The Design and Conduct of Cohort Studies of Peoples With CKD – International Perspectives From iNET-CKD

Abstract

Prospective cohort studies are important for understanding epidemiology, associations, and incidence of adverse outcomes of long-term conditions, including chronic kidney disease (CKD). Here, we draw on the combined expertise of the leaders of cohort studies within the ISN’s International Network of CKD Cohorts (iNET-CKD) to provide insights into the design and conduct of prospective cohort studies in people with CKD. Formulating research questions that will remain relevant years after the cohort launch, when data will be available, is the first, and perhaps most challenging step of the study design. These questions will inform participant selection, sample size, duration of follow-up, outcome definitions, as well as frequency of encounters and data collection. Consideration should be given to the data elements to be collected, balancing burden with data value. At least 1 marker of glomerular filtration rate (GFR) (preferably both serum creatinine and cystatin C concentrations) and a measure of proteinuria (preferably urine albumin-to-creatinine ratio) are essential. Standard operating procedures should be developed for all assessments and for biosample collection and storage. Robust governance arrangements should be established to ensure effective management and sustainability of these complex projects, along with sufficient funding for the proposed study duration. Facilitating future collaboration with other cohort studies should be born in mind, to expand scientific output. This includes adoption of standard definitions and assessments, as well as making provisions for sharing of data and biosamples. A clear communication strategy that includes all stakeholders is essential to ensure maximum clinical impact.

Introduction

Prospective randomized controlled trials (RCTs) are the gold standard for informing evidence-based treatments. Much effort has been devoted to optimizing the design of RCTs to ensure that the findings are robust and performed efficiently, in the shortest possible duration. However, RCTs have important limitations. First, they are designed to test a specific hypothesis, usually regarding the efficacy and/or safety of a novel treatment or intervention. Consequently, the research question is necessarily narrow. Second, RCTs tend to have restrictive inclusion and exclusion criteria, resulting in highly selected study population with uncertainty about the applicability of the results to other populations. Third, RCTs are designed to answer the research question in as short time-period as possible and, therefore, often are not suitable to assess long-term outcomes.

In contrast, prospective cohort studies are designed to assess the role or the predictive value of 1 or several exposures in predefined outcomes. Based on a priori hypotheses, prospective cohort studies may contribute to the understanding of a disease and generate insights into its underlying mechanisms, associations with other diseases, and long-term consequences. They often provide scientific output for generating new hypotheses and informing the design of interventional studies. Like RCTs, cohort studies require considerable effort to be initiated and sustained; however, they have unique challenges that can prevent attainment of their full potential. Perhaps because of the perceived lower status of prospective cohort studies in the hierarchy of evidence generation compared with RCTs, less attention has been given by the broader scientific community, including funders, and journal editors, to optimizing prospective cohort study design and conduct.

The International Network of CKD cohort studies (iNET-CKD) is a distinct committee within the ISN Research Working Group, which fosters collaboration across numerous prospective cohort studies from all regions of the globe. Currently, the collaboration includes 32 prospective cohorts from across all continents. Supplementary Table S1 includes data from 27 cohorts that responded to a questionnaire on data availability conducted between March 3 and July 21, 2024, from which an infographic was developed (https://www.theisn.org/ cohort-studies-inet-ckd/). The goal of the network is to enhance the understanding of how CKD progression and its consequences affect individuals, with a specific focus on variations between regions and countries, ultimately gaining insights that will facilitate and inform future research.

Two of the stated aims of iNET-CKD are to enhance research capacity around the world through education, especially because it pertains to the conduct of observational cohort studies of CKD; and to assist investigators in establishing CKD cohort studies. We therefore have drawn on the combined expertise from iNET-CKD cohort study leaders to share best practices and lessons learned from previous CKD cohort studies.

Cohort Study Design

Research Questions

Prospective cohort studies intend to answer research questions, based on a priori hypotheses. These questions constitute the basis for participant selection, sample size determination, data collection, length of follow-up, and outcome definitions. This contrasts with retrospective cohort designs that build upon data from electronic health records and registries, primarily collected to document patient care, optimize billing, and monitor epidemiological trends. Consequently, prospective cohort studies are less prone to measurement and confounding biases than other observational study designs.¹ Because they are resource-intensive and require long-term commitment from both investigators and participants, they typically aim to address a broad range of topics.²

The capacity of investigators to foresee the questions that will remain important years after the cohort launch, when data on outcomes will be available, determines the potential of the cohort study to impact patient care and public health.³ Over the past decade, game-changing drugs have been developed, and recognition of the importance of patient-reported outcomes (PROMs) have led to the emergence of new models of care.⁴ Their long-term impact, both in terms of efficacy and safety, needs to be thoroughly assessed in more representative, diverse populations. Future studies should focus on challenges emerging from environmental, cultural, and socioeconomic dynamics.^5,6 The continuous development of bioassays, imaging techniques, and sensor-rich wearable devices, along with advances in computing capacity and multimodal data integration, paves the way for more ambitious questions, including the development of more performant predictive tools, interaction of risk factors, and personalized treatment responses.⁷

All stakeholders, including patients, should participate in the process of identifying research questions for a new study. The ISN⁸ and other organizations advocating for kidney health^4,9 have provided a set of research recommendations that may be useful in designing relevant and coordinated cohort studies.

Participant Selection

In observational studies, one of the main goals of participant selection is obtaining a representative sample of its target population.² Strictly speaking, a representative sample is obtained when all individuals from a given population have an equal probability of being included in the study. This implies identifying or constructing a census framework, conceiving a survey design, and recruiting randomly selected participants with weighting for nonparticipation. Most, if not all, past or ongoing CKD cohorts cannot claim this level of representativeness. Although a strictly representative sample is required for the correct estimation of descriptive statistics, it is not necessary, and may even be counterproductive, for analytical purposes.¹⁰ Instead, a prospective cohort study sample should achieve representativeness in the sense that “the results estimated in that sample are generalizable to the target population.”¹¹ This pragmatic approach relies on assumptions that must be clearly stated in the study protocol and the discussion of results. It implies generalizability of study interpretation, and provided that the distribution of key confounding factors is the same as in the target population, generalization of the estimated effect sizes is reasonable. Prediction models developed within the cohort samples may, however, require recalibration to achieve clinical relevance. Because research results may vary from one population subgroup to another, adequate representation of specific characteristics of interest is paramount for advancing personalized CKD care; and may require selective overrepresentation of some subgroups.

Inequities in health outcomes and a lack of evidence to support existing interventions in certain groups underscore the need for greater diversity in participant selection.¹² Women and the elderly represent a significant proportion of people living with CKD; however, these groups are underrepresented in clinical trials.^13,14 Prospective cohort studies have played an important role in identifying differences in outcomes and treatment responses related to sex and age. At the other extreme of the age range, children present specific challenges (e.g., kidney function assessment, diverse and rare disease etiology), which could benefit from studies based on prospective cohorts.15, 16, 17 Ethnic minorities bear a greater burden related to CKD than the overall population because of multiple factors, including systemic racism and unequal representation in clinical research.¹⁸ There is an urgent need to disentangle the role of genetic factors, which are inconsistently correlated with geographic origin and physical traits, from that of the numerous social determinants that strongly underlie the constructs of race and ethnicity. An additional knowledge gap is the contributors to disproportionate prevalence and associated mortality of CKD in low- and low-middle income countries.¹⁹ Community engagement, decentralization, and the use of digital tools may enhance the reach and the diversity of participants in longitudinal studies.²⁰ The representation of all population groups among study investigators, grant evaluators, journal editors, and reviewers is key for preventing bias in research design, funding, and publication.

Sample Size and Length of Follow-Up

Sample size calculation is a tool to find the best trade-off between sufficient statistical precision to produce relevant research findings and practical constraints. It should be appreciated, however, that sample size calculation often relies on several unknown parameters within closed formulae that ignore the complex nature of observational data analysis (confounding, missing data, and informative censoring, etc.).21, 22, 23

The length of follow-up is a crucial aspect of determining the sample size. The required follow-up duration is intrinsically related to the study objectives—evaluation of short-, medium-, or long-term outcomes, predicting clinical events within a given time horizon, and the target population—although logistical aspects may weigh considerably in this choice. Given that CKD tends to progress slowly over years and decades, it is essential to plan for an adequate period of follow-up. It is important to consider the frequency of follow-up assessments. Again, this requires a balance between the scientific benefit of follow-up data and the logistical burden of arranging follow-up visits.

Besides the length of follow-up, sample size estimation for time-to-event analysis will require estimations of exposure frequency (or variance, in the case of a continuous exposure), event incidence in the reference group,²⁴ and hazard or event ratio. Studies focusing on statistical significance and power will further determine alpha and 1 − beta levels. However, the power approach for sample size calculation does not guarantee the relevant confidence intervals for study interpretation or meaningful translation to clinical practice. In the observational study framework, type II error (insufficient statistical power) is less likely to be the reason for a negative result than are confounding, measurement, and other biases.^22,25 Several methods have been proposed to estimate sample size to obtain a given precision,²⁵ including in the context of clinical prediction model development.²⁶ Perhaps, the simplest is that based on the desired width of the confidence interval relative to the effect size (Figure 1).²⁴ It is common (and advised) practice to inflate the sample size proportionally to the expected loss to follow-up.^27,28

Figure 1.

Relative width of the 95% confidence interval at each side of the expected rate ratio for the comparison of 2 equally sized groups as a function of sample size and annualized event rate (according to Dos Santos Silva²⁴).²¹ Annualized event rates were set to 0.05, 0.10, and 0.20 patient-year; study duration, to 6 years. No variance inflation factor has been considered. As an example, considering the comparison of 2 equally sized groups and an annualized event rate of 0.10 patient-year, relative widths of the 95% confidence interval at each side of the expected rate ratio of 26%, 18%, 12%, and 8% would require sample sizes of 500, 1000, 2000, and 5000, respectively. For an expected rate ratio of 1.20, these would correspond to 95% confidence intervals of 0.95 to 1.51, 1.02 to 1.42, 1.07 to 1.34, and 1.11 to 1.30.

Accounting for confounders in study analyses typically increases the variance of estimates. As the number of confounders to consider increases, so do the sample size requirements to maintain precision and avoid model sparsity.²⁹ In the context of prediction model development, the number of candidate parameters—covariates, but also potential modalities (for nominal variables), transformations, and interaction terms—is a key parameter to be considered in sample size calculations. Riley et al.²⁶ provide guidance on the consideration of shrinkage parameters and optimism in the model performance when calculating sample size, which may be particularly useful when developing prediction models with high-dimensional datasets, such as omics.

Data Elements

Primary data collection is a hallmark of cohort studies, involving the direct gathering of information from participants through clinical visits, surveys, and self-administered questionnaires. This approach allows researchers to collect detailed, high-quality data tailored to the study’s objectives, ensuring consistency and rigor. To ensure consistency, it is important to develop standard operating procedures for all assessments.

Data collection in a prospective cohort study is structured around 3 aspects, namely, exposures or predictor candidates, outcomes, and confounders. Consequently, the data to be collected will depend on the study’s research questions. Logistical and financial barriers often require trade-offs between quality and quantity. To date, efforts to generate consensus on the data elements for CKD research has been limited to outcomes in clinical trials, as developed by the Standardized Outcomes in Nephrology initiative,³⁰ and for specific topics such as vascular access.³¹ Previous publications from iNET-CKD have illustrated the lack of consistency of data in the existing CKD cohort studies. Approximately 40% of the participating cohorts lacked information to be included in the most comprehensive multivariable adjusted models of 2 distinct analyses.^32,33 Furthermore, these collaborations required substantial effort, and sometimes, loss of information to harmonize the definition of variables. This was particularly the case for albuminuria, which inconsistently collected as albumin- or protein-to creatinine ratio or dipstick, had to be categorized for analyses.

Recommending a minimum dataset for prospective CKD cohort studies is probably a far too ambitious goal for the present article. Instead, we provide a summary of the type of information collected by cohorts participating in the iNET-CKD (Table 1) and share insights from our experience with the collection and analysis of the most common CKD biomarkers and PROMs.

Table 1.

Summary of the information collected by cohorts participating in the iNET-CKD according to whether they were ascertained at baseline, over the follow-up, or as part of outcome assessment

Frequency of data collection among iNET-CKD cohorts (n = 27)	Assessed once	Repeatedly assessed over time	Outcome assessment
Always (n = 27)	Age, Sex, Ethnicity
	Hypertension, Diabetes
	Creatinine, serum albumin, phosphatea
Very frequently (27 > n ≥ 22)	Weight, height	Blood pressure	Kidney failure (any)b,c
	Blood pressure	Antihypertensive and lipid lowering therapies	All-cause death
	Smoking status	Creatinine, serum albumin, phosphate, hemoglobin, total cholesterola	Cardiovascular death (any)c
	History of cardiovascular disease		Cause-specific death (any)c
	Antihypertensive and lipid lowering therapies		Myocardial infarction (any)c
	Spot urine ACRa		Stroke (any)c
	Hemoglobin, total cholesterol, HDL cholesterol, LDL cholesterol, CRP, calcium, urea/BUN, PTHa		Peripheral artery disease (any)c
			Hospitalization for heart failure (any)c
Frequently (22 > n ≥ 16)	Waist circumference	Weight, height	All-cause hospitalizations
	Education attainment	Anemia and anticoagulant treatments	Patient reported outcomesd
	Alcohol consumption	Spot urine ACRa
	History of cancer and chronic liver disease	HDL cholesterol, LDL cholesterol, urea/BUN, bicarbonate, ferritin, PTH, Hemoglobin A1ca
	Anemia and anticoagulant treatments
	Spot urine PCRa
	Cystatin C, ferritin, TSAT, bicarbonate, alkaline phosphatase, 25-OH Vitamin D, Hemoglobin A1ca
Occasionally (16 > n ≥ 6)	Hip circumference	Waist and hip circumferences	Cardiovascular death (adjudicated)
	History of chronic lung- and thyroid diseases	Gastric acid inhibition, opioid pain medication	Acute kidney injury (any)c
	Gastric acid inhibition, opioid pain medication	Spot urine PCR, dipstick proteinuria, 24-h urine protein excretiona	Packed red cell transfusions
	Dipstick proteinuria, 24-h urine protein excretiona	Cystatin C, alkaline phosphatase, TSAT, CRP, 25-OH Vitamin D
	FGF-23a
Rarely (n < 6)		FGF-23a	Kidney failureb (adjudicated)
			Myocardial infarction (adjudicated)
			Stroke (adjudicated)
			Hospitalization for heart failure (adjudicated)
			Peripheral artery disease (adjudicated)

ACR, albumin-to-creatinine ratio; BUN, blood urea nitrogen; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; FGF, fibroblast growth factor; HDL, high-density lipoprotein; iNET-CKD, International Network of Chronic Kidney Disease Cohort Studies; LDL, low-density lipoprotein; PCR, protein-to-creatinine ratio; PTH, parathyroid hormone; TSAT, transferrin saturation.

^aEither centrally measured by protocol (80%–40% of measurements, depending on the biomarker) or collected, when available, from medical records.

^bIncludes dialysis, kidney transplantation, or an eGFR < 15 ml/min per 1.73 m².

^cIncludes adjudicated and nonadjudicated events.

^dHealth, quality of life, or functional status assessed through self-administered questionnaires.

Estimated GFR

GFR is a key variable for population selection and outcome ascertainment in CKD studies. Furthermore, the discovery of new risk factors and biomarkers of CKD progression requires controlling for baseline GFR, which may vary among study participants. Given the burden of GFR measurement, most cohorts have relied on GFR estimating equations associating laboratory values and demographics. However, the use of estimated GFR (eGFR) increases random error, which may contribute to misclassification (potentially differential, if categories are used) and residual confounding.^34,35 eGFR may also be biased, with the direction and magnitude of this bias depending on participant characteristics.^36,37 Efforts to achieve more accurate GFR estimation are ongoing. For example, the interest of including cystatin C, in addition to creatinine, to obtain reliable eGFR estimations without consideration of race or sex has been recently put forward by the CKD Epidemiology Collaboration and the European Kidney Function Consortium.^36,38 More work is needed to reach consensus about when to use each of these equations.³⁹

We recommend, therefore, that studies collect laboratory values of GFR markers to allow the use of different GFR estimating equations, including those that may be developed in the future. Whether creatinine, cystatin C, or both are considered, the assays should be traceable to the international standard reference materials. Repeatedly collecting data can help mitigate eGFR measurement errors when studying eGFR-based outcomes, such as slope, absolute, or relative eGFR decline, using linear mixed models. This class of model can further accommodate unequal number and spacing of measurements, and deal with noninformative censoring (data missing at random or completely at random).⁴⁰ Completing cohort information with data generated in the routine health care context can be tempting; however, some caution is needed with regard to the introduction of selection-, information-, or residual confounding–related biases.^41,42

Proteinuria

Proteinuria, in particular, albuminuria, is a major predictor of adverse outcomes, and under certain circumstances, of treatment response in CKD.^43,44 Yet, it is often not measured in routine care or research.⁴⁵ Among CKD cohorts, even after combining information on albuminuria, proteinuria, and urine dipstick testing, missing data for proteinuria can reach up to 60%.^33,43 Obtaining an indicator of albuminuria should be a priority in CKD cohort studies; albumin-to-creatinine ratio is most accurate and reliable when assays are standardized, but protein-to-creatinine ratio and dipstick can be an acceptable alternatives in financially constrained settings.⁴⁶

Cause of CKD

The 2024 Kidney Disease: Improving Global Outcomes guidelines on CKD evaluation and management upholds the CGA system (cause, GFR, and albuminuria categories) to define and classify CKD with respect to its severity and risk.⁴ The importance of this is emphasized by studies reporting how outcomes vary according to the primary renal diagnosis.⁴⁷ However, it may be difficult to identify the cause in many cases. In addition, multiple causes may be identified in the same study participant; up to 3 diagnoses were reported in the CKD-REIN study, with only 24% of them being biopsy-proven (ranging from 73% for glomerular nephropathy to 9% for diabetic nephropathy).²⁸ Furthermore, in the absence of a standardized system to group diverse CKD etiologies, it can be difficult to define how to implement the collection of this information in a cohort study. In contrast, CKD research has largely relied on the principle of disease progression mechanisms that are common to all etiologies.⁴⁸ The study questions should therefore drive the collection of information on primary kidney disease. If the etiology of CKD is a major consideration, every effort should be made to obtain the most reliable diagnosis, preferably based on a kidney biopsy.

PROMs

PROMs have been increasingly recognized as crucial for supporting patient-centered care and informing health policy decisions, such as drug approval and reimbursement.^49,50 They refer to validated, self-administered instruments that tackle concepts from the patient perspective, including symptoms, functioning, well-being, health-related quality of life, and psychological issues. Despite the high priority given to PROMs by patients and caregivers, they are rarely assessed in the setting of CKD clinical trials.³⁰ Prospective cohort studies remain our primary source of understanding of PROMs in patients with CKD.⁵¹

Choosing the PROMs for a cohort study, however, may prove to be challenging. Besides considering the study objectives and the psychometric properties of the instruments (validity, reliability, responsiveness), investigators may want to ensure comparability with previously published data. In a systematic review including 449 articles that assessed health-related quality of life and/or symptom burden in CKD (all stages), 4 different health-related quality of life instruments were identified, the most frequent being the 36-Item Short Form Health Survey (227/361 studies) and the Kidney Disease Quality of Life (100/361 studies).⁵¹ Regarding symptoms, heterogeneity was even more pronounced: 54 PROMs were used to measure a total of 68 symptoms, with relatively limited coverage of specific CKD-related symptoms and poor overlap across instruments. Instrument-related factors (length, complexity, and lack of sensitivity) and participant-level barriers (poor health literacy, language, or cultural differences) may represent a significant burden in PROM assessment.^52,53 Providing clear instructions before instrument implementation and offering feedback about research results to participants are key to enhancing response, as are regular reminders to respond.

Biosample Collection and Storage

Because cohort studies are, by necessity, many years in duration, at study initiation, it is not always known what types of data elements will be important during the study. Sample collection (and storage under proper conditions) provides some protection against this uncertainty. The sample collection strategy must, therefore, align with study objectives, ensuring that enough material (e.g., plasma, serum, urine, cells, tissue) is collected to meet current and future research needs. In contrast, excessive biosample collection can prove wasteful as long-term storage is costly. It is essential to develop standard operating procedures that include the shortest possible time between sample collection and freezing (ideally < 2 hours). Cataloguing samples using standardized identifiers (e.g., barcodes) reduces tracking errors. Storage conditions must account for the type of biospecimen (e.g., serum, plasma, urine), and methods should be chosen to preserve its integrity during long-term storage. For example, repeated freeze-thaw cycles can degrade biomarkers and introduce variability and aliquoting samples into smaller volumes at the time of collection minimizes risk. In addition, storage facilities must have backup power and disaster recovery plans to prevent the loss of samples.⁵⁴

External validation is a key aspect of biomarker discovery. Within the iNET-CKD, 19 out of 21 cohort studies that collected biosamples would be willing to conduct central (n = 14) or local (n = 17) additional analysis on stored material (Figure 2).

Figure 2.

Frequency of biosample collection in cohorts participating in the International Network of Chronic Kidney Disease cohort studies (iNET-CKD), by type.

Implementation and Logistics

Cohort Study Management and Leadership

Effective management and leadership are critical to the success of cohort studies. Leadership should involve a multidisciplinary team with clear roles and responsibilities, including scientific oversight, project management, data governance, and communication. A well-structured framework should ensure regular and transparent reporting to stakeholders, fostering trust and accountability. Leadership structures should include contingency planning to mitigate risks such as staff turnover, funding delays, or unforeseen operational challenges. Establishing a governance committee ensures accountability and facilitates decision-making for major milestones, such as protocol amendments or data access approvals.⁵⁵

A critical aspect of effective leadership is meaningful stakeholder involvement. This includes involving representatives from funding bodies, regulatory agencies, academic institutions, and patient advocacy groups throughout the study lifecycle. Engaging stakeholders early in the planning phase enables alignment on priorities, ensuring that the study design addresses both scientific and real-world needs. Regular updates and inclusive discussions foster collaboration and maintain stakeholder engagement.⁵⁶

Patient representation is particularly vital to the success and relevance of cohort studies. Including patients or their advocates as active contributors to study design and governance ensures that the research aligns with their needs and perspectives. Patient input can help prioritize research questions, improve recruitment strategies, and enhance the interpretation and dissemination of findings in ways that are directly applicable to those affected by CKD. Initiatives such as patient advisory boards or codesign workshops can provide structured mechanisms for this involvement. By embedding stakeholder engagement and patient representation into the leadership framework, cohort studies can achieve higher levels of transparency, inclusivity, and impact, ultimately improving their ability to translate findings into meaningful clinical and policy changes.⁵⁷

Expanding the Scope of Cohort Studies Through Data Linkage and Real-World Evidence Integration

One of the key strengths of cohort studies is their ability to link primary data to external datasets, expanding the scope of analyses. External datasets, such as hospital records, dialysis and mortality registries, and genomic databases, provide additional insights into comorbidities, long-term outcomes, including survival, and molecular mechanisms. Effective data linkage requires robust technical and legal frameworks, including accurate record-matching algorithms (e.g., probabilistic or deterministic matching) and compliance with privacy laws such as the General Data Protection Regulation in Europe and Health Insurance Portability and Accountability Act in the US Legal agreements, such as data use agreements and material transfer agreements, combined with data deidentification and secure environments, safeguard participant confidentiality and ensure compliance.⁵⁸

Although real-world evidence (RWE) complements cohort studies by enabling medico-economic assessments and extending findings to populations that are less likely to commit to long-term follow-up, cohort studies remain distinct because of their structured and prospective design. Cohort studies collect high-quality, tailored data to address predefined research questions, enabling stronger causal inference than RWE, which often relies on retrospective data with inherent variability. Furthermore, the depth of data available in cohort studies, such as biomarkers, genetic information, and patient-reported outcomes offers unique insights not typically captured in RWE sources. By linking cohort studies with RWE datasets, researchers can combine the rigorous, detailed data of cohorts with the broader, real-world context of RWE, enhancing generalizability while preserving data quality.

Regulatory and Ethical Considerations

Cohort studies, particularly those involving multiple centers or international collaboration, need to navigate complex regulatory and ethical issues, summarized in Figure 3. By addressing these regulatory and ethical challenges, cohort studies can maintain high standards of integrity and reliability, enabling meaningful contributions to clinical practice and public health policy.

Figure 3.

Typical regulatory and ethical aspects aimed to promote integrity and reliability in cohort studies conduct.

Funding – Initial and Sustaining

Obtaining sufficient funding for cohort studies is a significant challenge. It may be relatively easy to secure funding to establish a cohort, but funding for long-term follow-up, data management, and biosample storage is often more difficult to secure. Establishing a diversified funding portfolio reduces the risk of financial shortfalls. Initial funding often comes from grants, while long-term sustainability may require partnerships with industry or public-private collaborations. A clear policy on how industry partners are to be involved, particularly in access to data or publication rights, must be established to maintain scientific independence and transparency. Seeking philanthropic support or patient advocacy funding can provide supplementary financial resources while increasing stakeholder engagement.^59,60

Collaboration for Multicenter Studies

Collaboration in cohort studies offers significant benefits, including increased generalizability and statistical power.61, 62, 63, 64, 65 By combining data from multiple centers, researchers can analyze larger and more diverse populations, yielding findings that reflect real-world conditions more accurately. In addition, collaboration facilitates the sharing of specialized resources, such as bioinformatics pipelines or assay platforms, which may not be available at all sites. Resource-sharing fosters skills development and knowledge exchange, enhancing the overall quality and efficiency of the research.⁶³ The iNET-CKD, along with other initiatives (e.g., The CKD Prognosis, CKD Gen, as well as the CKD Biomarkers consortia, the Dialysis Outcomes and Practice Patterns Study), are the living proof of how collaborating studies can push forward the boundaries of what we know about CKD. To date, specific contributions of iNET-CKD have been focused on heterogeneity in blood pressure control,³² hemoglobin levels,³³ and kidney and cardiovascular outcomes across CKD cohorts around the world.⁶⁶ These results suggest that the “one size fits all” approach to CKD management is unlikely to be effective, and encourage new research efforts to develop a more tailored CKD care.

Despite the advantages, collaboration presents challenges. Harmonizing protocols across sites can be difficult because of differences in local infrastructure and resources. Developing a core protocol with optional elements for site-specific adaptation can address disparities while maintaining consistency in key aspects of data collection. Managing multicenter data requires careful coordination; centralizing data processing with built-in quality assurance checks reduces variability across sites. Alternatively, distributed data models enable local data storage while allowing centralized analysis.⁶⁴ Data sharing is a critical aspect of collaboration that necessitates establishing clear agreements on intellectual property and publication rights to maintain trust among collaborators. Using repositories with tiered access levels ensures compliance with privacy regulations while facilitating secure data sharing. Finally, interlaboratory variability in assays can introduce bias, which can be mitigated by using reference laboratories or calibrating assays against common standard statistical techniques, such as interrater reliability analysis, further help adjust for assay differences and ensure consistency across sites.⁶⁵

Dissemination and Impact

Findings from cohort studies play a critical role in translating research into practice by generating new hypotheses and filling gaps left by RCTs. Cohort studies yield valuable evidence on long-term outcomes and real-world applicability, providing important context for clinical guidelines. This ensures that research findings align with current practices and improve patient care. Beyond clinical guidelines, evidence from cohort studies has important policy implications. These studies can highlight disparities in care, inform cost-effectiveness analyses, and guide the allocation of healthcare resources. Policymakers often value data from large, diverse cohorts because it helps address equity and access challenges.

Successful dissemination of findings from cohort studies requires timely and strategic communication. This typically occurs through peer-reviewed publication. Delays in publishing results, however, can reduce their relevance; therefore, mechanisms such as preprints and open-access publications are essential for bridging the gap between data generation and implementation. Engaging with diverse stakeholders, including clinicians, policymakers, and patient advocacy groups, ensures that research findings are effectively communicated and applied. Tailored dissemination strategies, such as lay summaries or policy briefs, maximize the impact of these findings. Use of social media accelerates research dissemination while broadening its audience and feeding user-generated content.^67,68 Furthermore, assessing the long-term influence of cohort study findings on clinical practice and policy provides valuable feedback for future research efforts, enhancing the overall impact and utility of these studies.

New Perspectives in Cohort Studies

Technological advancements in computational power and data acquisition strategies have shaped evolutions in cohort study design by enabling continuous and more granular data collection while facilitating data collection, integration, and privacy protection.69, 70, 71 Beyond these technological foundations, contemporary cohort studies have increasingly emphasized stakeholder engagement and interdisciplinary collaboration to enhance patient centeredness, efficient translational research and comprehensive evidence generation that can easily be incorporated into practice guidelines.⁷¹ Cohort studies are increasingly aligned with policy development, generating information that informs policy decisions. In addition, they have incorporated transparency features, such as providing individuals with access to their own data,⁷² and serving as infrastructure for Trials within Cohorts designs.⁷³

From a collaborative standpoint, recent efforts in data harmonization across existing cohorts have relied on the Observational Medical Outcomes Partnership common data model.⁷⁴ One of the most interesting features of the Observational Medical Outcomes Partnership common data model is its association with a concept vocabulary that compiles numerous common standards (e.g., ICD-9, MedRA codes) and with open-source software tools for data extraction, transformation, loading, and subsequent analysis. Mapping concepts to standard definitions has, however, been particularly challenging in prospective cohort studies because of data heterogeneity and complexity. Practical examples of innovative solutions include the use of text mining augmented by language detection in international cohorts,⁷⁵ and natural language processing allied to machine learning for harmonization based on metadata. Collaborative analyses have increasingly benefited from federated approaches, more privacy-compliant than approaches based on individual or aggregated data sharing.^76,77

Summary and Conclusion

Well-designed cohort studies are essential complements to randomized trials, offering unique insights into the long-term natural history, heterogeneity, and patient-centered outcomes of CKD. Drawing from the collective experience of iNET-CKD, this article outlines best practices in design, implementation, and governance that can help new and existing cohorts maximize scientific value while ensuring inclusivity, sustainability, and impact. Future efforts should prioritize the standardization of data collection, patient and stakeholder engagement, and the strategic integration of real-world data sources. To advance global kidney health, we call on the international nephrology community, through networks such as iNET-CKD and organizations such as ISN, to champion the development of harmonized frameworks for cohort studies. These should include consensus on core data elements, outcome definitions, and governance principles to promote interoperability, equity, and data sharing. By fostering collaboration around shared standards, we can accelerate progress toward more precise, patient-centered, and policy-relevant research in CKD.

Appendix

List the International Network of Chronic Kidney Disease Cohort Studies (iNET-CKD)

4C: Franz Schaefer; BIS: Elke Schaeffner and Natalie Ebert; CanPREDDICT: Adeera Levin and Mark Canney; CKD-JAC: Takahiro Imaizumi; CKDopps Brazil: Viviane Calice-Silva; CKDopps Germany: Helmut Reichel; CKDopps-US: Roberto Pecoits-Filho; CKD-REIN: Natalia Alencar de Pinho; CKiD: Susan Furth; CORE-CKD: Chagriya Kitiyakara; CRIC: Laura M. Dember and Krista Whitehead; C-STRIDE: Luxia Zhang and Jinwei Wang; EQUAL: Nick Chesnaye and Vianda Stel; GCKD: Peggy Sekula; ICKD: Vivekanand Jha; KNOW-CKD: Kook-Hwan Oh; KNOW-pedCKD: Yo Han Ahn and Hee Gyung (Aurea) Kang; MMKD: Florian Kronenberg; NRHP-URU: Laura Sola and José Boggia; NURTuRE-CKD and RRID: Maarten Taal; PECERA: Jose Luis Górriz; PROGRESER: Alberto Martinez-Castelao; PROVALID: Gert Mayer and Susanne Eder; PSI BIND-NL: Martin de Borst; SKS: Philip Kalra and Rajkumar Chinnadurai.

Disclosure

The authors have received research support from Agencia Nacional de Investigación e Innovación, the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, and the Patient-Centered Outcomes Research Institute. Additional support related to the CKD-REIN cohort study has been received from the “Cohortes Investissements d’Avenir” program (ANR-IA COH-2012/3731), the French Ministry of Higher Education and Research, Boehringer Ingelheim France, Fresenius Medical Care, GlaxoSmithKline, Novo Nordisk, and Vifor France. The authors have received consulting fees, research funding, or honoraria from Alucent Biomedical, Alpine, AstraZeneca, Bayer, Baxter, Biogen, Boehringer Ingelheim, Chinook, CSL-Behring, Eli Lilly, Kyowa Kirin Company, Merck, Nipro, Novartis, Otsuka, Timberlyne, Vantive Brasil, Vera, Vertex, and Visterra. The authors have participated in advisory boards or leadership roles for the American Society of Nephrology, the Brazilian Society of Nephrology, the ISN, the Latin American Dialysis and Transplantation Registry and the SharE-RR Advisory Group (ISN). All other authors declare no competing interests.

Supplementary Material

Supplementary File (PDF)

Table S1. Cohort-level responses to a data availability questionnaire conducted as part of the ISN’s International Network of CKD Cohorts (iNET-CKD) activities between March 3 and July 21 2024.