In a paper published last year in the New England Journal of Medicine (NEJM), Messerli noted "a surprisingly powerful correlation between chocolate intake per capita and the number of Nobel laureates in various countries," which he postulated as "most likely" related to the cognitive enhancing benefits of chocolate.1 This article in one of the premier journals in the world received extraordinary press reports around the time of its release including in television (BBC News, CBC) and print (Time, Reuters, Associated Press, and Forbes) media. We found substantial methodological and conceptual errors in the Messerli article that we brought to the attention of the editor of the NEJM shortly after its publication. However, much to our surprise and dismay, no letter to the editor (including ours), note or article addressing these concerns have appeared in the NEJM since publication of the original article, although several authors have subsequently brought several statistical criticisms to light in other journals,2-5 including Practical Neurology. While several authors who have criticized the Messerli article have done a wonderful job of highlighting the limitations of association studies, no one, to date, has addressed two issues we feel are even more important to the scientific process: 1. Messerli's use of secondary sources (i.e., Wikipedia) for his data and 2. Messerli's selective use of data to support his hypothesis. In his article, Messerli notes, correlation does not "prove" causation and he notes he lacks the specific chocolate intake of individual Nobel Laureates. However, correlation between 2 variables is frequently used for hypothesis generation and Messerli suggests several hypotheses to account for his correlation. First, hypotheses generated from data are, of course, no better than the data on which they are based. Therefore, it is essential that authors use primary sources when possible. Messerli used data on Nobel laureates from Wikipedia6, which were based upon a BBC source, instead of data obtained directly from the Nobel Prize website7 which is readily available. The distinction is important and changes the resultant number of awardees based upon whether one chooses to count a Nobel laureate who is born in a country or where the person is located at the time of the award. For example, there are 323 Nobel laureates (1901-2012) who were affiliated with an organization in the USA at the time they were awarded the Nobel prize and 247 Nobel laureates who were born in the USA.7 Second, and perhaps even more important than the sources from which data are drawn, is the use of complete data sets and not just those that support a hypothesis. The selective use of data from secondary sources has no place in science to support, refute or generate hypotheses. This is a serious methodological lapse and is called into question by the following: 1. Wikipedia lists 71 countries with per capita Nobel laureate data6, yet only 23 were selected for Messerli's analysis; and 2. Data on chocolate consumption used were from non-contiguous years (2002, 2004, 2010 and 2011) although data from the same sources are available for multiple intervening years and per capita chocolate consumption for countries also change from year to year but Messerli chose certain data while excluding other data. Which data were selected skews the correlation substantially. For example, if one selects the top 10 chocolate consuming countries and using Messerli's data, the correlation is non-existent (r2= 0.04). It is important to remember in any hypothesis generating study that bias is easily introduced into that study through design flaws (e.g., selective use of data or use of secondary sources) as well as known statistical artifacts (e.g., ecological analysis2). Such bias can render the study completely useless or, worse, damages the integrity of the scientific process and the peer-review process of publication. As with others in the medical and scientific community2-5 and the press, we did not see Messerli's article as a joke intended by the author and the NEJM. However, we assert that if an article is published in a medical journal draws widespread publicity and is interpreted inappropriately as a valid scientific study, it is the responsibility of the editors of that journal to clarify the misinterpretation. Otherwise, the integrity of all scientific journals may become suspect. This will be the subject of an editorial in Practical Neurology. Competing Interests: None Contributorship statement: This article was originally conceived by BDP and was discussed with the other two authors who helped in the analysis and writing of the original article and its reviews. Each author contributed to the writing and revision of this revised article and approved the final version.

References 1. Dunstan, F. Nobel Prizes, Chocolate and Milk: The Statistical View Pract Neurol 2013;13:206-207. 2. Messerli, F.H. Chocolate Consumption, Cognitive Function, and Nobel Laureates N Engl J Med 2012;367:16:1562-4. 3. Maurage, P., Heeren, A., Pesenti, M. Does Chocolate Consumption Really Boost Nobel Award Chances The Peril of Over-Interpreting Correlations in Health Studies" J Nutr 2013;143:931-933. 4. Kayser, M. Editor's pick: Christmas is coming-time for chocolate to get ready for your Nobel Prize Investigative Genetics 2012;3:26. 5. Linthwaite, S., Fuller, G.N., Milk, chocolate and Nobel prizes Pract Neurol 2013;13:63. 6. List of countries by Nobel laureates per capita [Internet]. San Francisco: Wikipedia.-[accessed 2012 October 25]. Available from: http://en.wikipedia.org/wiki/List_of_countries_by_Nobel_laureates_per_capita 7. The Official Website of the Nobel Prize [Internet]. [accessed 2013 May 30]. Available from www.nobelprize.org.

Conflict of Interest:

None declared

I thank the authors for highlighting this difficult topic with a very thought provoking article.

In immunosuppressing a patient with neurological disease, as with anything a physician does to a patient, there are potential risks to take into account and to be weighed up against the proposed benefits. Indeed, this sort of "cost-benefit" analysis is a central tenant of decision making in much of life outside of medicine also.

It is generally accepted that in order to proceed, the benefits should outweigh the "costs". It is clear that PCP prophylaxis is effective at reducing the incidence of PCP, but at what cost? In patients with solid -organ transplant for example, this is fairly clear as the incidence appears high, as noted in the authors references.

The problem is that we have little knowledge of the incidence of PCP in patients immunosuppressed for myasthenia gravis, for example. My personal feeling is that this risk is much lower, but there is little evidence in the literature to support this.

When considering that around 15% of patients develop side effects whilst taking trimethoprim-sulfamethoxazole, and these can (in rare cases) be fatal, it is not clear in my mind that the cost-benefit analysis for patients with neurological disease swings in favour of the widespread use of PCP prophylaxis. Put simply, would enough cases of PCP be prevented to outweigh the risks posed by the side effect profile of the medication?

To help us make this decision we need clearer data on the incidence of PCP in patients with neurological diseases who are being prescribed immunosuppressive medication.

Conflict of Interest:

None declared

We welcome the article by Jones et al, and in particular the chance to compare the protocol used in their institution with the one we hope to implement (1).

In Northern Ireland, we are in the process of drawing up a regional protocol for the management of status epilepticus, due to wide variation in current practice. We carried out preliminary work to see where the gaps in knowledge are, so that we can create a relevant protocol to assist junior staff, hopefully highlighting the need for a protocol.

We surveyed post finals - final year medical students (9 students) foundation (28 doctors) and core medical trainees (24 doctors) and got 61 responses to a questionnaire (~85% returns). Only 10% of these knew the appropriate dose and rate of administration of two first line drugs.

Only 54% would have given an initial 4 mg dose of lorazepam, and only 36% knew the correct dose of phenytoin (15-20mg/kg); 44% acknowledged they were unaware of the dose. One third knew the appropriate initial infusion rate of phenytoin.

Knowledge of the side effects of phenytoin was generally good, although worryingly 30% thought sedation was a side effect of phenytoin, which might lead to complacency in managing the apparently "post ictal" patient.

It was also evident that a significant issue for junior doctors is making the actual diagnosis of status epilepticus. Without clinical experience junior doctors lack ability in making a confident and accurate diagnosis, for which they are prepared to give drugs which are sedating and are perceived to be potentially dangerous. In addition doctors in the early stages of training had limited appreciation of potentially discriminant features such as back arching, biting the tip of the tongue, open eyes during the event and the synchrony of limb movements, in making a clinical diagnosis of seizure versus non epileptic attack disorder.

Better educational strategies that inform, everyone who manages status epilepticus, about seizure semiology (2) would go hand in hand with dissemination of treatment protocols and would therefore be useful.

We wish to add another vitally important point to the protocol by Jones et al, and that is the clear instruction to seek immediate urgent advice from the nearest neurology service in all cases of status epilepticus.

(1) Jones S, Pahl C, Trinka E, et al. A protocol for the inhospital emergency drug management of convulsive status epilepticus in adults. Practical neurology 2014;14:194-197

(2) Seneviratne U, Ding C, Bower S et al. Video based training improves the accuracy of seizure diagnosis. J Neurol Neurosurg Psychiatry 2014;85:466-470

Conflict of Interest:

None declared

I read with interest the debate between Nightingale el al. and Schott about whether we should routinely test for HIV in patients presenting with cognitive impairment 1,2 . I assume that while the authors have taken rather polarized views to state their point, the correct answer lies somewhere in between. Testing for HIV shall not be the foremost in my mind in a 75-year-old male presenting with slowly progressive cognitive impairment for the past few years accompanied with typical features of a cortical dementia (aphasia, apraxias and agnosias). In such a patient I shall rather harness my clinical and diagnostic skills to differentiate between Alzheimer's disease, frontotemporal dementia and Lewy body dementia. On the other hand a 45-year-old male presenting with cognitive impairment accompanied by a movement disorder, certainly HIV testing shall be included in my diagnostic armamentarium. One must though not forget that positive HIV serology is not synonymous with HIV encephalopathy or AIDS dementia complex which is a clinical syndrome comprising of a combination of cognitive, behavioral and motor dysfunction. Testing a patient with cognitive impairment for HIV shall have a high positive predictive value only when the test is carried out after taking the patient's age at presentation, history (not just of risk factors but also of the rate of progression of symptoms) and examination findings into consideration.

References

1. Nightingale S, Michael BD, Defres S, Benjamin LA, Solomon T. Test them all; an easily diagnosed and readily treatable cause of dementia with life-threatening consequences if missed. Pract Neurol. 2013;13:354-6.

2. Schott JM. HIV testing in dementia: test some, perhaps more, but not all. Pract Neurol. 2013;13:357-8.

Conflict of Interest:

None declared