Periodontal Myths and Mysteries Series (VI) – How to Address Residual Confounding

What might lead to a Eureka! moment among our scientific and professional community should rather be seen as matter of serious concern. The two recent papers on teeth, or rather lack thereof, and possible relationship with cardiovascular events or even death have sparked the expected excitement. In one editorial of a Dutch heart journal, the editor-in-chief even trumpets in the title of his editorial, “Tooth brushing for a longer and healthier life.” He thus repeats, in other words, the notorious Floss or die! claim of 1996. I had forecast, some time ago, that relentlessly beating the drum by our thought leaders will ultimately lead to loss of any interest of the medical profession (in particular cardiologists) in the perio-systemic link, in particular as the American Heart Association had already concluded, in their comprehensive systematic review of 2012, that

[e]xtensive review of the literature indicates that PD [periodontal disease] is associated with ASVD [atherosclerotic vascular disease] independent of known confounders. This information comes mostly from observational studies, however, and therefore does not demonstrate that PD is a cause of ASVD, nor does it confirm the contention that therapeutic periodontal interventions prevent heart disease or stroke or modify the clinical course of ASVD. Although a contribution of PD to ASVD is biologically plausible, periodontal and cardiovascular diseases share multiple risk factors that are prevalent and powerful promoters of disease, including tobacco use, diabetes mellitus, and age. (Emphasis added.)

The Darapladib Trial

But let’s have a look at the two papers on lack of teeth and cardiovascular events. The one which has sparked so much enthusiasm (Vedin et al. 2016) is in fact a large randomized trial, STABILITY (STabilization of Atherosclerotic plaque By Initiation of darapLadIb TherapY), evaluating Darapladip, an oral inhibitor of lipoprotein-associated phospholipase A2 as compared to a placebo. A huge number of almost 16,000 participants had been randomized at 661 sites in 39 countries on five continents. According to inclusion criteria, all had so far (when randomized) stable chronic coronary heart disease and were 60 yr or older. The mean observation period in report by Vedin et al. (2016) was a mere 3.7 yr. Number of teeth was self-reported in a lifestyle questionnaire which included questions about education, psychosocial factors, general health, diet, alcohol consumption and physical activity.

Note that in November 2013, GlaxoSmithKline had announced that Darapladip had failed to meet Phase III endpoints in the trial  of patients with acute coronary syndrome. The respective paper indicating not significantly lower incidence of primary composite outcome in the test group can be found here. In an attempt to comprehend the lack of effect of Darapladip, authors discuss that

[i]t is possible that the coronary risk among patients in our study may already have been minimized by concurrent therapy. The trial was designed to test the incremental effect of a new treatment administered in patients who were receiving a high level of standard of care for secondary prevention at baseline. Thus, more than a third of the patients had an LDL cholesterol level of less than 70 mg per deciliter (1.81 mmol per liter) at baseline, and revascularization had been performed in 75% of the patients before randomization. High rates of the use of evidence-based medications were maintained throughout the trial. These standards of care are consistently higher than those that were observed in patients with stable chronic coronary heart disease who were included in previous large international registries.16-18 These factors probably reduced event rates in the two study groups and may have reduced the proportion of events that were modifiable.

Another consideration is that 96% of the patients at trial closeout were taking statins, which have been shown to reduce levels of lipoprotein-associated phospholipase A2 by up to 35%.19-21 In addition, in the Long-Term Intervention with Pravastatin in Ischemic Disease (LIPID) study, among patients with stable chronic coronary heart disease, more than half the treatment effect of pravastatin in reducing rates of death from coronary heart disease or myocardial infarction was estimated to be due to an association with a reduction in levels of lipoprotein-associated phospholipase A2.22 Incremental benefits of inhibiting lipoprotein-associated phospholipase A2 activity, if present, could be less in patients treated with statins. (Emphasis added.)

In the paper by Vedin et al. (2016) a secondary analysis of the data is therefore reported (in other words, it is not “the first study to prospectively assess the relationship between tooth loss and outcome in patients with [cardiovascular disease]”, as van der Wall has it almost enthusiastically, see below). Patients were not randomized according to number of teeth. In particular, it was not planned to assess number of teeth as potential risk factor for cardiovascular disease. Use of statins and possible association with tooth loss and periodontal disease is controversial. Smoking was simply recorded as current or former. Whether tooth loss was an effect modifier was not investigated. (Note that the first who actually looked at a possibly decreased [sic] of edentulism with cardiovascular events as compared to subjects with periodontitis in a representative longitudinal study in the US were Hujoel et al. in 2001 who couldn’t find a different risk as compared to that of periodontitis).


But data provided by Vedin et al. (2016) may speak for itself anyway. Fig. 1 (above) shows cumulative incidence rates of various outcomes as related to numbers of teeth present at the outset. There was a strong dose-effect relationship for major adverse cardio effects, cardiovascular deaths, stroke and even all-cause death with more teeth lost consistently related to higher incidence rates. Authors indicate that myocardial infarction was not related to self-reported number of teeth. Authors conclude that “mechanisms leading to tooth lost, most importantly PD [periodontal disease], may contribute to a worse prognosis in CHD [coronary heart disease], although causality cannot be established from our results.”

Well, authors’ interpretation on “mechanisms leading to tooth loss, most importantly periodontal disease,” may be too hasty. Patients in the STABILITY trial had been randomized in 39 countries from North and South America, Western and Eastern Europe, Asia and the Pacific Rim. But it is certainly not a representative sample. It is undeniable that tooth loss is largely dependent on access to dental care and therefore strongly related to socio-economic factorsHujoel et al. (2007) had pointed  to the fact that disparities in socio-economic status of edentulous subjects in the US had not changed for decades. Interestingly, one cannot find a respective note in the paper by Vedin et al. (2016) that the respective Cox proportional hazard models had been adjusted for socio-economic status. Hazard ratios for every increase in self-reported tooth loss level (26-32, 20-25, 15-19, 1-14 or no teeth) in the final model (III) were adjusted for (I) treatment, (II) age, systolic and diastolic blood pressure, body mass index low and high-density lipoprotein cholesterols, history of diabetes, prior myocardial infarction, smoking status, waist hip ratio, (III) estimated glomerular filtration rate, family history of coronary heart disease, alcohol consumption, years of education, level of physical activity and country income level (but the latter may be prone to ecological fallacy). It must be stressed that all point estimated dropped further in model III except for the model where stroke was the outcome. It might be speculated that further adjusting for socio-economic status would have attenuated associations with tooth number further.

The EPIC-Potsdam Study

Is it even plausible to relate edentulism to more-or-less established association between systemic inflammation caused by periodontal disease and cardiovascular events? Isn’t it rather so that it might be expected that lack of any teeth is related to lower levels of systemic inflammation due to periodontal disease?

In contrast to Vedin et al. (2016), in a 13-yr, properly designated retrospective, study of a cohort of more than 24,000 middle aged subjects between 35 and 64 yr  (not necessarily representative of the Eastern German population), Oluwagbemigun et al. (2015) have recently found an inverse relationship between lack of teeth (or  18-23 teeth) as compared to full dentate subjects, 28 and more, and risk for myocardial infarction when models were extensively adjusted for possible confounders including socio-economic and lifestyle factors. The picture as regards stroke was even more complicated as, after extensive adjustment, associations with number of teeth became insignificant. Additional analyses indicated a non-linear relationship, though. There was no association between number of teeth and diabetes or cancer.

In last month’s further analysis of two subsamples of participants of EPIC-Potsdam (about 18 years apart), Oluwagbemigun et al. (2016) showed that the relationship of lower numbers or lack of teeth with myocardial infarction was not due to systemic inflammation as indicated by, say, high sensitivity C-recative protein (hsCRP) levels in serum. They conclude that “number of teeth has a weak association with hsCRP, if any, thereby excluding this marker of low-grade systemic inflammation as a possible explanation for the association between number of teeth and myocardial infarction.” As they emphasize in the Introduction of their paper, the association between reduced number of teeth and increased hsCRP, found in some studies, may vary across sexes, body mass index categories, socioeconomic status, or race. Other studies had not revealed an association with number of teeth. Likewise (and rather confusing), hsCRP was significantly higher in edentulous subjects in one study but not different from dentate individuals in another. They also stress that risk factors for tooth loss are multifactorial and complex, including alcohol consumption and flavonoid-rich foods (I would intuitively add also bad dentistry). Some risk factors are related to increased hsCRP. Moreover, tooth loss and hsCRP are risk factors for other chronic diseases and thus potential mutual confounders. Oluwagbemigun et al. summarize (in the Introduction of their paper) that the interrelationship between number of teeth and hsCRP remains unclear (and after having presented their results, it had not become clearer).

Some Remarks

So, is it justified to relate (as van der Wall does in his editorial) the linear relationship between number of teeth and poor outcomes (except myocardial infarction) in stable coronary heart disease, which was  found in the paper by Vedin et al. (2016), after insufficient confounder adjustment of a very large industry-funded novel drug trial in a huge number of countries (with participants who are certainly not representing the entire populations) which had failed to reject the null hypothesis about the main outcome, to tooth brushing (for a longer and healthier life)? Well, that seems to be outright nonsense despite the fact that teeth may indeed lead, in many cases, to some improvement in quality of life. (In others, in particular when considering high prevalence of periodontitis and insufficient treatment on a global scale, they may in fact increase the risk for a large number of ailments). An observation period of a median of 3.7 years is anyway not sufficient to draw definitive conclusions.

Fortunately, Pussinen and Könönen (2016) did not rule out, in the title of their commentary to the paper by Vedin et al. (2016), a confounded association.

10 April 2016 @ 10:11 am.

Last modified April 11, 2016.





Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s