In an analysis of the data from Health Survey for England, people who managed more than seven helpings of fruit and veg did best. Since I struggle to make even three or four helpings, this meta analysis came as something of a relief.
The science of meta-analysis has developed substantially over the past decade and frankly, I have no idea about Egger, Begg, 'contour enhancement' or 'trim and fill':
"There were no missing studies imputed in regions of the contour enhanced funnel plots. Egger’s linear regression test indicated that the P value was significant. No publication bias was found for Begg’s rank correlation test. Also, the application of the trim and fill method did not change the average effect size, further suggesting that results were not affected by publication bias."
Fortunately the rest of the paper is a bit easier to understand.
When I was young (i.e. in the 1980s) I heard an eminent psychiatrist say that smoking prevalence among patients with schizophrenia approached 100%. Later, I got involved in tortuous debates about whether it was cruel to forbid smoking in long stay psychiatric wards - surely it was their home, and they should be free to smoke there?
So I was interested in this paper even before I noticed that the lead author was Judith Prochaska, inventor of the famous 'Stages of Change' model. Let's spend a bit of time on the 'P' of Population, Intervention, Comparator and Outcome (PICO). It's an inpatient psychiatric population: but let's delve a bit deeper.
"The mean age of the sample (n = 224) was 40 years (SD = 14 years); ... Major psychiatric diagnosis groups were unipolar depression (47%), bipolar depression (25%), and schizophrenia spectrum disorders (15%)... Hospital stays averaged 7.4 days...
Reasons for hospitalization were danger to self (75%), grave disability (10%), and danger to others (2%); 13% were voluntary admissions. Most participants (72%) had a previous psychiatric hospitalization."
So this is a short stay population, mostly with depressive disorders rather than schizophrenia. And it's not clear why they all needed to be on a locked ward. Smoking prevalence was 20% among the patients randomised, which seems low till you remember that this is California.
The study ran from 2006 to end 2008, with 18 month follow up, so the fieldwork was complete by mid 2010. I've complained before about academics who take years to publish research which could save lives so I won't do it again here (oops...).
Folic acid is a B vitamin. Folate is the naturally occurring form but it's not as bioavailable as folic acid.
Blood consists of cells and plasma; serum is plasma without the clotting factors.
Folate concentrations in plasma (or serum) vary from day to day. The concentration in red blood cells is less variable, but can only be measured in specialist laboratories.
So now we can turn to this paper, which used a somewhat complicated method but tackles an important problem. The aim was to relate concentrations of folate in the red blood cell (rbc) to the prevalence of neural tube defect. We need to know this to make sensible judgments about folate supplementation. Is there, for example a plateau beyond which further increase makes no difference? Or a lower threshold?
The study used a three stage design:
a. the Community Intervention Trial encouraged women to take folic acid tablets during pregnancy. It monitored number of pills taken and neural tube defect outcomes, but didn't measure rbc folate levels.
b. the Folic Acid Dosing Trial gave data relating rbc folate concentrations to folic acid supplementation but had nothing on pregnancy (indeed pregnancy was an exclusion).
c. The second trial was used to estimate rbc folate levels for each woman in the first trial, given her known intake of folic acid pills. An adjustment was needed because there is a gene (MTHFR) which affects rbc folate concentration at any given level of folic acid intake, and the prevalence of the gene varies between different populations.
This is modelling, not direct measurement, and you should always check a model to see how well it performs against real life. The researchers checked against data from the USA, where two data points were available for folic acid intake and ntd prevalence (before and after mandatory fortification of flour with folic acid); and against a study done in Ireland 30 years ago.
All of which provides the necessary science if you're trying to decide whether fortification of flour should be mandatory.
The 'rule of 10' tells us to suppress the information in any cell of a table where the number is less than 10, and what you're counting is people. It's been standard practice in census outputs for many years; we also do it for tables of data on HIV infection. The aim is to avoid inadvertently identifying an individual - there may for example only be three males aged 35-44 of Chinese ethnic origin in a particular census zone, so by a process of elimination you might be able to identify a particular individual. This is called 'deductive disclosure'.
The authors of this paper worried about what widespread suppression of data might do to geographical analyses, particularly since suppression is more likely in sparse populated i.e. rural areas.