The authors of a paper showing a link between immune response and depression requested a retraction after they realized they’d merged two spreadsheets with mismatching ID codes.
Here’s the notice for “Lower CSF interleukin-6 predicts future depression in a population-based sample of older women followed for 17 years,” retracted in February 2014:
This article has been retracted: please see Elsevier Policy on Article Withdrawal (http://www.elsevier.com/locate/withdrawalpolicy).
This article has been retracted at the request of the Editor-in-Chief and Author.
The authors informed the journal that the merge of laboratory results and other survey data used in the paper resulted in an error regarding the identification codes. Results of the analyses were based on the data set in which this error occurred. Further analyses established the results reported in this manuscript and interpretation of the data are not correct. Based upon this new information, this manuscript is retracted.
The paper has been cited four times, according to Google Scholar. Here’s the abstract:
Objective
The literature regarding cerebrospinal fluid (CSF) cytokines in geriatric depression is sparse. The aim of this study was to examine associations between CSF interleukin-6 (IL-6) and related proinflammatory cytokines and current and future depression in a population-based sample of older women who were followed for 17 years.
Methods
83 non-demented women aged 70–84 years who participated in the Prospective Population Study of Women in Gothenburg, Sweden took part in a lumbar puncture in 1992–3. CSF- IL-6, interleukin-1β (IL-1β), interleukin- 8 (IL-8) and tumor necrosis factor-α (TNF-α) were measured. Psychiatric symptoms were rated with the Comprehensive Psychopathological Rating Scale at baseline and at three subsequent face-to-face examinations. Depression (major or minor) was diagnosed in accordance with DSM-IV/DSM-IV research criteria.
Results
At baseline, women with ongoing depression had lower levels of IL-6 (p < 0.04), IL-8 (p < 0.05) and TNF-α (p < 0.05) compared with those without depression. In women without depression at baseline, lower CSF IL-6 levels predicted depression at one or more follow-up examination (p < 0.03). Results from the generalized linear mixed logistic model using all baseline and follow-up data on depression status and Mini Mental State Examination score showed a significant relationship between IL-6 and depression (p = 0.005 OR 0.370 CI [0.184–0.744]).
Conclusion
Lower levels of CSF IL-6 were associated with current depression and with future depression during a follow-up of almost two decades. Our findings suggest that lower levels of CSF IL-6 may be related to depression vulnerability in later life.
Author Silke Kern let us know the paper was republished in the same journal in May of this year. Here’s the abstract of “Higher CSF interleukin-6 and CSF interleukin-8 in current depression in older women. Results from a population-based sample,” which appears to come to the opposite conclusion:
Abstract
OBJECTIVE:
The literature regarding cerebrospinal fluid (CSF) cytokines in geriatric depression is sparse. The aim of this study was to examine associations between CSF interleukin-6 (IL-6), interleukin-8 (IL-8) and depression in a population-based sample of older women who were followed for 17years.
METHODS:
86 dementia-free women aged 70-84years who participated in the Prospective Population Study of Women in Gothenburg, Sweden took part in a lumbar puncture in 1992-3. CSF IL-6 and CSF IL-8 were measured. Psychiatric symptoms were rated with the Comprehensive Psychopathological Rating Scale at baseline and at three subsequent face-to-face examinations. Depression (major or minor) was diagnosed in accordance with DSM-IV/DSM-IV research criteria.
RESULTS:
At baseline, women with ongoing major (n=10) or minor depression (n=9) had higher levels of CSF IL-6 (p=0.008) and CSF IL-8 (p=0.007) compared with those without depression (n=67). Higher CSF IL-8 was related to higher MADRS score (p=0.003). New cases of depression were observed in 9 women during follow-ups. No associations between CSF cytokine levels and future depression could be shown in women without depression at baseline.
CONCLUSION:
Higher levels of CSF IL-6 and IL-8 were associated with current depression in this population-based sample. CSF IL-6 and CSF IL-8 may play a role in depression in late life.
The paper also includes this nice nod to transparency:
The reader should take note that this paper replaces our retracted paper (Kern et al., 2013) in which an error had occurred when merging laboratory and survey data.
We don’t always see that in republished retracted papers.
Hat tip: Rolf Degen
Wait a minute… I’m not at all a statistician, but I would be curious to know how wrong and correct raw data can provide opposite results:
Wrong data —> “Lower levels of CSF IL-6 were associated with current depression […]”
Correct data —> “Higher levels of CSF IL-6 [and IL-8] were associated with current depression […]”
Does that mean that, regardless of the correctness of raw data, it was possible to get statistically significant conclusions (a.k.a. “proper conclusions” or “positive results” or whatever name you call it) ?
I don’t see the inevitable conflict here. Take a look at the number of subjects reported in the original paper: n=10, n=9, only the control group stands out (n=67). If they truly messed up data and accidentally merged their relatively small datasets with irrelevant values, then the outcome could have been easily altered.
This is very easy to understand IF CSF was in one dataset, and depression levels were in the other. Then merging them without keys would result in an arbitrary relationship, which happened to be higher. That is the point of the bad merge – the appropriate rows were not joined together.
This is, yet again, the same deal as with Potti: Doing an operation that you are not expert at, in this case data management, and then going on to write a paper. I’m a statistician. No one would dream of hiring me to do Western blots, but they do their own statistics all the time. Dreadful garbage, irreproducible results, and in the long run, bad science is the result.