Correlation of the Dietary Protein Intake between Those Estimated from a Short Protein Food-Recall Questionnaire and from 24-Hour Urinary Urea-Nitrogen Excretion in Stages 3-4 Chronic Kidney Disease Patients

Introduction High protein intake may accelerate progression of chronic kidney disease (CKD). Estimation of dietary protein intake (DPI) is indispensable for management of CKD, but to achieve optimum DPI is quite challenging in routine clinical practice. We recently studied a beneficial effect of utilizing integrated care on the management of CKD at the rural community level. In that study, we created a short protein food-recall questionnaire (S-PFRQ) as a working tool to estimate DPI of the CKD patients during home visit by community health personnel. Herein, we reported the initial evaluation of the reliability of S-PFRQ from our previous study. Objective We compared the amount of DPI obtained from S-PFRQ with that obtained from protein-equivalent of total nitrogen appearance (PNA). Methods In the previous ESCORT-2 study, 914 patients with CKD stage 3 or 4, who were living in the rural area of Thailand, were prospectively followed while receiving integrated care for 36 consecutive months. During home visits by community nurses from subdistrict health centers, dietary food recall was made, recorded in S-PFRQ, and DPI was obtained. Among these, sixty patients were randomly selected, and 24-h urine was collected for urinary urea-N and estimation of PNA. A correlation was made between DPI obtained from S-PFRQ and PNA. Results The DPIs derived from S-PFRQ and PNA were 28.8 ± 14.8 and 39.26 ± 17.79 g/day, respectively. The mean difference and 95% CI between the 2 methods was −10.43 (−7.1 to −13.8) g/day, respectively (P < 0.001). Interclass correlation between these 2 methods was 0.24, P = 0.007. The difference between the 2 methods remained constant across different amounts of DPI. Conclusion The DPI estimated from S-PFRQ significantly correlated to that from PNA. However, the S-PFRQ method yielded a DPI value which was about 10 g of protein or 25% less than the PNA method. Despite this amount of difference, this S-PFRQ is user-friendly and could be used during field work as an easy and simple tool for DPI estimation in resource-limiting condition.


Introduction
Chronic kidney disease (CKD) is a growing healthcare problem worldwide.It is estimated that about 800 million population on earth harbor CKD at any stage [1].CKD is associated with increased risks of cardiovascular diseases and premature mortality.Moreover, CKD leads to poor quality of life of individuals and their families [2].Expense and resource utilization incurred from CKD management is extremely high [3].In Tailand, about 17% of the total population could have CKD stages 3-5 (nondialysis) [4].Hemodialysis is the main renal replacement modality in Tailand and cost about 60 USD per session or about 7800 USD per patient per year [5].Kidney replacement therapy expenditure was approximately as high as 3% of total health expenditure of Tailand [6].At an early stage of the disease, CKD prevails not only at secondary-or tertiary-care hospitals but also in the rural community across the country.Dietary modifcation is recommended in the practice guideline for delaying progression of CKD [7].Patients with CKD should have moderate to stringent restrictions on protein and salt consumption [8].Previous studies reported that patients with CKD who had received consultation from dietitians had slower progression of CKD [9,10].However, accessibility to such dietary care is usually confned to developed countries or at major tertiary-care hospitals where qualifed dieticians are abundantly available.In a recent cross-sectional survey study from United States of America, the rate of referral for medical nutritional therapy was only 50.7% [11].In developing countries, counseling a dietician for patients with CKD is practically low [12].Tus, it is mandatory to identify a simple and pragmatic way to estimate dietary protein intake (DPI) among patients with CKD in resource-limited countries.
At present, several alternative methods have been utilized as a tool for estimating DPI, namely, 3-day dietary recall, dietary food record, 24-h urine collection, and short food recall questionnaire [13,14].Te frst 2 methods require much patient's understanding and compliance with dietitian's instruction.Accuracy of the methods also relies on consistency of dietary ingredients of the food recipes.Tus, it might be subject to estimation error when it is applied to oriental foods, which commonly do not have exact portion of the ingredients.Moreover, an experienced dietitian is required to calculate the amount of nutrient intake.Collection of 24-h urine, for measuring urine urea-nitrogen (urea-N) and estimation of protein-equivalent of total nitrogen appearance (PNA), seems to be a more accurate method and less dependent on specialized personnel [15].However, it is inconvenient for patients, subject to collection error, and not practical to use in a primary care setting.
In our previous studies among patients with CKD stages 3-4 who were residing in rural communities of Tailand, we could demonstrate that our model of an integrated care for CKD conducted by healthcare personnel available in the rural healthcare setting (i.e., consisting of visiting a multidisciplinary care team at the district hospital, in conjunction with receiving home visit by community nurses and village health volunteers from the subdistrict health centers) could be efective in delaying CKD progression [16,17].In the latter study [17], the community nurses from subdistrict health centers were assigned to use a short protein food recall questionnaire (S-PFRQ) as a simple tool for estimating an amount of protein consumed by each patient to provide appropriate nutritional counseling.In this communication, we intended to correlate the results of DPI obtained from S-PFRQ and PNA.

Methods
Te ESCORT-2 study protocol was approved by the Ethics Committee of Institutional Review Board, Ministry of Public Health, Tailand, and registered with https://www.clinicaltrials.in.th (TCTR-20160614001).Details of the study had been described elsewhere [17].Importantly, patients who had heavy proteinuria (more than 2+ of proteinuria on urine dipstick or 3.5 g per day) were not eligible for that study.In that communication, patients with stage 3 or 4 CKD were taken care of by a multidisciplinary care team at district hospitals.In addition, they also received home visits twice a year from a community care network team (CCNT).Te CCNT comprised a community nurse from subdistrict health center and a village health volunteer who was responsible for primary health care of the corresponding households.CCNT advised the patients on how to conduct healthy lifestyles with respect to CKD, interviewed about their daily activities, their consumption of salty or highprotein foods, and other unhealthy behavior.Of all patients who were enrolled in the ESCORT-2 study, patients with their research identifcation numbers ended with 0, 15, 30, 45, 60, etc., were randomly selected for this study.If any of these patients were not available, those who had a research identifcation number next in order were invited instead.Written informed consent was obtained prior to the enrollment.

Development of a Short Protein Food Recall Questionnaire (S-PFRQ).
In the beginning, an interview was conducted with nutritionists who were working at the local district hospitals.Eight food items containing high protein, which were commonly consumed by rural Tais, were identifed.Tese included egg white, whole egg, mackerel, meat, processed meat, seafood, meatball, and milk.For a patient's convenience, the unit of expression of each food item was a common unit which was used daily by and easily understandable among villagers, e.g., a number of eggs or fsh, a number of soup spoons, etc. Prior to initiation of the trial, members of CCNT were trained by a renal dietician on how to interview patients and how to record on S-PFRQ.A 7-day period was set as a time interval for food intake recall.Records on S-FPRQ were made every 6 months during home visit throughout the 3-year study period of the ESCORT-2 study.

Estimation of DPI Based on S-PFRQ.
Te total amount of food items consumed during a 7-day recall was converted from the local units as such to a standard protein portion, i.e., one portion of protein intake or about 30 g of meat contains about 7 grams of protein [18].Ten, this amount was averaged to a daily DPI.Te details of short protein food recall questionnaire (S-PFRQ) are shown in Figure 1.

Calculation of Dietary Protein Intake from the 24-h Urine
Collection.Troughout the ESCORT-2 study, the enrolled cases in this study were asked to collect 24-h urine, while staying at home, and every 6 months thereafter.Te urine specimens were sent to the Department of Laboratory Medicine, Kamphaeng Phet Hospital, Kamphaeng Phet Province, for analysis of creatinine, and urea-N.Creatinine was assayed with an enzymatic method.Calculation of

Results
Among 914 patients enrolled in the ESCORT-2 study, 60 cases were eligible for the current study (Figure 2).Altogether, concurrent S-PFRQ interview and PNA estimation could be achieved in 272 episodes over the 36-month followup period.Te mean age was 62.0 ± 7 years, with a male to female ratio of 1 : 2.5.Twenty-one, 28, and 11 cases were classifed as CKD stages 3A, 3B, and 4B, respectively.Diabetes, hypertension, and cardiovascular diseases were present in 52%, 97%, and 3% of cases, respectively.Te baseline eGFR was 40.3 ± 9.73 mL/min/1.73m 2 body surface area.Te BMI was 25.1 ± 2.83 kg/m 2 .Te baseline serum albumin was 4.27 ± 0.41 g/dL.Fifty-eight percent of participants had positive proteinuria detected by urine dipstick.Te 24-h urine volume was 1581 ± 612 ml (Table 1).
Overall, DPIs estimated from S-PFRQ and PNA were 28.8 ± 14.8 and 39.26 ± 17.79 g/day, respectively.Tese were equal to 0.46 ± 0.03 and 0.59 ± 0.09 g/kg/day of protein intake estimated from S-PFRQ and PNA, respectively.Te overall mean diferences of DPI estimation between these 2 methods (S-PFRQ-PNA) were −10.43 g protein/day (95% CI −7.1 to −13.8).Te mean diferences of DPI estimation from the 2 methods in patients with stage 3A, 3B, and 4 CKD were −18.35, −7.18, and −16.83 g protein/day, respectively (Table 2 and Figure 3).Te mean diference of DPI estimation was statistically lower in the stage 3B patient group than stage 3A and 4 patient groups.Te mean diference of DPI estimation in patients with diabetes was −12.26 g protein/day which was higher than patients without diabetes but not statistically signifcant.Also, patients with proteinuria tended to have higher diference of DPI estimation than patients without proteinuria (Table 3).
Te Pearson correlation coefcient of DPIs estimated from these 2 methods was 0.126 (P � 0.021) (Figure 4).Te mean diferences of DPI estimation from the 2 methods were −13.28 g protein/day at the baseline and subsequently were Figure 1: Short-protein food recall questionnaire English version.Te frst column is the list of 8 groups of common protein diet, the 2 nd column is the portion size, the 3 rd column is the conversion factor for each food group, and the last column is the estimated equivalent portion of protein intake.
−12.21, −9.93, and −8.98 in frst, second, and third years, respectively.Te intraclass correlation −0.12, −0.09, 0.12, and 0.51 at the baseline, at the frst, second, and third years, respectively (Figure 5).Overall, intraclass correlation (ICC) between these 2 methods of DPI estimation was 0.24 (95% CI � 0.041-0.392,P � 0.007).Te Bland-Altman plot was done to elucidate bias and limit of agreement of these 2 methods of DPI estimation as shown in Figure 6.About ninety-fve percent of the values lie within the limit of agreement.Te diference of estimated DPIs was relatively stable across a broad range of the mean DPIs of the 2 methods.

Discussion
CKD is now a signifcant health problem in both developed and developing countries.In the latter country group, limited access to kidney replacement therapy due to several factors signifes that delaying kidney disease progression is of prime importance.Among these disease-modifying factors, protein restriction is one of the key recommendations [7].Te proper amount of DPI recommended in clinical practice guidelines is quite consistent [7,19,20].Dietary assessment tools are important for improving patient adherence to nutritional advice.However, identifying a practical, sustainable, and user-friendly tool for accessing DPI has remained a problematic issue.In this study, we could demonstrate that frst, the DPI obtained from S-PFRQ correlated reasonably well with that from the PNA method, with a mean diference of only −10.43 g of protein per day, or DPI estimated from S-PFRQ was about 25% lower than DPI estimated from PNA (Figure 6).Second, ICC of DPI estimation from these 2 methods progressively increased over time.Lastly, up to 95% of the values lie within the limit of agreement indicating that this diference is consistent across a broad range of the amount of DPI (Figure 6).
Estimation of DPI with the S-PFRQ method could be lower than the PNA method due to several possible explanations including imperfect recall leading to underreporting of protein intake and consumption of protein from other foods not listed in the questionnaire.Food recall is a simple and convenient dietary intake assessment method.Yet, it relies on patient's memory which could be imperfect particularly in our study participants who are aging and undereducated population.A previous study conducted in non-CKD patients showed that daily protein intake estimation from 24-hour food recall was lower than urinary nitrogen excretion by 25.5% [21].A study done in CKD patients also showed that the food record method   International Journal of Nephrology underestimates protein intake compared to 24-hour urine collection [22,23].Our fndings are in accord with these publications.Besides patients' memory, interviewer skill and experience are also important.ICC between these 2 daily protein estimation methods is increased over the time of follow-up suggesting that the skill and experience of healthcare providers in interviewing food recall had been improving over time, resulting in more accurate results.Some participants might tend to underreport protein intake to show that they have good compliance with healthcare worker nutritional counseling.Studies done in dialysis and nondialysis CKD patients showed that underreporting energy intake was found in more than 50% of patients [24,25].Te eight food items enlisted in the S-PFRQ were only the most consumed among villages.Certainly, there could have been many other kinds of food, which might contain a considerable amount of protein, and might not have been included in the questionnaire.Tus, part of food protein might have been omitted from the calculation.Nevertheless, the diference of DPI values obtained from the two methods of estimation is only 10 g of protein per day.Moreover, this diference as such was stable across a wide range of dietary    Proteinuria is another factor that could afect estimation of protein intake by using the PNA method.In this study, we did not estimate proteinuria quantitatively.Tus, mean diferences of DPI estimation from 2 methods could be higher than our fndings.However, in this study, patients with heavy proteinuria were not eligible for participation [17].Tus, the efect of proteinuria on our results should be minimal.Furthermore, patients with proteinuria tended to have higher diferences of DPI but not reaching statistical signifcance (Table 3).Whether the presence of proteinuria could exert a clinically signifcant efect on DPI estimation needs further study.
Tere are several limitations in our study.Due to 24hour urine collection technical difculties, we therefore randomly selected only 60 patients for initial evaluation of our food recall questionnaire performance.Secondly, proteinuria was assessed with the urine dipstick test.Proteinuria was therefore not quantifed in the estimated daily protein intake from the PNA method.Lastly, our food recall questionnaire was designed for a 7-day period which might be too long period of time to recall.However, a 7-day period is intended to reduce selection bias and daily protein intake is supposed to be an average of 7-day period.
Dietary protein modifcation would be efective for delaying CKD progression if it is implemented at an early 6 International Journal of Nephrology stage of the disease when the patient is symptom-free.
Having this simple questionnaire tool in hands, it will be convenient for primary health personnel, such as community nurses at subdistrict health ofce or the other, to give advice or instruct the patients to avoid consuming high amounts of dietary protein.Our S-PFRQ could be an example of a simple, but moderately efcient dietary assessment tool that could be adapted to use in other resourcelimiting countries.It would create better healthcare professional-patient communication and promote adherence to low protein diet counseling [26,27].Nevertheless, S-PFRQ needs to be according to diferent patterns of foods of each country.In addition, a dietary intake survey needs to be done prior to enlisting high protein food items into the questionnaire.A commonly used unit and a picture of each food item should also be utilized in the questionnaire.
We hope that implementing this simple and friendly used tool in primary care settings would enable local health care personnel to be a meaningful driving force on delaying CKD progression at an early stage as expected.International Journal of Nephrology

List of common protein diet 1 .
Egg white (medium size) 2. Whole egg (medium size) 3. Mackarel or other fshes (Each about 6-8 inches in length) 4. Pork, Chicken, beef, or other meat 5. Processed meat (e.g.white pork sausage, Chinese sausage, sour pork) 6. Seafood ex.shrimps, crabmeat, mussels (3-4 cm in size) 7. Meatball (standard size) 914 patients in ESCORT-2 study 281 episodes of DPI by PNA were obtained 342 episodes of DPI estimated by SPFRQ were obtained 1 : 15 randomly selection 60 patients were randomly selected Had DPI estimated from both SPFRQ and PNA every 6 months for 36 months follow up period 272 episodes of matched DPI estimation by SPFRQ and PNA were analyzed

Figure 3 :
Figure 3: Box and Whisker plot shows comparison of mean DPI estimation from the 2 methods in diferent stages of CKD.

Figure 5 :Figure 4 :
Figure 5: Diference of estimated DPI (Bar) and ICC (dotted line) between 2 methods over study time period.

Figure 6 :
Figure 6: Bland-Altman plots between the DPI estimated from PNA and S-PFRQ.

Table 1 :
Participants' demographic and biochemical data.

Table 3 :
Estimated daily protein intake (DPI) by various clinical factors.CKD � chronic kidney disease, DPI � daily protein intake, PNA � protein-equivalent of total nitrogen appearance, S-PFRQ � short-protein food recall questionnaire.Mean diference � DPI estimation from S-PFRQ minus DPI estimation from PNA.

Table 2 :
T-test comparison of DPI estimation from the 2 methods in diferent stages of CKD.