Levels of Renal Function: Epidemiological Comparison

Introduction

The field of epidemiology mainly concerns itself with the incidence of diseases within certain populations, morbidity as well as other factors that involve or determine health (Coggon et al, 2003, pp.1608-1615; Fletcher et al, 1988), and epidemiological data is used to develop plans that strategize how best to prevent illnesses while guiding patients whose diseases have already progressed (Coggon et al, 1997; Moon et al, 2000; Stewart, 2000). Study designs may vary with factors for determining one’s choice such as cost, time and the particular question of interest (Moon et al, 2010:9). Epidemiological research has played an essential role in furthering the general understanding of certain medical issues, and it is no exception in this assignment, where two epidemiological studies that assess the eGFR’s association with the timing of dialysis, progression of kidney disease and mortality in elderly patients will be critically appraised and compared. In detail, it will compare a cohort study (Stel et al, 2010) that determines whether age, gender, primary renal disease etc. affects the level of renal function at the start of dialysis, and what among these factors could be used as determiners of initiation of dialysis in patients with a GFR of 4-10 mL/min, with a randomized controlled trial (Cooper et al, 2010) that determines whether the timing of the initiation of dialysis could affect the improvement, survival, maintenance, quality of life and mortality of patients 18 years old and above with progressive chronic kidney disease. It will not aim to answer the issues posed by both studies, or to clarify where medical research currently stands on the decision-making process in starting dialysis, but rather to demonstrate the depth of understanding this author has with regards to the different types of epidemiological studies, as well as this author’s capacity to compare and analyze them.

Focus of Assignment: Comparison and critical appraisal of two epidemiological studies

In 2010, a group of researchers led by Dr. Bruce Cooper set out to fill the need for data from randomized, controlled clinical trials. They researched on the prevalent practice of early dialysis initiation and the studies regarding the timing of dialysis commencement, saying that these could possibly be limited due to having biases with patient selection, lead time and referral time (2); but more importantly, they focused on the effects of early vs. late initiation of maintenance dialysis.

In the same year, Dr. Vinda Stel et al reviewed the guidelines set up by the National Kidney Foundation – Dialysis Outcome Quality Initiative (NKF-DOQI) on the initiation of dialysis based on estimated glomerular filtration rates (GFR) and renal function levels.

Both cohort study (Stel et al, 2010) and randomized controlled trial (Cooper et al, 2010) were defined by Beaglehole et al’s categorizations of quantitative study designs (1993), and will be appraised further using Elwood’s schema as the primary mode of epidemiological analysis.

Description of the evidence

According to Moon et al (2010), both cohort and randomized control studies are analytical and observational in their design: analytical because they both try to determine the associations of the medical condition and the potential causes (Moon et al, 2010), and observational because the researcher is limited to observing the events as they happen (Fletcher et al, 1988).

Cohort studies are defined by a large population sample, and could either be historical or prospective in design (Coggon et al, 1997); they answer the question ‘what should happen next (Crombie, 1996)?’ In contrast, randomized control studies are mostly retrospective and smaller in size (Fletcher et al, 1988), while the selection process is based on a randomized process.

The randomized control trial measured the association between the timing of initiation of dialysis and recovery, maintenance or mortality of the patient from ESRF, while the cohort study measured the association between the level of renal function and baseline characteristics of patients. Cooper et al’s study looked at association and effect at only one point in time, while Stel et al was longitudinal looking at information from more than a single point in time. Both, however, looked at retrospective data and the incidence of chronic kidney disease. Cooper et al’s study concluded that there was no difference between early and late initiation of dialysis, for the level of renal function is not affected by the timing of initiation; Stel et al’s is basically inconclusive in that they were unable to pinpoint a single definite cause for the varying levels of eGFR, although country seems to be the leading cause, and yet even this is inconclusive since this could be caused by the differences of measuring creatinine levels in each country.

Level of Renal Function in Patients Starting Dialysis: an ERA-EDTA Registry Study by V.S. Stel, C Tomson, D Ansell, F.G. Casino, F Collart, P Finne, G.A. Ioannidis, J De Meester, M Salomone, J.P. Traynor, O Zurriaga and K.J. Jager, published by Oxford University Press on behalf of ERA-EDTA, 2010.

The setting for the study for this study was 8 different national or regional registries. Population includes 11, 472 patients who started dialysis in 1999 and 2003. This study’s aim was to determine which baseline characteristic affected levels of eGFR at the start of dialysis. The time scale of the study is 1999 and 2003; it was designed to be a retrospective cohort registry study.

Results

In this study, it was established that older patients, males, patients with diabetes mellitus, hypertension or renal vascular disease as primary renal disease, ischaemic heart disease and patients starting on peritoneal dialysis had higher levels of eGFR at the start of dialysis, as well as eGFR varied depending on the country.

A Randomized, Controlled Trial of Early versus Late Initiation of Dialysis by B Cooper, P Branley, L Bulfone, J.F. Collins, J.C. Craig, M.B. Fraenkel, A Harris, D.W. Johnson, J Kesselhut, J Li, G Luxton, A Pilmore, D.J. Tiller, D.C. Harris, and C.A. Pollock, published in The New England Journal of Medicine 2010; volume 363: no. 7.

Having been conducted in Australia and New Zealand, the setting for the study for this study was 32 different centers scattered across those two countries. Population includes patients with progressive chronic kidney disease and estimated GFR from 10-15 mL/min and 5-7 mL/min aged 18 and above – a total of 828 adults.

This study’s aim was to determine whether early initiation of dialysis could affect the condition of the patient when compared with patients who started dialysis at a later time. By extension, this study questions the validity of previous research on when to start dialysis. The predicted primary outcome is death. The time scale of the study is from July 2000 to November 2008 (100 months) and was designed to be a randomized, controlled clinical trial.

Results

In Cooper et al’s study, it is shown that the primary outcome was death from any varying cause except from uremia, and so it is proven that there the early or late timing of initiation of dialysis makes no difference in the outcome.

Internal Validity

According to Crombie (1996), validity concerns itself with the accuracy of data and how likely the study measures what it is supposed to measure, and internal validity is concerned with the accuracy of the data about those being studied (Fletcher et al, 1996), as well as the structure and methodology of the study (Burns and Grove 1993). According to Moon et al (2010), there are three main factors that influence internal validity: bias, confounding and chance.

Bias is defined as ‘a process at any stage of influence tending to produce results that depart systematically from the true values (Murphy, 1976).’ It mostly occurs during subject selection and data collection (Crombie, 1996; Moon et al, 2010) due to faults in study design (Coggon et al, 1997; Stewart, 2002). Bias can possibly occur in any kind of study design and is basically anything that gives preference or an unfair disadvantage to one side, whether accidental or deliberate, known or unknown (Follman and Schron, 2001).

Selection bias in the control trial is minimized by the fact that subjects were randomly selected; however, the fact that there were patients ineligible for selection suggests selection bias and could potentially mean that the findings could not be applicable to the complete populace. Cooper et al (2010) describe complete follow up for more than 95% of the cohort, and therefore only a small proportion was lost to follow up, minimizing form and observational bias that could affect the results. It also limited the number of clinicians involved and making sure all involved groups had restricted, limited knowledge of the whole study, reducing information bias.

Despite the fact that the cohort study was retrospective, selection bias is minimal for Stel et al’s study because the data was taken from renal registries. Given the size of the study’s population, information bias is also reduced.

Neither study acquired information solely through the patients or their relatives, so recall did not pose any challenges. Both studies seemingly used standard, institutionalized forms for data collection relative to their fields and environments.

According to Cogon et al (2003), non-randomised sampling of subjects further adds to the issue of confounding, therefore it can hardly occur with Cooper et al’s trial due to its randomization, but Stel et al’s study admitted that there is in fact room for confounding to occur since only eGFR was measured as basis for initiation of timing of dialysis and not other factors that could very well affect the results.

As for chance, it is also minimal with the randomized control trial because of the low P value (less than 0.05), giving it statistical significance as previously mentioned; however, the P values were not shown in the cohort study, making it seem susceptible to chance. Data was not given for both studies to indicate confidence intervals, which proved to be a disadvantage because such intervals provide a wider range of possible values for the extent of any difference, whereas the P value merely points out the statistical significance of the differences without indicating what they actually are. In studies such as these, it is always essential to have both values to reduce the probability of the results being affected by chance (Coggon et al 2003).

External Validity

External validity refers to the capacity of the studies to be generalized to the whole population involved or affected by the data (Burns and Grove, 1993). Both cohort and randomized control trial provided specific details such as age, sex, ethnicity, renal function level an other conditions, so the findings could only be applied to people who meet the same criteria, or is at least within its range, and has kidney disease. Given that these terms are common in elderly people, both studies are easy to replicate. Although a hierarchy of evidence exists when choosing appropriate study designs (Greenhalgh, 2001), there are many various advantages and disadvantages in all kinds. Appendix 1 depicts the main differences between the two studies.

As for the comparisons of the research to other studies, Elwood (1998) identifies four key points of comparison: consistency, specificity, plausibility and coherence. In the hierarchy of evidence (Greenhalgh, 2001), the two studies belong in different levels, making them difficult to compare with one another. Specificity comparisons are also difficult to make since no specific hypothesis was actually tested (Elwood, 1998). Results from both studies are plausible, but only when isolated to certain practices in the medical field.

Conclusion

According to Beaglehole et al (1993, pp.36-9), there are two kinds of studies: observational and experimental. Two sub classifications are under observational studies – the descriptive analysis and the analytical – and Cohort studies fall under the latter. These studies, also known as incidence studies, observe a population or a cohort that is free of disease through time to determine if exposure to certain risk factors affects the incidence of diseases (Gerstman, 1998, pp.146-8). From this definition, we understand that Stel et al’s work is observational by nature, and as such, prioritizes the acquisition of data regarding frequency, distribution and patterns of disease (Moon and Gould, 2000), which it adequately fulfills. Randomize control trials, on the other hand, are under experimental studies, and they are considered to be the most scientifically rigorous research, as well as the most troublesome due to the difficult of applying it to the larger population (Gerstman, 1998, pp.152-5). This is why the cohort study seems insufficient in some points in comparison to Cooper et al’s work, and why it seems to hold more validity than the former.

Both study designs were sufficient to probe the respective issues investigated. The cohort study involved over 11,000 patients – more than enough to qualify the research as representative of the population. Follow up period was also appropriately long for the randomized control trial, and it had an adequate population size as well. Nothing negative could truly be said for Cooper et al’s research since it provided sufficient data; however, the cohort study could’ve provided more data to verify its internal validity (ex: P values to rule out chance etc.).

Data collection was standardized in both studies as well, therefore minimizing the possibility of information bias at the point in time when data was being collated. Despite what has been previously stated regarding the consistency, thoroughness and validity of the randomized control trial, however, it shares the main weakness of the cohort study in that both used retrospective data and heavily relied on precise documentation, giving way to the possibility of information bias.

Although both studies can still be criticized further on a number of grounds, they still provide good examples of the various research designs and analytical approaches used in epidemiological studies.

As for the initiation of dialysis, the issue remains unclear since there is still no standardized measure as to the proper time to initiate renal replacement therapy (RRT) (Lameire et al, 2002, p.1), though a number of conclusions can be derived from past studies regarding maintenance hemodialysis – a field which has been introduced since the 1960’s (Scribner, 1960, p.114). It has always been the tradition to view the presence of uremia and its signs and symptoms, coupled with results of “biochemical measurements in serum and plasma” (Cooper et al, 2010, p.1), as indicators for initiating dialysis (Hakim et al, 1995, pp.1319-28); however, it has recently come to light that initiating dialysis at early stages may actually be harmful (Beddhu et al, 2003, pp.2305-12; Kazmi et al, 2005, pp.887-96; Lassalle et al, 2010, pp.700-7; Stel et al, 2009, pp.3175-82; Travnor et al, 2002, pp.2125-32), although there are still cohort and case-studies that suggest otherwise (Bonomini et al, 1986, pp.267-71; Tattersall et al, 1995, pp.283-9). Today there is still only very little scientific consensus regarding the timing of initiation of dialysis, as well as on which clinical parameters to base such decisions (Fenwick et al, 2004, p.1022); perhaps this is the case because, as Cooper had previously written, “data from randomized, controlled trials that establish the optimal timing for the initiation of dialysis are lacking (2010, p.2).”

Using and comparing the results acquired from the medical research of Dr. Hyung Kim et al (2009), Venturelli and Brunori (2010), NECOSAD (2002), Fenwick et al (2004) as well as the two epidemiological studies discussed in this assignment, the following conclusions can be drawn: Future studies regarding the decision of nephrologists as to which factor should determine the initiation of dialysis should take into consideration the “heterogeneous nature of the End Stage Renal Failure population” (Fenwick et al, 2004, p.1022) while still considering eGFR and disregarding time of initiation; it would also help if a standard measurement of creatinine would be institutionalized and implemented globally.

Reference List

Beaglehole R, Bonita R & Kjellstrom T. (1993). Basic Epidemiology. World Health Organization, Geneva.

Beddhu S, Samore MH, Roberts MS, et al. (2003). Impact of timing of initiation of dialysis on mortality. J Am Soc Nephrol, 2305-12.

Bonomini V, Feletti C, Stefoni S, Vangelista A. (1986). Early dialysis and renal transplantation. Nephron, 267-71.

Burns & Grove. (1993). The Practice of Nursing Research: Conduct, Critique & Utilization. W.B. Saunders co.

Coggon D, Rose G, Barker DJP. (1997). Epidemiology for the uninitiated. London: BMJ Publishing Group.

Coggon D, Harris EC, Pool J & Palmer, KT. (2003). Extended follow up of a cohort of British chemical workers expose to formaldehyde. J. National Cancer Institute, 1608-1615.

Cooper B, Branley P, Bulfone L, Collins JF, Craig JC, Fraenkel MB et al. (2010). A Randomized, Controlled Trial of Early versus Late Initiation of Dialysis. The New England Journal of Medicine, Volume 363: no. 7.

Crombie IK. (1996). The pocket guide to critical appraisal: a handbook for health care professionals. London: BMJ Publishing Group.

Elwood J. (1998). Critical Appraisal of Epidemiological Studies and Clinical Trials. Oxford University Press.

Fenwick S, Saxena R & Harper JM. (2004). Dialysis: when to start or when to stop? Nephrol Dial Transplant, 1022.

Fletcher R, Fletcher S & Wagner E. (1998). Clinical Epidemiology: The Essentials. Williams and Wilkins.

Follmann DA & Schron E. (2001). Essentials of Randomized Clinical Trials. Pacing and Clinical Electrophysiology, 254-259.

Gerstman B. (1998). Epidemiology Kept Simple: An Introduction to Modern Epidemiology. New York: Wiley-Liss.

Greenhalg T. (2001). How to read a paper: the basics of evidence based medicine. London: BMJ Publishing Group.

Hakim RM & Lazarus JM. (1995). Initiation of dialysis. J Am Soc Nephrol, 1319-28.

Kazmi WH, Gilbertson DT, Obrador GT, et al. (2005). Effect of comorbidity on the increased mortality associated with early initiation of dialysis. Am J Kidney Dis, 887-96.

Lameire N, Van Biesen W and Vanholder R. (2002). Initiation of dialysis—is the problem solved by NECOSAD? Nephrol Dial Transplant, 1550–1552.

Lassalle M, Labeeuw M, Frimat L, et al. (2010). Age and comorbidity may explain the paradoxical association of an early dialysis start with poor survival. Kidney Int, 700-7.

Moon G et al. (2010). Epidemiology: An Introduction. London, Milton Keynes: Open University Press.

Murphy EA. (1976). The logic of medicine. Baltimore: Johns Hopkins University Press.

Scribner BH, Buri R, Caner JEZ, Hegstrom R & Burnell JM. (1960). The treatment of chronic uremia by means of intermittent hemodialysis: a preliminary report. Trans Am Soc Artif Organs, 114–119.

Stel VS, Dekker FW, Ansell D, et al. (2009). Residual renal function at the start of dialysis and clinical outcomes. Nephrol Dial Transplant, 3175-82.

Stel VS, Tomson C, Ansell D, Casino FG, Collart F, Finne P, et al. (2010).Level of Renal Function in Patients Starting Dialysis: an ERA-EDTA Registry Study. Nephrol Dial Transplant, 3315-25.

Tattersall J, Greenwood R, Farrington K. (1995). Urea kinetics and when to commence dialysis. Am J Nephrol, 283-9.

Appendix 1

Comparisons between the two studies based on the Elwood Scheme