Slide 1

I will prepare the field for Kitty Jager and Friedo Dekker that will go in to more detail in some epidemiological themes. My scope today is to focus in general terms on this new document that has been produced for epidemiologists, for clinicians if they want to produce neat papers, papers correctly reporting the strengths and the quality of the observed association.

Slide 2

Now, let’s start with observational studies, what they are and their scope.

Slide 3

The scope of observational studies is to explore cause-effect associations the very old egg problem or to identify prognostic factors, which factors predict the storm? Or to verify the magnitude of previously reported association. The association between smoking and heart disease has a hazard ratio of two. That of cancer with smoking is ten times higher, this should be confirmed several times. There are of course other scopes.

Slide 4

Importantly, observational Studies are the most frequent study type in Nephrology  Journal (>80%).

Slide 5

Observational studies can be started for several reasons..

Slide 6

They may be started  on the basis of clinical observations. For example nephrologists may note that xerostomia seems to be frequent in dialysis patients.

Slide 7

And they may therefore start a case-control study to confirm that this is indeed  the case.

Slide 8

In this particular study systematic testing of salivary and lacrimal secretion revealed an increased risk for xerostomia and  xeroftlamia in ESRD.

Slide 9

They may be started  because of biological insight.

Slide 10

Fetuin is a calcification inhibitor and in transgenic rats with double deletion of the fetuin gene  organ system and CV calcification is pervasive.

Slide 11

On the basis of this knowledge investigators by Aachen set up a cohort study  that demostrated that survival patients with low fetuin have an increased risk of death which is line with experimental models.

Slide 12

Finally observational studies may arise from informal look at available data. The investigator may note something special in the data he or she has… and performs further analyses to generate a new hypothesis

Slide 13

The clinical trial is the unbeatable standard for assessing the value of treatment  but even here  some epidemiologists argue that observational studies  can be useful to measure the effect of interventions.

Slide 14

Indeed the effect size...

Slide 15

...measured in a series of selected clinical trials...

Slide 16

...almost coincided with the estimate made in well conceived observational studies looking at the same problem.

Slide 17

But before proceeding further let’s me contextualize the area of application  observational studies. Clinical epidemiology research explores etiology of diseases, evaluates  diagnostic and prognostic tests  and assesses the efficacy of treatments. Etiology and prognosis is the typical area were observational studies  are applied.

Slide 18

We do clinical research  because  we hope to be useful to our patients ….we hope that our work serves to advance scientific knowledge and hence - directly or indirectly - lead to improvements in prevention and treatment of disease......

Slide 19

These are noble scopes but there is a background complaint that notwithstanding the advamcements of modern sciences, the quality of medica research, clinical research in particular, is poor. Altman goes into saying that we should face a scandal and puts the question squarely: what should we think about researchers ...

Slide 20

Clinical trials have been a weak research area for a long time.

Slide 21

In a survey made by Altman about 10 years ago, a systematic inquiry on fundamental methodological items from study eligibilibity to subgroups comparison, from a quarter to about 90% of trials hads pecific  problems with these items. 

Slide 22

To react to this situation methodologists founded CONSORT, an initiative aimed at alleviating the problems related with poor reporting of clinical trials.

Slide 23

And a practical spin-off of this initiative was the production of a checklist that trialists should consider when reporting a clinical trial.

Slide 24

The same problem exists with diagnostic studies.....

Slide 25

the methodological quality of diagnostic studies is mediocre at best....

Slide 26

and gain the response to this problem was the STARD initiative, that is an initiative providing standards for reporting of diagnostic accuracy.

Slide 27

In 2004 Pockock in the pages of the BMJ exposed the problem of poor reporting in epidemiological studies…..

Slide 28

This survey raises concerns regarding inadequacies in the analysis and reporting of epidemiological publications in mainstream journals…

Slide 29

This problem was well known to nephrologists  because two papers published in AJKD one in 1996 on the fallacies in the use of Cox regression analysis

Slide 30

and another on problems related  with prognostic studies  showed that the quality of reporting was suboptimal.

Slide 31

In the last two years  a group  of distinguished epidemiologists and methodologists convened several times  to produce guidance aimed at improving reporting of epidemiological studies and the result of this effort  is strobe, i.e. strengthening reporting  of observational studies in epidemiology.

Slide 32

The manuscript was published in a long and in a short format on several major journals including the lancet and annals of internal medicine

Slide 33

and  was accompanined by a checklist that included 22 recommendations.

Slide 34

As you know  we have three types of observational studies, cohort studies,

Slide 35

case-control studies

Slide 36

and cross-sectional studies.

Slide 37

For reason of relevance and time I  will show some recommendations related  with cohort studies...

Slide 38

Describe the setting, location, relevant dates including periods of recruitment, exposure, follow up and data collection, participants, appropriate description of the cohort, definition of quantitative variables, statistical methods and degree of adjustment. These three are related to confounding. These are important for controlling for confounding. Finally, recommendation about the description of limitation of these studies.

Slide 39

Now let’s start with an example related with the setting.

Slide 40

This is a paper that was published in Kidney International last year. It describes risk factors for CKD in a community based population. A ten year follow up. This is a Japanese study.

Slide 41

You see the authors correctly specified that participants were seen in the Annual Health Examination held in the Ibaraki prefecture between 1993 and 2003. This is let’s say a sort of Japanese equivalent of NHANES that is the periodical survey the Americans made but in a randomised way. So it’s adamant that describing the setting is important because Japan is different from ---, Japan is not the States.

Slide 42

So setting and location is important to assess the context of the study, when and where the study took place and over what period.

Slide 43

Second point,

Slide 44

description of the cohort, eligibility, criteria, source and methods of selection of participants.

Slide 45

You see this study aimed at assessing the prognostic value of the New York Health Association classification in ESRD.

Slide 46

The authors specified that this study was based on a registry. The registry started collecting data from 1994 onwards and here is specified the criteria that was followed for considering patients eligible to enter the registry. Finally, it is specified that the study was based on incident patients rather than on prevalent patients.

Slide 47

So a detailed description of the study participants helps the reader in understanding the applicability of the results. Here the criteria were irreversible ESRD and this was a study based on registry data and on incident patients.

Slide 48

Slide 49

Now these 3 points, as I anticipated, are related with confounding.

Slide 50

Let me remind you what confounding is. Confounding means confusion of effects.

Slide 51

Smoking and heart disease. Smoking is a cause of heart disease. Poverty can be a confounder because poverty facilitates smoking. People who are poor frequently smoke. But poverty is also a cluster of factors that may lead to heart disease, so poverty is a confounder for the interpretation of the effect of smoking on heart disease. Please don’t forget that confounding should not be an intermediate factor in between the exposure and the outcome.

Slide 52

Now, in observational studies we use multivariable analysis to deal with confounding in the analytical phase. You see this explains this recommendation on quantitative variables and adjustment. The recommendation is please explain how quantitative variables were handled in the analysis and if applicable, describe which groupings you have chosen.

Slide 53

Now, confounding poverty. We may use income as a measure of poverty. We may use income as a continuous variable from 0 to 100 million euros.

Slide 54

If we adjust the analysis for income, especially as a continuous variable, we may eliminate the confounding effect of poverty. But if we use income in categorical terms, in binary terms, that is people are poor when they have an annual income less than 30 million euros and not poor if they have a higher income,

Slide 55

You see we may have residual confounding because this annual income or this annual income or this annual income and this annual income this cannot remove the confounding,

Slide 56

so the confounding remains there.

Slide 57

So categorising variables increases the likelihood of living their residual confounding.

Slide 58

Now the adjustment. Statistical adjustment. The recommendation is to give unadjusted and adjusted estimates.

Slide 59

An example. We know that inflammation causes atherosclerosis and Chlamydia Pn has been suspected as an agent responsible for inflammation and atherosclerosis.

Slide 60

But the effect of Chlamydia maybe confounded by age and smoking because Chlamydia is frequently found in smokers and old people.

Slide 61

For the authors of these papers reported the association between Chlamydia measured as IgA titre in unadjusted and adjusted form. In the unadjusted analyses you see the higher the titre, the higher the risk of all cause death. But when the analysis was adjusted for age and smoking, the association collapsed meaning, implying that these were confounders for the interpretation of the effect of Chlamydia on all cause mortality.

Slide 62

And finally limitation.

Slide 63

Investigators should face the issue of study limitation.

Slide 64

In this study, for example, you see the authors emphasised the limitation of their study, of their analysis because of the mortality and censoring approximately 25% of patients could not repeat in this case it was an echocardiographic study. So the cohort that entered the follow up study that aimed to establish the prognostic value of serial echocardiographic studies had a lower risk than the original cohort. This is a limitation which is frankly faced and recognised.

Slide 65

This is emphasised that the identification and discussion of the limitation of a study are an essential part of scientific reporting. We should describe and discuss the sources of bias and confounding and also discuss the relative importance of different types of bias.

Slide 66

So, I can now summarise observational studies are the most frequent study type in nephrology journals over 80%. In nephrology like in other specialities the quality of observational studies is suboptimal. Transparent and complete reporting of observational studies is important to provide fair information to readers. This STROBE statement provides helpful recommendations for reporting observational studies and if you are planning to submit a paper to the American Journal of Kidney Disease please take into account that this journal already asks that you apply all items that are described in the STROBE statement. I suspect, I am sure that also all the nephrology journals will adopt the STROBE statement very soon. Thank you for your attention.