergamo), and the staff of their outpatient clinics for their major contribution to patient screening and selection.

Funding. This research was supported by grants from the Istituto Superiore di Sanita/National Institutes of Health Collaborative Projects of the Italian Ministry of Health, the Bakewell Foundation, and the Longer Life Foundation (an RGA/Washington University Partnership).

The sponsors had no role in conducting study and reporting the results.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. P.R. wrote the final version of the manuscript. P.R., M.A., S.R., G.R., and L.F. wrote the study protocol. P.R., M.A., G.R., and L.F. contributed to data analyses and interpretation. P.R., G.R., and L.F. had the original idea. M.A., B.R., S.R., M.T., C.A., A.Pa., and I.P.I. identified, treated, and monitored study participants and contributed to data recording. M.A. and G.P. prescribed CR or SD and monitored compliance to the recommended diets. M.A. and L.F. wrote the first draft. A.Pe. and A.R. performed the statistical analyses. O.D. monitored the study. D.M. prepared the database and helped in data handling. A.C., F.C., S.F., and N.S. performed the GFR measurements and laboratory tests. All of the authors had direct access to original data, critically revised the draft, and approved the final manuscript. G.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.