The paper [ Proc. Natl. Acad. Sci. vol. 108 pp. 8189 – 8193 ’11 ], if true, provides a plausible explanation for (1) why attempts to replicate reports linking various genes to depression fail so often (2) ditto for studies of drug efficacy for depression. It also points to a far better way to test drugs for efficacy in a variety of psychiatric conditions. Clearly important stuff. If you don’t have it already, youl’ll need to pick up some background in molecular biology. This should be all you need — https://luysii.wordpress.com/category/molecular-biology-survival-guide/. Also, it might be worth a look at this — https://luysii.wordpress.com/2010/08/29/some-basic-pharmacology-for-the-college-student/
The biological susceptibility model posits that some genotypes are highly susceptible to environmental influences (orchids) while others are not (dandelions). The claim is made that postpartum depression (PPD) occurs in 10 – 20% which I think is extremely high (depends on how it is diagnosed). Major depression is defined in the DSM-IV as 3 or more depressive symptoms lasting over 2 weeks during which there is dysphoria or anhedonia (aren’t these depressive symptoms?). The sample of 1,206 mothers gave an incidence of PPD of 17% using these criteria in the first year after the birth of a child (1 year is quite long. the postpartum period is medically defined as the time maternal physiology returns to near pre-pregnancy status, which is about 6 weeks).
The Fragile Families and Child Wellbeing study (FFCWS) is a sample of children born in large cities between 2/98 and 9/00 with an oversample of kids born to unmarried parents (75%). Interviews were conducted within 3 days and subsequent interviews were conducted when the child was 1, 3, 5, and 9. DNA samples (saliva) were taken at age 9. The marker used for socioeconomic status (SES) was maternal education (not out of wedlock births) — either seems reasonable to me and I’ll analyze their results both ways. They used post-high school education as the determinant of SES. It’s a little hard to figure out how many women of the 1206 were in the high SES group. So I had to eyeball figure 1.D (p. 8190) to get a number around 30% giving 362 women of ‘high’ SES.
The study looked at gene variants (polymorphisms) of the serotonin transporter
Previous studies have focused on just one polymorphism of the serotonin transporter gene at a time– this study looked at two — however 6 different polymorphisms of the gene are known. One was in the 5′ regulatory region of the gene (called 5HTTLPR) the other was a 17 nucleotide variable number of tandem repeat (VNTR) in the second intron — (called STin2 VNTR).
The 5HTTLPR polymorphism: there is a short (S) 14 repeat and long (L) 16 repeat of a 23 nucleotide incomplete repeat. Other less common repeats are also found. At any rate the S allele has been associated with higher rates of mental health poblems including depression in many studies.
The number of S and 12 alleles was determined for each mother and totaled. So each mother could have 0, 1, 2, 3, 4 of the ‘risk’ alleles for depression. The effect of 0, 1, or 2 risk alleles was neglible (in both high and low SES groups) but effects were seen if the mothers had 3 or 4 of the ‘high risk’ alleles. So the 3 and 4 high risk allele group are orchids while the 0 – 2 group are dandelions.
There are all sorts of calculations in the supplemental material, but I’d love to see the actual numbers for the women with 3 and 4 of the bad risk alleles — how many were there? How many in this group developed depression? We’re talking 25% with married parents (bringing the total down to 300 in this group). Then reduce this to 17% (the average rate of depression in the total group) and we get an expected 51 cases. Next reduce number 300 by the % of the study population having the 3 of the high risk alleles and we have even less — from figure 1.C it looks to be around 35% (again I had to eyeball this, I couldn’t find an explicit number. So we now have an expected 17 cases in this group. Not a great number to play around with.
Another way to look at is is to figure out the percentage of women with more than a high school education (this is the actual marker they used for socioeconomic status), which looks to be around 30% (see figure 1.D p. 8190) this gives a group of 362 women, and at a 17% rate of depression gives 61 expected cases. How many in this group had the 3 or 4 risk alleles? Again about 35% of them do making 128 educated women with risk alleles (this contains possibly an unwarranted assumption to be discussed later). Again the expected number of cases of depression would be 128 *.17 = 21 cases. Presumably there were fewer cases of depression than expected by chance — but how many fewer – 1, 2, —- 10? I can’t find the data. They are making a very important conclusion based on at most 21 missing cases of depression. If there were 3 cases of depression in the groups the results would be conclusive, if there were 19 cases instead of 21 the results would be unimpressive. Which was it?
Another interesting way to look at the data, would be to see if the high number of risk alleles was over represented in any subgroup, married vs. unwed, high SES vs. low SES — they certainly have this data, but I couldn’t find it. The calculations in the previous paragraphs assumed that the distribution of risk alleles is identical in all groups — married, unmarried, high SES, low SES and various subgroups made from these (married and low SES, married and high SES etc. etc.). The authors certainly have this data, but I couldn’t find it. It should be a simple matter to put it on a spread sheet and print the results.
( It is an interesting mathematical question to compare the histogram of risk factor distributions in the various populations married/unwed, high SES/low SES and see if they are essentially the same. How would you explicitly define two distributions as being ‘essentially the same’ ? I’m sure this has been done, and that someone out there knows the answer. If you do how about a link? )
This is yet another example of why I hated reading the medical literature when I had to (for some other horrible examples, if you have the stomach see — https://luysii.wordpress.com/2009/10/05/low-socioeconomic-status-in-the-first-5-years-of-life-doubles-your-chance-of-coronary-artery-disease-at-50-even-if-you-became-a-doc-or-why-i-hated-reading-the-medical-literature-when-i-had-to/.) The results could be a statistical fluke.
Small numbers are no barrier to a definitive conclusion. Here’s a very recent example — a recent study on prophylactic use of antiretrovirals in couples one of whom was HIV1 positive. 1793 couples were studied in which one partner was HIV1+. Half started antiretroviral therapy (type not specified), the other half didn’t (presumably all the partners didn’t have clinical AIDS or a low CD4 lymphocyte count, as it wouldn’t be ethical to withhold treatment from them, and presumably if they did develop this over the course of the study, treatment would have been started). 6 years on the study was stopped (it planned to last 10 years) when preliminary data was analyzed. Of the 39 uninfected partners who become infected, 28 were infected from their regular partners (there’s a way to know if they got their partners HIV1). 27 of those infected from their partners were in the group where the partner had not begun antiretrovial therapy. Absolutely unequivocal despite the small numbers. A slam dunk.
Perhaps this is true of the depression study as well, but I can’t tell from the paper. Too seminal to ignore, too fantastic to believe without the actual numbers. Irritating — yes. Fairly typical — unfortunately.
There’s tons more for the drug developer and pharmacologist to think about. Perhaps the results of this paper (if true) account for some of the problems replicating risk factor and efficacy results of other studies.