Article Text

Download PDFPDF

Practice corner: will it happen again doctor? Prognosis after a first seizure
  1. Mike Crilly, MD, MPH, MRCGP, MFPHM
  1. University of Aberdeen Medical School, Aberdeen, UK

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


    Susan attends with her mother for a consultation at the end of Monday morning surgery. On the previous Friday evening, 22 year old Susan was alone in the kitchen when her mother heard a clatter of kitchen utensils. She went through to find Susan unconscious on the floor in a pool of urine with her arms and legs shaking. This subsided after a couple of minutes leaving Susan drowsy and confused. Susan takes no regular medication and has had no serious illnesses in the past. There is no suspicion that alcohol or illicit drugs are involved. This is the first time that something like this has ever happened. She was seen in the local accident and emergency department and allowed home after being reassured that her blood tests, heart tracing, and clinical examination were all entirely normal. A neurological outpatient appointment is being organised, and Susan has been told that she mustn’t drive.


    Both Susan and her mother are very worried that she’s going to have another convulsion. We discuss some of the implications of her recent convulsion and the difficulty of adopting an essentially “wait and see” approach pending further neurological assessment. I’m reluctant to hazard a guess as to the likelihood of a recurrence. My clinical intuition is that the risk of recurrence is quite high (if pushed I’d estimate 70% over 12 months). A straw poll of 6 general practitioner (GP) colleagues over morning coffee produced a wider range of estimates (10, 20, 40, 60, and 70%). But I’m conscious that none of our estimates are based on extensive clinical experience. Primary care physicians only see around 1 new patient with a first seizure every couple of years.1


    The PubMed Clinical Queries filters include a focused search for research relating to “prognosis.” The search can set to be either broad (include a lot of irrelevant articles, but unlikely to miss many relevant ones) or narrow (fewer irrelevant articles, but also likely to miss a few relevant ones). I used the free text term first convulsion (figure 1), reasoning that a clinical researcher interested in prognosis would probably use the term convulsion (a clinical presentation) rather than epilepsy (implying a more certain diagnosis).

    A narrow search yielded 260 references (table), which is far more than I’ve got time to scroll through (a broad search locates even more citations, 586). Selecting the alternative “systematic reviews” option reduces the number of references to 29. The seventh article is “an evidence based approach to the first unprovoked seizure.”2 Unfortunately the PubMed abstract contains no prognostic data, but it does indicate that the article includes a 1 page critical appraisal of the topic (CAT) based on a published meta-analysis.

    Number of articles retrieved by PubMed searching (26 October 2004)*

    PubMed often has direct links to full text journals, but not this time. Google ( is the quickest way that I know to check if a journal is available electronically (Google is so useful for evidence-based medicine [EBM] that I’ve installed the free toolbar on my web browser). Googling Canadian Journal Neurological Sciences rapidly locates their website, which is “celebrating 30 year of publishing” by providing free online access ( The selected article2 was actually written to illustrate the EBM approach and it critically appraises the published meta-analysis against explicit criteria.4


    Based on pooled results (of 1930 patients from 13 studies with ⩾12 months’ follow up), the overall risk of seizure recurrence after a first unprovoked seizure is 42% (95% CI 39% to 44%) after 2 years’ follow up.2 An increased risk of recurrence is associated with abnormal neurological findings (which Susan doesn’t have) and an abnormal EEG.2 A quick look at the PubMed abstract for the original meta-analysis (also picked up on my search) adds little further information, other than generalised seizures having a lower recurrence risk than partial seizures.3 Although the meta-analysis was published rather a long time ago (1991), it was still the only systematic review available when the CAT was produced in September 2001.2 The journal Neurology is available online (Googling journal neurology immediately locates, but unfortunately electronic full text is only available from January 1999.


    Google ( is also very useful for quickly locating other important documents. For example, Googling fitness to drive immediately locates the DVLA (Driver Vehicle Licensing Agency) current medical standards for driving ( The latest version (updated September 2004) indicates that a car driver in the UK with a “first epileptic seizure/solitary fit [incurs] one year off driving with medical review before restarting driving.”

    If I had known at the time of my initial search that SIGN (Scottish Intercollegiate Guidelines Network, had produced epilepsy guidelines (April 2003) or that Clinical Evidence ( covers the topic of epilepsy (March 2003), then I could have Googled rapidly to their summaries of the same meta-analysis.3


    I noticed on my “first convulsion” search that many of the citations identified used the word “seizure” rather than “convulsion.” The table above shows how some alternative search terms panned out. “Clinical Queries” is clearly a much better option than a standard PubMed search as it filtered out more than 75% of the citations. However, both the broad and narrow clinical query searches (respectively filtering out 75% and 90% of citations) still produced far more citations than I’d time to scroll through in the clinic. The “Systematic Reviews” option consistently produced the most manageable number of citations, filtering out some 99% of the citations while still including my 2 key references.2,3

    The “systematic reviews” option pastes “systematic[sb]” into the search (sb stands for subject subset). This filter preferentially identifies systematic reviews, meta-analyses, evidence-based guidelines, and pre-appraised resources (such as the ACP Journal Club). The search involving “Seizures [MeSH]” was the most useful approach for identifying the smallest number citations, but it took slightly longer to construct (and it only filtered out an additional 3 citations compared with my original search). It involved using the “MesH Database” option in PubMed to look up the appropriate Medical Subject Heading [MesH] for my term “convulsion.”


    I was puzzled that neither of the 2 key references2,3 from my “first convulsion” search actually contained the word “convulsion” anywhere in their PubMed entry. How had I managed to pick them up with my simple search? Checking “PubMed Help” revealed that unqualified searches (like “convulsion”) are mapped against something called a MeSH Translation Table. This automatically matches such words against clinical synonyms, English language variants, and the relevant Medical Subject Headings (MeSH). The “details” option (figure 2) reveals what goes on behind the scenes in PubMed. My simple use of the word “convulsion: was translated into:

    "seizures"[MeSH] OR "convulsions"[MeSH] OR convulsion[Text Word].

    With its in-built filters (and automatic term mapping), PubMed Clinical Queries is a key resource for rapid searching. It performs equally well for questions of diagnosis, therapy, and harm.5 It’s particularly useful for peripatetic physicians. On any web linked PC I can rapidly access PubMed (just Google PubMed) without the need to remember any usernames or passwords.


    Clinical practice is an inherently uncertain activity. Although EBM can’t remove this uncertainty, it can at least help to reduce it. A recurrence risk of 40% (at 2 years) is considerably lower than my initial estimate of 70% (at 1 year). While I still can’t predict the future for Susan, I can at least contribute the best currently available evidence to our discussions. I telephoned Susan and her mother to explain that the outlook was more optimistic than I had originally thought.


    View Abstract