Open access Protocol Rationale and population-based prospective cohort protocol for the disadvantaged populations at risk of decline in eGFR (CO-DEGREE) Marvin Gonzalez-Quiroz,‍ ‍ 1,2,3 Dorothea Nitsch,3 Sophie Hamilton,4 Cristina O'Callaghan Gordo,5,6,7 Rajiv Saran,8,9 Jason Glaser,10 Ricardo Correa-Rotter,11 Kristina Jakobsson,12,13 Ajay Singh,14 Nalika Gunawardena,15 Adeera Levin,16 Giuseppe Remuzzi,17 Ben Caplin,‍ ‍2 Neil Pearce,18,19 on behalf of the DEGREE Study Steering Committee To cite: Gonzalez-Quiroz M, Abstract Nitsch D, Hamilton S, et al. Strengths and limitations of this study Introduction A recently recognised form of chronic Rationale and population- kidney disease (CKD) of unknown origin (CKDu) based prospective cohort ►► We propose a prospective generic cohort protocol is afflicting communities, mostly in rural areas in protocol for the disadvantaged for populations affected by chronic kidney disease several regions of the world. Prevalence studies are populations at risk of decline in of unknown origin in which the sampling frame eGFR (CO-DEGREE). BMJ Open being conducted in a number of countries, using a consists of the entire at-risk population. In addition, 2019;9:e031169. doi:10.1136/ standardised protocol, to estimate the distribution of the use of this standardised protocol will allow for bmjopen-2019-031169 estimated glomerular filtration rate (eGFR), and thus regional and international comparisons. identify communities with a high prevalence of reduced ►► Prepublication history and ►► Serial estimated glomerular filtration rate measure- glomerular filtration rate (GFR). In this paper, we additional material for this ments in an apparently healthy population will al- propose a standardised minimum protocol for cohort paper are available online. To low the description of the evolution of disease and view these files, please visit studies in high-risk communities aimed at investigating reduce problems associated with recall bias and the journal online (http://​dx.​doi.​ the incidence of, and risk factors for, early kidney reverse causation when assessing potential risk org/​10.​1136/​bmjopen-​2019-​ dysfunction. factors. 031169). Methods and analysis This generic cohort protocol ►► Samples will be analysed in a single batch at the provides the information to establish a prospective end of the study to minimise time-dependent mea- BC and NP contributed equally. population-based cohort study in low-income settings surement errors. with a high prevalence of CKDu. This involves a baseline ►► A biobank is expected to be created in each centre Received 19 April 2019 Revised 03 July 2019 survey that included key elements from the DEGREE to store biological samples for future analyses. Accepted 05 July 2019 survey (eg, using the previously published DEGREE ►► As for any cohort, loss to follow-up could pose a methodology) of a population-representative sample, threat to validity of the study, and every effort must and subsequent follow-up visits in young adults be made to mitigate this. (without a pre-existing diagnosis of CKD (eGFR<60 mL/min/1.73m2), proteinuria or risk factors for CKD at baseline) over several years. Each visit involves Introduction a core questionnaire, and collection and storage of A mysterious form of chronic kidney biological samples. Local capacity to measure serum disease (CKD) is afflicting young adults, creatinine will be required so that immediate feedback mostly in rural communities in a number of on kidney function can be provided to participants. After completion of follow-up, repeat measures of creatinine low-income and middle-income countries should be conducted in a central laboratory, using (LMICs).1–10 This disease has been termed reference standards traceable to isotope dilution mass CKD of undetermined cause (CKDu). spectrometry (IDMS) quality control material to quantify Several definitions for CKDu exist; the the main outcome of eGFR decline over time, alongside criteria typically include demonstration © Author(s) (or their employer(s)) 2019. Re-use a description of the early evolution of disease and risk of renal damage using biomarkers in the permitted under CC BY. factors for eGFR decline. absence of diabetes, severe hypertension or Published by BMJ. Ethics and dissemination Ethical approval will be evidence of alternative renal diagnoses.11–14 For numbered affiliations see obtained by local researchers, and participants will This syndrome has caused thousands of end of article. provide informed consent before the study commences. deaths and reduced the life expectancy Participants will typically receive feedback and advice among young adults in Mesoamerica, South Correspondence to on their laboratory results, and referral to a local health Dr Marvin Gonzalez-Quiroz; Asia and possibly in other tropical/subtrop- system where appropriate. ​m.​quiroz@​ucl.​ac.​uk ical regions of the world.7 15–19 The cause(s) Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 1 Open access of CKDu are not yet established, but proposed aetiol- study group, without requiring that a DEGREE survey is ogies include recurrent dehydration/heat stress, pesti- conducted first. cides, infections and heavy metals.1 20–22 In addition, We are already conducting such a cohort study in Nica- there is no evidence that these forms of CKDu have a ragua,25 26 and have had many challenges to address, unified causality or are due to different aetiologies in including (1) community engagement, awareness of diverse parts of the world. conditions, political unrest and ethics; (2) follow-up over Although a broad range of cross-sectional studies time (frequency and minimising loss to follow-up); (3) investigating prevalence of CKDu have been conducted fieldwork and laboratory standards to ensure decline is in Mesoamerica, South Asia and other regions of the detected and (4) regular feedback information on study world,1–7 9 17 these have generally not used standardised progress. We will draw on our experience in Nicaragua methodology, and therefore do not allow for valid inter- in presenting both the generic CO-DEGREE protocol national comparisons. A recently published standardised and observations on the practical issues involved in protocol (the Disadvantaged Populations eGFR Epide- conducting such studies in a particular population. miology Study (DEGREE) protocol) for estimating the population distribution of glomerular filtration rate (eGFR) has addressed this concern, and is being used in communities suspected to have a high prevalence of Objectives reduced eGFR. The DEGREE protocol makes it possible Studies using this generic cohort protocol, and contrib- to undertake comparisons internationally, by mandating uting to the wider DEGREE collaboration, will aim to a population-representative sample and standardised 1. Investigate the evolution of, and risk factors for, kidney collection of information on sociodemographic factors, function decline over time among populations at risk occupational and environmental exposures, body compo- of CKDu. sition and kidney function.23 To date, studies using the 2. Compare the evolution, and risk factors for kidney DEGREE methodology have been conducted in four function decline, in different populations and regions countries (Peru, Sri Lanka, India and Malawi), with a at risk of CKDu. number of future projects in preparation or in progress.17 3. Establish a framework for international collaboration A recent meta-analysis highlighted the lack of robust and promote a network for future work on the causal- studies that have considered risk factors for early kidney ity of CKDu. damage in CKDu.24 This is of key importance as those with even apparently mildly damaged kidneys (eg, a Rationale for a cohort study of decline in eGFR borderline elevated serum creatinine (sCr) but no renal A representative sample of those at risk reserve) may experience progressive renal decline in Population-based cohort studies have several advan- response to a wide range of exacerbating insults (eg, tages:27 first, this type of study allows the recruitment episodes of dehydration/heat stress, nephrotoxic medi- of a representative sample of the at-risk population, for cation or other nephrotoxic exposures) making identi- example, it will include workers from a variety of occu- fication of causal associations challenging in those with pations (including unemployed) at the community level. existing kidney damage. Based on our experience,25 26 Assuming that the study sample is randomly selected we propose a generic cohort protocol to characterise the from the entire at-risk population based on a commu- decline in kidney function over time and conduct aetio- nity census, and there are no substantial problems with logical research in those without pre-existing CKD/risk non-response, these studies are unlikely to be affected factors at baseline but at risk of CKDu. Our focus is on by significant selection bias. Furthermore, in contrast conducting such cohort studies in populations that are to studies conducted solely in an occupational setting, at high risk for CKDu, that is, that have previously been differential loss to follow-up is likely to be less prob- classified as such by surveys based on cross-sectional eGFR lematic, particularly if workers are screened for kidney measurements. In general, this work would follow on disease within that setting and potentially denied further from a study using the DEGREE protocol, and hence we work. will use the term ‘CO-DEGREE’ (cohorts based on the Like all prospective cohort studies, to ensure the entire DEGREE study) for such studies. Indeed, in some situ- population is ‘at risk’, those with the outcome at baseline ations, a DEGREE survey may form the ‘baseline’, with should be excluded, although it is recognised that inves- a subgroup of DEGREE survey participants then being tigators may wish to follow up those with eGFR <60 mL/ selected for follow-up based on age, a single measure- min/1.73 m2 and those with established risk factors for ment of eGFR ≥60 mL/min/1.73 m2 (accepting that this CKD for other purposes (see below). is likely a conservative cut-off for pre-existing kidney One general disadvantage of population-based studies dysfunction) and without clinical diagnosis or history of is that this approach typically requires large sample sizes hypertension, diabetes mellitus, obesity or other known and long-term follow-up if disease is not highly prevalent. risk factor that could potentially explain CKD. However, However, the focus of CO-DEGREE is on conducting the standardised protocol we propose here can also be studies in population with a high prevalence of CKDu used as a ‘stand-alone’ study design in any well-defined (see below).25 27 2 Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 Open access Handling reverse causation and recall bias Notably, the accuracy of sCr determinations is also an The problem of reverse causation (eg, modification of inherent problem (see further below). In addition, the behaviour or work tasks in response to the diagnosis of Chronic Kidney Disease Epidemiology Collaboration renal impairment) can be minimised in a cohort study by (CKD-EPI) or The Modification of Diet in Renal Disease focusing on people without pre-existing disease, and then Study (MDRD) equations used to calculate eGFR from following these initially apparently ‘healthy’ participants sCr28 have not been validated in many populations over time. Similarly, a cohort approach, unlike cross-sec- reported to be suffering CKDu,29 potentially further tional studies, is less prone to recall bias regarding increasing misclassification bias in cross-sectional studies. previous exposures. Alternative approaches based on serial eGFR measure- ments in the same person over time render between-person Measuring kidney function variation less problematic. If estimated across a period of Quantification of kidney function is most easily under- taken by determining sCr concentration, which is rela- time using multiple measures with sustained preanalytical tively easy and cheap to measure, and then calculating and analytical quality, this will also reduce the influence the eGFR. A case of CKDu is typically defined by an eGFR of the within-person factors that are not directly related <60 mL/min/1.73 m2 (sustained for at least 3 months to to kidney damage. In summary, an approach using serial confirm chronicity) in the absence of known causes of eGFR measures substantially improves the potential to kidney disease. However, this dichotomous definition has identify risk/causal factors for CKDu as well as allowing weaknesses in studies exploring the causation of CKDu, the description of the evolution of disease. as it is well established that substantial damage may have already occurred at the histological level before serum biomarkers of renal dysfunction become abnormal (and other markers such as proteinuria are often absent in this Core protocol disease). Furthermore, repeat measures after 3 months Study design are not always performed in cross-sectional surveys, and This is a prospective cohort study protocol for studying sCr levels are modified by multiple non-renal factors decline in kidney function over time in populations such as high animal protein intake, strenuous exercise, with high reported prevalence of CKDu, primarily in changes in plasma volume, body mass index, sex, age, LMICs. We consider the following study design issues: ethnicity and some drugs28; thus, cross-sectional studies (1) population sampling strategy and follow-up interval, examining associations with reduced eGFR based on a (2) questionnaire development and delivery, (3) clinical single sCr measurement may be prone to a significant measurements and biosampling and (4) data manage- degree of misclassification, especially in smaller studies. ment and reporting25 (see figure 1). In addition, we Figure 1 Flowchart and study procedures of CO-DEGREE protocol. Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 3 Open access discuss (1) sample size and follow-up duration and (2) (seated, average BP ≥140/90 mmHg on second and third ethical considerations. of three readings).34 In addition to self-report of CKD, those with previously Population, sampling strategy and follow-up interval detected eGFR <60 mL/min/1.73 m2, proteinuria (eg, In Mesoamerica, CKDu typically affects young men on the albumin/creatinine ratio (ACR) >300 mg/g or dipstick Pacific Coast. This population is dying in their 40s, often 3+ or greater)35 on testing at baseline should be excluded younger, from end-stage renal disease.15 30 The disease from the study. It is recognised that a proportion of partic- appears to occur at a later age in South Asia, with few ipants not excluded by these criteria may still have some cases occurring in men in their 20s.7 31 Nevertheless, one form of underlying kidney abnormality (eg, low-level might expect preliminary changes in GFR to occur early proteinuria), and some of those excluded due to a low in adulthood. In general, the study population should eGFR at baseline may go on to recover function, but this include participants who are old enough to experience represents a pragmatic approach to excluding those with an identifiable decline in kidney function, but not older significant pre-existing renal disease at baseline. Further- age groups (eg, >60 years old) where the prevalence of more, for practical, ethical or scientific reasons (eg, to CKD is already high in many populations globally (eg, up gain insight into progression of established CKDu or to 10%). Thus, inclusion criteria should be tailored to other non-communicable disease research aims), investi- the local disease profile, but the default approach should gators may wish to study an entire population (including be to recruit participants aged 18–40 years old (though those with pre-existing clinical diagnosis of, or newly 18–30 might be more appropriate in Central America, and 18–50 may be more appropriate in areas such as identified, CKD, diabetes mellitus and hypertension), South Asia where age of onset appears older). The ratio- but in that case, it is important to ensure that there are nale for including people ≥18 years old was based on sufficient ‘disease-free’ participants included at base- definition on adult life, and may be lowered, especially line to meet the sample size requirements (see table 1). in populations where the working life starts years earlier. Although the disease is generally more common in men, A population-census should be conducted to identify women with CKDu are of strong scientific interest in that all potential participants in the appropriate age range they may suggest alternative risk factors, or help to rule and either the entire population recruited, or a random out some that have been previously proposed. Hence, sample selected. In either case, response rates by age and recruitment should in general involve equal numbers of sex should be reported. men and women, though women who are pregnant at The focus of these studies is to conduct aetiological recruitment are also excluded, since pregnancy-related research in those without traditional CKD/risk factors at changes in eGFR are challenging to interpret. baseline; thus, the sample size estimates (see below) are The baseline study visit will require the administration based on following a cohort in which those with evidence of the core questionnaire, with additional context-spe- of pre-existing CKD, diabetes or hypertension have been cific additions, clinical measurements and biological excluded.25 Diabetes can be diagnosed by self-report, samples. Subsequent to the baseline visit, follow-up visits use of medication or lab tests (fasting serum glucose: should be conducted at least annually for a minimum ≥7.0 mmol/L or HbA1C ≥48 mmol/mol),32 33 and hyper- follow-up of 2 years to evaluate the study outcome and tension by self-report, use of medication or measurement keep close contact with the participants and update their Table 1 Sample size calculations Parameters Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Scenario 6 Scenario 7 Scenario 8 Population frequency of 0.04 0.06 0.08 0.10 0.04 0.06 0.08 0.10 eGFR decline Proportion population 0.5 exposed OR associated with 2 3 exposure P (outcome|unexposed) 0.027 0.04 0.053 0.066 0.02 0.03 0.04 0.05 P (outcome|exposed) 0.054 0.08 0.106 0.132 0.06 0.09 0.12 0.15 Group size 993 686 405 436 463 317 243 200 Sample size 1986 1372 810 872 926 634 486 400 Calculations based on equal proportion of the population exposed/unexposed for simplicity. No adjustments made for loss to follow-up or multiple testing. P indicates probability, assumes 1-β=0.80; =0.05. eGFR, estimated glomerular filtration rate. 4 Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 Open access Table 2 Details and procedures of the baseline study visit and subsequence follow-up Follow-up period (variable) Baseline visit At Items (0 month) 12 months 24 months 36 months 48 months completion Community census X – – – – – Participants enrolment X – – – – – Informed consent X           Update personnel contact information X X X X X   Anthropometric measurements X X X X X   Biological samples X X X X X   Baseline core questionnaire X – – – –   Follow-up questionnaire   X X X X   Local serum creatinine measurement X X X X X   Results feedback X X X X X   Biobank X X X X X   Batch testing of serum creatinine           X contact information. This will help minimise the loss to (online supplementary file 2), provide biosamples and follow-up at each study point. Substantial seasonal varia- update their contact information. tion in eGFR has been reported in a number of settings (both CKDu related and unrelated).26 36–38 Therefore, the Clinical measurements conduct of additional study visits at a 6-monthly interval Blood pressure and heart rate should be measured on the (eg, at beginning and end of summer season) might be right arm after 5 min rest in the sitting position using an useful in explaining within-person eGFR variation as well automated sphygmomanometer, WHO validated for the as providing important information for the wider popu- clinical setting (eg, Omron HEM-907XL sphygmoma- lation on the significance of kidney function testing at nometer), and the average of the first, second and third different time points in the year (perhaps for a subset of of three readings is recorded. Subjects height and weight participants or a proportion of the follow-up period) (see (in meters and kilograms) should be measured (without table 2). shoes) using a stadiometer and digital calibrated scales. Questionnaires Biosamples The purpose of the baseline core questionnaire is to Fasting blood and urine samples will be collected at each obtain a minimum dataset to explore associations with study visit and stored in the field into coolers with icebox decline in kidney function and make comparisons within (4°C) no more than 4 hours before processing. and between persons. The baseline core questionnaire Dipstick urinalysis should be performed by using elec- (online supplementary file 1) is based on the question- tronic readers (urine chemistry analyser) where possible, naire used in the DEGREE protocol and has been used or otherwise at least 10% of tests should be re-analysed by in DEGREE-related studies in a number of settings. The a second investigator. Parameters that should be reported baseline core questionnaire represents a minimum dataset are urinary specific gravity, pH, protein, blood, leuco- and it will provide basic information on exposures such cytes, nitrite, glucose, etcetera. Investigators with access as sociodemographic factors, occupational and environ- to ACR measurements may wish to perform these assays mental exposure, lifestyle, diagnosis of infectious diseases (at least at baseline and annually). and medication. Local research teams may decide to add Samples for serum analysis should be centrifuged at data items of specific interest to the core dataset, partic- 3500 rpm for 10 min within 4 hours of collection, and ularly items of relevance to societal and occupational subsequently separated into at least four aliquots of context and/or environmental samples. They also have 1–2 mL and stored at ≤−20°C (ideally −80°C). One aliquot the responsibility to translate, validate and to make any should be used for contemporary sCr measurements, local contextual changes. Training procedures for the for example, by using the modified Jaffe assay (ideally field staff should be documented. also using standards traceable to isotope dilution mass Researchers will return to field (at least) annually for spectrometry (IDMS) reference material). At baseline in-person follow-up visits. At these follow-up visits, partici- and during each study visit, a cross-checking of local lab pants are invited to respond to a follow-up questionnaire quality control is highly recommended to ensure that sCr Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 5 Open access determinations are comparable as these lab results may population sample in Europe or the USA with similar guide referral to clinical care for participants during the exclusion criteria, there would be very little or no decline follow-up period. A further aliquot should be stored for a of kidney function in the young adult population. In repeat batch measurement of sCr in all samples (a subset contrast, in our Nicaragua study of apparently healthy of samples from each study visit will be adequate if IDMS adults aged 18–30 years,25 26 there was a clearly distinct referenced methods are used on initial measurement) at subgroup which experienced a marked decline in kidney the end of follow-up using a method traceable to an IDMS function over a short time, whereas the eGFR in the other reference material (and potentially also cystatin C). study participants was relatively stable. Given this distri- The CO-DEGREE group suggests the storage of at least bution of such eGFR trajectories in the population, we a further two 1–2 mL aliquots of serum and a similar would expect any analysis of risk factors to be conducted amount of urine in addition to those described above. using a prospective case-control approach. Additional samples and analyses should be pursued Therefore, the sample size requirements to detect an depending on the priorities of the local research team. association with an exposure at any given power will be All samples for future analysis should be stored at ≤−20°C determined by the following factors: (ideally −80°C) in a local or international biobank. Such 1. Proportion of the population that experience ‘substan- a biobank requires an uninterruptible power supply to tial’ decline. In turn the power to detect ‘substantial’ protect the samples. decline will depend on the following: Investigators should assess (as part of their public a. The rate of eGFR decline in those affected. engagement efforts) and, if appropriate, obtain consent b. The duration of follow-up. from participants for future use of samples for further c. The number of eGFR measures. (specific and/or more general) use both locally and 2. Proportion of general population exposed to any ex- internationally (eg, through the DEGREE collaboration) posure of interest. as well as ensure that storage capacity is available. 3. Effect size of any exposure. 4. The study retention rate. Data management and reporting Taking a simplistic approach, the duration of the study Questionnaires and samples will be labelled using a should be designed so that those affected have sustained a unique bar-code to maintain participant confidentiality. clinically important loss of kidney function, for example, Electronic data capture systems such as Open Data Kit39 20% of normal eGFR. Therefore, if CKDu in the study may be the most resource efficient method to capture population is predicted, from a baseline of ≥60 mL/min questionnaire data but where hard copies are used, eGFR, to lead to a loss of eGFR of a magnitude of 5% double data-entry should be undertaken to minimise the each year (~7 mL/min/1.73 m2/year), the study dura- transcription errors. tion should be 4 years. If, alternatively, loss is predicted The CO-DEGREE protocols are openly available to to be 10% each year, study duration could be as short as interested research teams. Although primarily designed 2 years. Additional eGFR measures, over and above the to be used in population-based studies, similar approaches suggested annual frequency, will reduce error associated could also be used in an occupational or other selected with determining trajectory (and might be performed for cohorts. the reasons discussed above) but either way a minimum Each centre will be the ‘owner’ of their data and follow-up of 2 years is recommended. expected to publish the results of their study inde- After basing the study duration on the expected rate of pendently. However, where a study is registered as part of eGFR decline among those affected, the sample size can the DEGREE collaboration, the coordinating centre will then be calculated on the basis of the expected frequency request a digital copy of anonymised individual-level data of ‘substantial’ decline among the population and the to allow the undertaking of international comparisons. In effect size of any proposed exposure that it is desirable addition, a summary of local contextual information and to detect. A number of scenarios are outlined in table 1. a description of the population characteristics along with A further (eg, 20%, depending on local circumstances) response rates will be requested. The importance of such increase in target recruitment is advised to allow for loss information is emphasised. to follow-up. Finally, these initial sample sizes will need adjustment Sample size and follow-up duration for exclusions based on estimates of the prevalence of The overall size of the cohort will be largely dependent previously unknown CKD (based on eGFR/albuminuria on the proportion of the ‘healthy’ population which is tests), diabetes, hypertension or other known causes of expected to experience a ‘substantial’ decline in eGFR CKD at baseline (unless these data are already available over time in the community as a result of CKDu. As from a previously conducted cross-sectional study). It is discussed above, demonstrating that reduced renal func- worth considering whether people who may have CKD tion without diabetes, hypertension or known kidney (or CKD risk factors) will be aware of this, as this may diseases is prevalent on a cross-sectional basis is a neces- affect the numbers of participants that will be retained sary first step before pursuing this work. If, for example, for the analysis following testing. For example, if there this study protocol was to be conducted in a general is screening for kidney problems (as in some Central 6 Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 Open access American Sugarcane mills or community-based screening and retention rates. As the workday starts very early in the in Sri Lanka), then potential cohort participants may be morning and finishes late in the afternoon, attempts were aware of their kidney function status and can be excluded made to conduct data collection during economically less from the study sample prior to recruitment. For example, active (eg, each side of the main sugar harvest) periods 5% of the target population in the community studied in of the year, so as to still capture approximately 30% of Nicaragua reported pre-existing CKD. Nevertheless, there participants who were employed at the time. Addition- was an additional 10% who had undiagnosed impaired ally, participants receive their kidney test results within a kidney function at baseline assessment based on their fortnight of the study visits and receive reimbursement laboratory findings, highlighting the importance of iden- of expenses and any lost income they have incurred to tifying an age group where CKDu is not already highly attend the study visit. Although study visits have been time- prevalent so as to satisfy a key inclusion criterion (absence tabled to occur outside of the harvest season, employees of CKD at baseline) when calculating sample sizes. still express the concern that their employment opportu- nities might be affected by taking part in the study. In an Ethics and dissemination attempt to mitigate against these types of consequences, Local research teams will ensure these studies are the study team have corresponded with local employers conducted in accordance with the Declaration of Helsinki explaining the content and extent of this study in order Principles and be responsible for assuring that the work is to reduce any concerns about workers’ participation. In approved by the local institutional review board. Written addition, the study team takes particular precautions to informed consent will be obtained from all participants maintain participant’s confidentiality during the study before taking part in the study. Information should be and beyond. transparent in terms of using the data and biosamples Conducting a follow-up study in a rural area remains stored for future research. Typically, a key aspect of the a major challenge. Alongside the logistical challenges ethical review of any protocol is a discussion surrounding of reaching geographically isolated neighbourhoods the provision of feedback and advice to participants when along poor-quality roads, a significant obstacle has been abnormal results become available. In most settings, these internal and external migration due to lack of employ- processes should be developed in partnership with local ment source or social unrest. Rural communities have a communities. Furthermore, mechanisms will need to be tradition of working with seasonal crops, and sugarcane established in collaboration with local health providers/ workers often leave their communities at the end of each healthcare systems to define pathways for participants harvest season, to go abroad or to other regions within needing referral for medical care. Findings from these the country in search of temporary employment. In our studies should be disseminated widely by publication in study, at the end of each harvest, up to 30% of the study peer-reviewed journals and presentations/representa- population had left their communities in search of alter- tions to relevant local stake holders. native employment during the non-harvest period in our study. Despite these problems, our team achieved Patient and public involvement attendance at 92% of all scheduled visits over 2 years.25 26 Patients or member of the public were not involved in However, the level of investment of time and resources the design of this protocol. Procedures will vary by loca- should not be underestimated. tion; however, the DEGREE Steering Committee would Finally, continuing community engagement and the encourage active involvement of lay members of study maintenance of good relationships between researchers, communities in additional design elements and imple- community leaders, participants and communication with mentation of these studies, particularly relating to the local healthcare system have been key. The development ethical issues above. For example, it is expected that study of standardised procedures for use by the research team participants will receive the results of their lab tests, expla- may be useful in this context, for example, a reference nations of them and a reference to the relevant health flowchart for communication with local health posts/ centre, if appropriate. However, the best mechanisms for primary hospital or hospital for persons with health prob- doing this will vary by location. lems detected during the study. Experience with the CO-DEGREE protocol in Nicaragua The protocol presented here is, by necessity, generic. The approaches and challenges of implementing the protocol Discussion will vary widely in different populations and regions of The CO-DEGREE protocol was developed in response the world. However, since we have already implemented to the highly prevalent form of CKD of unknown cause this protocol in a study in Nicaragua,25 26 we will make that is affecting Mesoamerica and other countries around some observations on the practicalities, and challenges, the globe. To date, the existing epidemiological studies or implementing the protocol in this context. of CKDu have provided an incomplete understanding of The Nicaragua study involved community-based the evolution of and risk factors for disease. This CO-DE- follow-up in Leon and Chinandega departments.25 A GREE protocol aims to provide a framework to address number of strategies were used to maximise response this. Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 7 Open access 11 This CO-DEGREE protocol is designed to capture the Department of Nephrology and Mineral Metabolism, National Medical Science and entire at-risk population by aiming to recruit men and Nutrition Institute Salvador Zubirán, Mexico City, Mexico 12 Department of Public Health and Community Medicine, Insitute of Medicine, women, and those who work across a variety of different Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden occupations. The main outcome measure of within-person 13 Occupational and Environmental Medicine, Sahlgrenska University Hospital, loss of eGFR over time, which means it should be possible Region Västra Götaland, Gothenburg, Sweden 14 to capture the earliest stages of disease, making associ- Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, USA ations with possible causal exposures (and exacerbating 15 World Health Organization Country Office, Colombo, Sri Lanka factors) less prone to reverse causation and recall bias. 16 Division of Nepohrology UBC, University of British Columbia, Vancouver, British We do not underestimate the challenges posed by the Columbia, Canada 17 lack of language-validated and standardised exposure 18 Instituto di Ricerche Farmacologiche Mario Negri - IRCCS, Milan, Italy questionnaires in this area. The accompanying question- Deparment of Medical Statistics and Non-Communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, UK naire represents a minimum, and most studies will use an 19 Centre for Global NCDs, London School of Hygiene and Tropical Medicine, London, expanded dataset. Currently, there is an absence of glob- UK ally generalisable instrument to capture environmental and occupational exposures; however, the DEGREE Twitter Marvin Gonzalez-Quiroz @@MarvinGonzlez16 group is undertaking further work in this area. Addition- Acknowledgements The authors wish to thank all DEGREE Steering Committee ally, short-term or long-term environmental measure- members. ments and/or novel biomarkers that capture exposure Collaborators DEGREE Steering Committee members: Neil Pearce (UK) (Chair), to heat, agrichemicals and/or infection in either the Ben Caplin (UK) (Co-chair), Jason Glaser (USA), Ricardo Correa-Rotter (Mexico), Kristina Jakobsson (Sweden), Ajay Singh (USA/India), Antonio Bernabe-Ortiz (Peru), community or workplace are likely to be valuable addi- Emmanuel Burdmann (Brazil), Marvin Gonzalez-Quiroz (Nicaragua), Vivekanand Jha tions to this type of study but are beyond the scope of this (India), Rick Johnson (USA), Phabdheep Kaur (India), Pronpimolk Kongtip (Thailand), basic protocol. Hans Kromhout (Netherlands), Adeera Levin (Canada), Magdalena Madero Rovalo (Mexico), Dorothea Nitsch (UK), Moffat Nyirenda (Ugand/Malawi), Cristina Finally, it should be emphasised that this protocol is not O’Callaghan-Gordo (Spain), Pablo Perel (UK/Argentina), Dorairaj Prabhkaran (India), suitable for studying the progression of CKD in general, Narayan Prasad (India), Giuseppe Remuzzi (Italy), Rajiv Saran (USA), Liam Smeeth due to the specific constraints introduced by excluding (UK), Vidhya Venugopal (India). those with hypertension, diabetes and CKD as well as Contributors The CO-DEGREE protocol was conceived by MGQ, BC, DN and NP. other known causes of CKD (ie, those with proteinuria Design of the study and drafting the protocol was done by MGQ, BC, DN, NP, SH, COG, RS, JG, RCR, KJ, AS, NG, AL and GR. All authors contributed to and approved and/or with reduced eGFR) at baseline. Indeed, in the final manuscript. settings where there is not a high prevalence of CKDu, a Funding This work was supported by grants from the UK Colt Foundation and the cohort comprising people without traditional risk factors UK Medical Research Council (MR/P02386X/1). for CKD or with CKD would be unlikely to identify any Competing interests None declared. detectable kidney function loss over time in the young– Patient consent for publication Not required. adult population. For studies outside the CKDu arena, Provenance and peer review Not commissioned; externally peer reviewed. investigators are advised to use alternative methodolo- gies using established protocols, for example, the CRIC Data availability statement Data are available in a public, open access repository. There are no data in this work. Data are available upon reasonable request. Data study.40 may be obtained from a third party and are not publicly available. No data are In conclusion, we have designed a CO-DEGREE available. All data relevant to the study are included in the article or uploaded as protocol that can be used in the different settings around supplementary information. the globe to investigate the evolution of CKDu and the Open access This is an open access article distributed in accordance with the associated risk factors for decline in kidney function. Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any These studies should provide important information purpose, provided the original work is properly cited, a link to the licence is given, on the early decline in kidney function across different and indication of whether changes were made. See: https://​creativecommons.​org/​ affected areas as well as key insight into the cause(s) of licenses/​by/​4.​0/. disease. Author affiliations References 1 Research Centre on Health, Work and Environment (CISTA), National Autonomous 1. Wegman D, Crowe J, Hogstedt C. Mesoamerican nephropathy: report from the second International research workshop on men. University of Nicaragua, León, Nicaragua Report No.: ISBN 978-9968-924-33-7. Heredia, CR: SALTRA/IRET- 2 Centre for Nephrology, University College London, London, UK UNA, 2016. 3 Department of Non-Communicable Disease Epidemiology, London School of 2. Torres C, Aragón A, González M, et al. Decreased kidney function Hygiene and Tropical Medicine, London, UK of unknown cause in Nicaragua: a community-based survey. Am J 4 Kidney Dis 2010;55:485–96. School of Public Health, Faculty of Medicine, Imperial College London, London, UK 5 3. O'Donnell JK, Tobey M, Weiner DE, et al. Prevalence of and risk Campus Mar, Instituto de Salud Global Barcelona, Barcelona, Spain factors for chronic kidney disease in rural Nicaragua. Nephrol Dial 6 CIBER Epidemiología y Salud Pública (CIBERESP), Madrid, Spain Transplant 2011;26:2798–805. 7 Universitat Pompeu Fabra (UPF), Barcelona, Spain 4. Orantes CM, Herrera R, Almaguer M, et al. Chronic kidney disease 8 Division of Nephrology, Department of Internal Medicine & Epidemiology, University and associated risk factors in the Bajo Lempa region of El Salvador: Nefrolempa study, 2009. MEDICC Rev 2011;13:14–22. of Michigan, Ann Arbor, Michigan, USA 9 5. Orantes CM, Herrera R, Almaguer M, et al. Epidemiology of chronic University of Michigan, Ann Arbor, Michigan, USA kidney disease in adults of Salvadoran agricultural communities. 10 La Isla Network, Washington DC, UK MEDICC Rev 2014;16:23–30. 8 Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 Open access 6. Jayasekara JM, Dissanayake DM, Adhikari SB, et al. Geographical 23. Caplin B, Jakobsson K, Glaser J, et al. International Collaboration distribution of chronic kidney disease of unknown origin in North for the Epidemiology of eGFR in Low and Middle Income central region of Sri Lanka. Ceylon Med J 2013;58:6–10. Populations - Rationale and core protocol for the Disadvantaged 7. Jayatilake N, Mendis S, Maheepala P, et al. Chronic kidney disease of Populations eGFR Epidemiology Study (DEGREE). BMC Nephrol uncertain aetiology: prevalence and causative factors in a developing 2017;18:1. country. BMC Nephrol 2013;14:180. 24. González-Quiroz M, Pearce N, Caplin B, et al. What do 8. Ganguli A. Uddanam Nephropathy/Regional nephropathy in India: epidemiological studies tell us about chronic kidney disease of preliminary findings and a plea for further research. Am J Kidney Dis undetermined cause in Meso-America? A systematic review and 2016;68:344–8. meta-analysis. Clin Kidney J 2018;11:496–506. 9. Jayasumana C, Orantes C, Herrera R, et al. Chronic interstitial 25. González-Quiroz M, Camacho A, Faber D, et al. Rationale, nephritis in agricultural communities: a worldwide epidemic with description and baseline findings of a community-based prospective social, occupational and environmental determinants. Nephrol Dial cohort study of kidney function amongst the young rural population Transplant 2017;32:234–41. of northwest Nicaragua. BMC Nephrol 2017;18:16. 10. Peraza S, Wesseling C, Aragon A, et al. Decreased kidney function 26. Gonzalez-Quiroz M, Smpokou E-T, Silverwood RJ, et al. Decline in among agricultural workers in El Salvador. Am J Kidney Dis kidney function among apparently healthy young adults at risk of 2012;59:531–40. Mesoamerican nephropathy. J Am Soc Nephrol 2018;29:2200–12. 11. García-Trabanino R, Cerdas M, Madero M, et al. Nefropatía 27. Ben C, Marvin G-Q, Neil P, SALTRA. Gaining insights into the mesoamericana: revisión breve basada en El segundo taller evolution of CKDnt from community-based follow up studies. In: del Consorcio para El estudio de la epidemia de Nefropatía en Second International workshop on Mesoamerican nephropathy. San Centroamérica Y México (CENCAM). Nefrología Latinoamericana José. Costa Rica: SALTRA, 2015. 2017;14:39–45. 28. Padala S, Tighiouart H, Inker LA, et al. Accuracy of a GFR estimating 12. Lozier M, Turcios-Ruiz RM, Noonan G, et al. Chronic kidney equation over time in people with a wide range of kidney function. disease of nontraditional etiology in central America: a provisional Am J Kidney Dis 2012;60:217–24. epidemiologic case definition for surveillance and epidemiologic 29. Inker LA, Schmid CH, Tighiouart H, et al. Estimating glomerular studies. Rev Panam Salud Publica 2016;40:294–300. filtration rate from serum creatinine and cystatin C. N Engl J Med 13. Rajapakse S, Shivanthan MC, Selvarajah M. Chronic kidney disease 2012;367:20–9. of unknown etiology in Sri Lanka. Int J Occup Environ Health 30. García-Trabanino R, Trujillo Z, Colorado AV, et al. Prevalence of 2016;22:259–64. patients receiving renal replacement therapy in El Salvador in 2014. 14. WHO. Workshop report: designing a step-wise approach to estimate Nefrologia 2016;36:631–6. the burden and to understand the etiology of CKDu in Sri Lanka. Sri 31. Jha V, Garcia-Garcia G, Iseki K, et al. Chronic kidney disease: global Lanka: WHO, 2016. dimension and perspectives. Lancet 2013;382:260–72. 15. Ordunez P, Nieto FJ, Martinez R, et al. Chronic kidney disease 32. American Diabetes Association. 2. Classification and Diagnosis of mortality trends in selected central America countries, 1997-2013: Diabetes: Standards of Medical Care in Diabetes-2018. Diabetes clues to an epidemic of chronic interstitial nephritis of agricultural Care 2018;41(Suppl 1):S13–S27. communities. J Epidemiol Community Health 2018;72:280–6. 33. Association American Diabetes. Updates to the Standards of Medical 16. Ordunez P, Saenz C, Martinez R, et al. The epidemic of Care in Diabetes-2018. Diabetes Care 2018;41:2045–7. chronic kidney disease in central America. Lancet Glob Health 34. Chobanian AV, Bakris GL, Black HR, et al. Seventh report of the 2014;2:e440–1. joint National Committee on prevention, detection, evaluation, and 17. Ekiti ME, Zambo J-B, Assah FK, et al. Chronic kidney disease in treatment of high blood pressure. Hypertension 2003;42:1206–52. sugarcane workers in Cameroon: a cross-sectional study. BMC 35. KDIGO. KDIGO 2012 clinical practice guideline for the evaluation Nephrol 2018;19:10. and management of chronic kidney disease. Kidney Int Supps 18. Ministry of Health Nutrition and Indigenous Medicine. Medical 2013;3:1–150. statistics unit. annual health bulletin of Sri Lanka 2015. Sri Lanka: 36. Wesseling C, Aragón A, González M, et al. Kidney function in Ministry of Health, Nutrition and Indigenous Medicine, 2017. sugarcane cutters in Nicaragua--A longitudinal study of workers at 19. Rajapurkar MM, John GT, Kirpalani AL, et al. What do we know about risk of Mesoamerican nephropathy. Environ Res 2016;147:125–32. chronic kidney disease in India: first report of the Indian CKD registry. 37. Wesseling C, Aragón A, González M, et al. Heat stress, hydration and BMC Nephrol 2012;13:10. uric acid: a cross-sectional study in workers of three occupations in 20. Correa-Rotter R, Wesseling C, Johnson RJ. Ckd of unknown origin a hotspot of Mesoamerican nephropathy in Nicaragua. BMJ Open in central America: the case for a Mesoamerican nephropathy. Am J 2016;6:e011034. Kidney Dis 2014;63:506–20. 38. Laws RL, Brooks DR, Amador JJ, et al. Changes in kidney function 21. Wesseling C, Crowe J, Hogstedt C, et al. The epidemic of among Nicaraguan sugarcane workers. Int J Occup Environ Health chronic kidney disease of unknown etiology in Mesoamerica: a 2015;21:241–50. call for interdisciplinary research and action. Am J Public Health 39. Open Data KIT. Longitudinal clinic study APP: GitHub, Inc, 2018. 2013;103:1927–30. Available: https://​opendatakit.​org/​use/​2_​0_​tools/​odk-​application-​ 22. Wesseling C, Crowe J, Hogstedt C, et al. First international research designer-​2-​0-​rev126/ [Accessed 28 Jul 2018]. workshop on Mesoamerican nephropathy (men. Report No.: ISBN 40. Feldman HIet al. The chronic renal insufficiency cohort (CRIC) study: 978-9968-924-06-1. Heredia, C.R: SALTRA/IRET-UNA, 2013. design and methods. J Am Soc Nephrol 2003;14:148S–53. Gonzalez-Quiroz M, et al. BMJ Open 2019;9:e031169. doi:10.1136/bmjopen-2019-031169 9