Editor’s Choice

Are surveys worthwhile?

To be worth publishing, surveys have to address an important question, and equally importantly, they have to be conducted properly. It is commonly assumed that surveys are an easy form of research, but they are more difficult to do well than many imagine. I learned this the hard way in the mid-90s when I encouraged a research fellow, assisted by a medical student, to carry out a survey of how women felt about being delivered by caesarean section. As I had never supervised a survey before, I thought it would be a simple task. If only I had had the opportunity to read Patrick Chien’s excellent commentary (page 1285) before we started, we wouldn’t have made so many mistakes. Instead of questioning every woman having a caesarean, we only recruited them Monday through Friday (we didn’t like working weekends), immediately introducing potential bias. We failed to assess the questions for reliability and validity and we didn’t pilot them. Although we found the results of the study interesting and informative, we never published them because these flaws were immediately spotted by referees. The research fellow is now a consultant obstetrician at a well-known London teaching hospital, and the erstwhile medical student is on our graduate training programme, so I meet them regularly, and each time I still feel embarrassed that they didn’t have better supervision. We learned some important lessons, but it would have been much better to have performed the survey properly in the first place. Unfortunately, at BJOG we still get sent the results of surveys that have been performed badly and are therefore uninterpretable. Luckily, my team’s failed survey took only 4 months, but I have known research fellows spend years on surveys only to obtain unpublishable results. I recommend everyone to read Dr Chien’s excellent summary of survey methodology, and if they are going to carry out a survey, they should read all the sources he quotes before embarking upon it. These include the excellent ‘Equator’ website (http://www.equator-network.org/), which we encourage all authors to visit before submitting their papers.

What do we look for when deciding which surveys to publish? Appropriate methodology is vital, and the questions asked need to address issues of at least national, and preferably international, significance. We receive many surveys which will have been valuable to the groups who carried them out, but are so location specific that they are not generalisable to our international readership. In contrast, we publish the survey by the EUROBS study group (page 1340) of obstetricians’ experience of late termination of pregnancy in eight European countries. Unsurprisingly, the most common indications for late termination of pregnancy were congenital anomalies and poor maternal health. However, I was surprised that in five of the eight countries surveyed, feticide was not generally used before the baby was delivered, while in France and the Netherlands, active euthanasia of liveborn babies was practised (In the UK for example, this would not be legal). Although a consensus of management has been developed in some countries, the results of the survey show that there are many variations in practice, both ethical and legal, between countries. It is important for obstetricians to be aware if they move from one country to another that practice varies widely if they are not to fall foul of local feelings and indeed the law.

The purpose of surveys can also be to compare the attitudes of different groups. We publish the results of a survey of women and professionals by Boormans et al. (page 1396) regarding prenatal diagnostic tests for fetal abnormality. The professionals preferred the most rapid tests (for trisomy 21), presumably on the assumption that this is what women and their families would choose. In fact, just over half the women preferred the test giving the most information (the whole karyotype) over the rapid test, even though abnormalities other than Down’s syndrome are rare. On reflection, one can hypothesise that the women having the tests are less likely to be influenced by considerations such as cost effectiveness (particularly when the test is provided by the state so the individual does not have to pay) or by the population performance characteristics of the test, than by what it offers them as individuals. This brings into focus the perennial problem of balancing what individuals want if given a free choice with what society is willing to pay for.

What causes preterm labour?

Of the many and varied causes of preterm labour, inflammation and infection have been among the most intensively studied over the last decade. While it is well established that acute infections can precipitate labour, the role of abnormal bacterial colonisation of the vagina remains controversial. One abnormal form of abnormal colonisation, bacterial vaginosis, has frequently been proposed as increasing the risk of preterm birth. A recent paper (Xu J. et al., Obstet Gynecol 2008;112(3):524–31) has suggested that bacterial vaginosis is linked with reduced levels of defensins (human neutrophil peptides 1–3) in vaginal fluid, indicating a reduced host response to abnormal bacteria. Both bacterial vaginosis and low defensins are more common in women of black African origin, and this (by promoting earlier birth and therefore smaller babies) may be in part an evolutionary response to their high rate of obstructed labour. Speculation about the role of bacterial vaginosis has led to trials of prophylaxis against preterm labour using antibiotic treatment, some of them reported in this journal. However, a 2006 Cochrane meta-analysis of 13 such trials involving 5300 women, by McDonald, Brocklehurst and Parsons (doi: 10.1002/14651858.CD000262.pub2), reported that treatment was not effective in reducing the incidence of preterm birth at any gestation. Donders et al. (page 1315) report their study of 1026 unselected pregnant women who had their vaginal microflora assessed between 9 and 16 weeks of pregnancy. They found that absence of lactobacilli was associated with an increased incidence of preterm birth (odds ratio 2.4, 95% confidence interval 1.2–4.8). Paradoxically, women with ‘full’ or classical bacterial vaginosis did not have an increased incidence of preterm birth, whereas those with bacterial vaginosis plus other abnormal vaginal flora did (odds ratio 3.3, 95% confidence interval 1.3–8). The presence of Mycoplasma hominis, although uncommon, was strongly associated with preterm birth, with six of 14 women carrying this organism delivering preterm. These results indicate the complexity of the situation which will make developing effective preventive therapies difficult. The disparate nature of previous studies of prophylaxis with antibiotics is well summarised in their discussion. Clearly, this area of research will run and run.

Do endometrial polyps detected at hysteroscopy need to be removed?

I have written before in these columns about the tendency of new technology to generate activity which is self-justifying, when what is needed is careful and critical evaluation of effectiveness. Commercial imperatives dictate that the use of new equipment, often developed at substantial cost, is heavily promoted. Delays in marketing imposed by the need to validate effectiveness are not welcomed. When hysteroscopy was first introduced, I remembered the glee with which many of my gynaecological colleagues described visualising hitherto unsuspected endometrial polyps. They gave every sign of enjoying devising ways of removing them, and were perplexed when I asked whether all this activity was really necessary. How many normal women, I asked, had these polyps and yet were asymptomatic? Would it harm women to keep their polyps? I was encouraged that my question had not been hopelessly naïve when I read the paper by van Dongen et al. (page 1387). They report that endometrial polyps are found in up to 10% of asymptomatic women, they may regress spontaneously, and removal can cause morbidity. The likelihood of cancerous change is low. After several decades of the use of hysteroscopy, there is little good quality evidence that removing such polyps is beneficial. Van Dongen et al. report the results of treating only 21 premenopausal patients but did at least find a modest but statistically significant reduction in the median monthly bleeding score. This positive finding is somewhat undermined by the fact that they excluded six of the original participants eligible for the study who had to be treated following surgery, with oral contraceptives or a levonorgestrel intrauterine contraceptive system, because of continuous bleeding problems – clearly, in these six, the surgery was not beneficial and may have exacerbated their condition. Accordingly, their study might be considered relatively slight justification for hysteroscopic polypectomy. But surely, women after the menopause should not have endometrial polyps, and therefore these must be removed if repeated investigation for recurrent bleeding is to be avoided? Timmermans et al. (page 1391) described a laudable attempt to address this issue with a prospective randomised controlled trial. They describe 361 participants who attended the outpatient clinic with postmenopausal bleeding, and 255 of these had outpatient hysteroscopy. However, of the 246 participants eligible for the trial, only 105 participants were informed about the trial before surgery. Of the 43 participants who had an endometrial polyp, only four gave permission for randomisation to resection or non-resection. Presumably, the participants agreed with many gynaecologists that ‘polyps should not be there and therefore should be removed’. The lawyers have a phrase for this, ‘res ipsa loquitur’, or ‘the thing speaks for itself’ (i.e. it is self-evident). But as I have often pointed out in these columns, things that are obvious are not always true. Risking the wrath of any gynaecologist who has read this far, I would suggest that we still don’t know whether hysteroscopy and polypectomy are useful or are simply a way of justifying expensive professional time.

How many hysterectomies are really necessary?

One of my favourite mock examination questions with trainees was ‘why do gynaecologists perform hysterectomies?’ Almost invariably, I would get a list of medical indications when what I was trying to get them to see was that some surgeons do hysterectomies for doubtful indications because they like operating, because they wish to give their trainees surgical practice, because they want to remove a ‘heart sink’ participant from repeatedly attending their clinics or (dare I mention it?) because in private practice it increases their remuneration. The huge variations from one region to another in the UK in the use of hysterectomy to treat menorrhagia, reported by Cromwell et al. (page 1373), makes it difficult to sustain the view that such differences are entirely due to medical factors.

And finally

Our instructions to authors says ‘We do not publish case reports unless they highlight …. previously unpublished complications of new techniques or medications’. By coincidence, we received close together two case series reporting vulval ulceration associated with the use of nicorandil, a drug used to treat angina. A PubMed search revealed no previous papers reporting this adverse effect, and so we thought we should bring it to the attention of our readers. The papers by Fraser et al. and Chan et al. are published on pages 1400–1405.

Over the last year or so we have had a number of papers questioning the validity of customised birthweight centiles in relation to parity and maternal body mass index. I have repeatedly pointed out that the ultimate test of their usefulness relates not just to their ability to predict size at birth, but perinatal mortality. Professor Gardosi, an early proponent of customised centiles, has, together with colleagues Clausson and Francis, used the Swedish birth registry database to study more than a third of a million births, and their study reported on page 1356 was therefore powered to use perinatal mortality as an endpoint. He concludes that use of customised centiles to determine smallness for gestational age does improve the identification of pregnancies at increased risk of perinatal death. Is this the last word?