Page 262 - Read Online
P. 262
interleukin* or interferon* or tumor necrosis factor quantitative and qualitative syntheses are quite weak in
or TNF* or NKκB) and (brain or cerebrospinal fluid Prisma, and their definitions fuzzy and because we were
or CSF)] and (“bipolar disorder” or mania or manic subsequently faced with an extreme heterogeneity of
or “recurrent depression”). There were no restrictions study designs and results that cannot be meta‑analyzed.
as to publication date or language. To be included,
the paper had to be an original article and carried out WHAT DID THE SEARCH PRODUCE?
in humans. We did not restrict our search to human
studies using the related PubMed function, because The PubMed search yielded 316 papers as of 11th of July
in our experience this practice does not exclude 2015. The search of the reference lists of these papers
complete animal studies but is likely to conceal some produced further 40 potentially interesting papers. Of
relevant human investigations instead. Hence, animal the total output of 356 papers, 100 were completely
studies were subsequently excluded on the basis of unfocused and were captured for the presence of
evidence. Furthermore, we excluded reviews and spandrels in the search strategy (however, we chose
meta‑analyses, opinion/speculative papers, animal not to modify the strategy by adopting a more specific
studies, case reports, as well as papers conducted approach, since this would result in a potential loss of
without respect to the 1964 Declaration of Helsinki otherwise eligible material), 98 were excluded because
principles of human rights, reviews and meta‑analyses, they were reviews or meta‑analyses, 16 were opinion
opinion/speculative papers, animal studies, as well as articles/editorials or speculative with no experimental
case reports. However, their references, especially those data, 55 were animal studies, 5 were case studies or case
of reviews/meta‑analyses were searched for possible series, 7 did not include patients with bipolar disorder
further includible papers. This review followed the separately or did not investigate neuroinflammation
bureaucratic Prisma statement [19,20] when appropriate. at all, 50 were investigations of peripheral markers,
We inverted the order recommended for screening, 1 was a duplicate, and 4 were carried out in vitro on
that is, first exclude duplicates and then the screen. human glial cells. The flow diagram in Figure 1 shows
We first classified studies according to their type, the search strategies and papers selected for review
then excluded animal studies, reviews and studies according to a modified Prisma algorithm. A total final
of peripheral immunity, and included only human number of 20 papers were found to be eligible and
studies with data, including clinical and postmortem, were analyzed. The search output spanned from 1982
but not in vitro experiments on cell lines. We also to 2015; recalling that the first occurrence in PubMed
excluded and classified as unfocused those postmortem of the term neuroinflammatory was in 1983 that of
studies investigating glia in the brain without telling neuroinflammation was in 1995, and that microglial
microglia from other types of glial cells. We excluded activation and related terms appeared during 1973‑
all nonpeer reviewed literature, as it could constitute 1975, we may presume that there was a dearth of
a source of bias. We considered as duplicates only focused papers in the 1st year, that is, prior to 2000.
studies that reported the same results in different However, one of the includible papers was dated
published reports. When different studies by the same 1997. This was the case (19 papers up to 1999 and
research group progressively report on increasingly 337 from 2000 on) and the trend has been increasing
larger samples and when the past used sample is used through years, witnessing an increasing interest of the
and accrued, we adopt the strategy to disregard the first scientific community in the neuroinflammation issue
appearing papers, including only the last study with in psychiatric disorders in general, and in bipolar
the larger sample, provided its quality of evidence is disorder in particular. Of interest, two of the included
high. Papers reporting on the same samples, but on papers were not identified by the PubMed search despite
different measures, were considered as different studies the strategy was appropriate for singling them out. The
and included if appropriate. Contrary to the distinction included papers are shown in Tables 1 and 2. They
made by the Prisma statement between records and fell into two broad categories, postmortem (n = 15)
full‑text articles, we considered all papers emerging and in vivo (n = 5), the latter comprising four studies
from our research as papers to obtain in full text and of neuroinflammatory mediators in the CSF and one
carefully searched for any data that conformed to our was an ingenuous positron emission tomography study
aims. Our experience is that you can never say if you of the binding of a neuroinflammatory marker in the
only read the abstract, especially for not so recently brain. The only study we excluded as a duplicate
published papers. was Rolstad et al. [21] which focused on the correlation
between CSF neuroinflammatory markers and cognitive
Contrariwise to what Prisma dictates, we did not label performance and admittedly reported data previously
our review as “systematic” and did not attempt to reported in Jakobsson et al., [18] despite differences in
carry out a meta‑analysis, both because the concepts of sample size.
254 Neuroimmunol Neuroinflammation | Volume 2 | Issue 4 | October 15, 2015