Letters

To the Editor: A recent article by O’Donnell and colleagues1 claimed contradictory results to a previous study which found that compensation was associated with worse health and return-to-work outcomes after injury.2 Their findings were similar to those of the previous study until they excluded a group of non-compensable patients because they had accessed private health insurance. The authors argued that “private health insurance was similar to other compensation agencies in that patients in this group had their health care costs met”. Using this argument, all patients would be compensable, as Australia has a universal health care system in which all Australians have their health care costs met. There is no precedent in the literature for such an exclusion. Compensation bodies provide additional payments beyond health costs, including payment for pain and suffering and income replacement. They also involve patients in a complex process with many features thought to influence outcomes (eg, the adversarial nature of making compensation claims and delays in receiving payments). We believe that the exclusion of private patients from the non-compensable group in the study by O’Donnell et al was incorrect and reduced the already small study sample, limiting the capacity to identify differences across groups. Furthermore, O’Donnell and colleagues found that compensable patients had higher anxiety levels at 24 months, until a supplementary analysis showed that, after controlling for stressful interactions with compensation agencies, compensation itself became non-significant. Surely stressful interactions are one of the mechanisms by which any compensation effect might be mediated. To say that an association is not significant once the mechanism of the effect is allowed for is akin to stating that smoking is not carcinogenic once the carcinogens are allowed for. O’Donnell and colleagues appear to be stating that simple access to compensation is not harmful, with which we agree, but fail to consider the complexities of compensation involvement. Both studies1,2 share a common limitation — that of comparing victims of transport-related injury with victims of other injury types. A recent study confirmed compensation and lawyer involvement as predictors of worse outcomes in a study of compensable and non-compensable transport-related trauma.3 A true understanding of the effect of compensation requires comparison of patients of comparable injury circumstances (eg, road trauma) and different compensation systems. Studies are clearly needed to establish a better understanding of the complexities of compensation delivery and the impact on outcomes. O’Donnell and colleagues’ conclusions have the potential to mislead compensation authorities and other stakeholders who should be focused on addressing this issue.

Belinda J Gabbe · Ian A Harris · Alex Collie · Peter A Cameron

To the Editor: In their recent study, O’Donnell and associates1 examined the effect of compensation, and the clinically vexing problem of interaction with insurance companies, on recovery after hospitalisation for trauma in Victoria. They concluded that access to compensation might not be associated with a poor outcome per se. We agree that the relationship between compensation and health outcomes is complex, but believe there are a number of conceptual and methodological issues that undermine their findings. First, as they note, this sample of injured people might not be representative of those making an insurance claim. In an earlier study of motor vehicle accidents in New South Wales,2 less seriously injured victims who only attended their general practitioner or spent less than a day in hospital comprised as much as 70% of those seeking compensation. Furthermore, the article by O’Donnell and colleagues provides no description of any differential attrition with respect to factors that may be associated with poorer psychosocial outcomes (such as previous psychiatric disorders), other than sex and acute hospital factors. The outcome measurement characteristics change in the course of the analyses, potentially undermining power to detect any differences. For example, the quality-of-life and disability measures become dichotomised using norms in the population for the modelling, rather than as scores in the baseline characteristics as in Boxes 2 and 3. This approach has the potential to conceal true differences between the groups because of regression to the mean and baseline differences in both groups. The main problem relates to the possibly post-hoc exclusion of privately insured subjects from one group. We do not believe that private health insurance can reasonably be considered to be “compensation”. It can provide money to cover the cost of inpatient treatment and very limited outpatient services, but provides no more recompense and retribution for injury than Medicare. Older and wealthier Australians disproportionately hold private health insurance. This group is likely to differ on a number of factors, many of which are associated with better psychosocial outcomes. Although the authors have evaluated some demographic factors, this is likely to have introduced some potentially significant confounding. Thus there is little justification for the removal of this group from the non-compensable group alone. We would be interested to see an analysis after the removal of subjects with private health insurance from both groups. This would allow a more rigorous examination of the effect of one factor — actual insurance compensation — on recovery outcomes.

Nicholas S Glozier · Matthew Large

To the Editor: O’Donnell and colleagues seek to extend and improve on previous research into the relationship between compensation status of injuries and medium-term health outcomes.1 Improvements are needed because much of the empirical analysis in this area has had major methodological limitations.2 Their analysis uses an impressive array of mental health measures to probe the “compensation effects”. However, several aspects of the study design raise questions. First, with very few exceptions, the transport accident compensation scheme in Victoria covers all injuries arising from transport accidents. It is therefore unclear how a quarter of patients in the non-compensable group could have suffered injuries due to motor vehicle accidents (MVAs) yet have fallen outside the scheme. Second, the purpose of control variables in a multivariate model is to address potential confounders of the relationship between the predictor of interest (MVA compensability) and the outcomes (measures of health status at 24 months). Using significant univariate differences between the predictor of interest and other covariates as the basis for selecting control variables is statistically inappropriate, and this approach may have affected the results of the regression analyses. Third, a key study finding is that significant differences in health outcomes were detected between MVA-compensable and non-compensable patients at 24 months after injury. These then “all but disappeared” when the non-compensable group was altered by shifting three patients who had accessed Transport Accident Commission compensation over to the MVA-compensable group and dropping 54 patients who had accessed “other forms of compensation”. The result casts the spotlight on the removed group. It suggests that their mean health status at 24 months was relatively high. But who were they? Little information is provided, other than that nearly two-thirds (36/57) had private health insurance and were dropped for this reason. (In our view, private health insurance should not be construed as compensation, because policies tend to be highly selective about services covered and generally do not provide payment for lost income or non-economic losses.) Another possible explanation, not addressed, is that with only 88 patients left in the non-compensable group, the multivariate analyses lacked power to find differences. The relationship between compensation availability and injury recovery is complex. Policy interest in the relationship looks set to increase in the next few years, as the federal government explores the merits of a national disability scheme.3,4 In this environment, the need for rigorous research and reliable findings will be greater than ever. O’Donnell and colleagues’ welcome contribution to the evidence base should stimulate further debate about how best to disentangle the effects of injury compensation systems on the health outcomes of Australians who call upon them.

David M Studdert · Harold Luntz · Genevieve Grant

To the Editor: As noted by O’Donnell and colleagues,1 there is a growing body of evidence suggesting that provision of compensation is associated with poor recovery after injury. Most of this evidence arises from international workers compensation jurisdictions. However, two recent Victorian studies have examined health and work outcomes in compensable and matched non-compensable groups after transport injury.1,2 Despite examining broadly similar patient groups and using broadly similar outcome measures, the two articles reach very different conclusions. There has been substantial community reaction to these findings. Gabbe and colleagues’ suggestion that compensation is associated with poor recovery2 provoked public criticism of its methodology from the Law Institute of Victoria, and a prominent plaintiff legal firm released a public statement3 3 days after publication of the study by O’Donnell et al. There is a disconnection in conceptualisation of this issue between the research community and those involved in compensation regulation and policy. Researchers are focusing on the question “Does compensation lead to poor health outcomes?”, while the more nuanced policy question attracting the attention of many injury compensation regulators is “Which, if any, aspects of the compensation scheme have a positive or negative impact on health, vocational and social outcomes?”. Close inspection of the published literature suggests that there are individual components of compensation systems that may have a negative impact on outcome, including the provision of payments for pain and suffering4 and the provision of income benefits.5 There are also examples of compensation organisations acting to improve outcomes via their broader remit as government regulators. For example, the Transport Accident Commission was a major driver of the reorganisation of the Victorian state trauma system, which has resulted in a significant reduction in mortality after road trauma.6 O’Donnell and colleagues1 note the complex relationship between compensation and health outcomes, with particular reference to patient characteristics. The compensation schemes themselves are also highly complex. However, there has been very little research effort directed towards identifying the impact of specific scheme components on patient outcome. In Victoria, the two major injury compensation regulators have funded the Institute for Safety, Compensation and Recovery Research to address this issue. This level of interaction between policymakers and researchers is needed to improve outcomes for those injured in transport- and work-related accidents.

Alex Collie · Niki Ellis

In reply: We thank the authors of the above letters for their comments. Our response will focus only on the major themes raised. We note the concerns about excluding people with private health insurance. The issue here is not whether having health care costs met by private health insurance is the same as having motor vehicle accident (MVA) compensation entitlements. Rather, it is whether access to private insurance payments for health care is the same as not having any compensation at all. We argue that injury patients with private insurance have access to a broader range of health care services and providers than those in the public system, and can access these services more quickly because they avoid long public sector waiting lists. The suggestion that patients who are dependent on public health care in the 2 years following injury (non-compensable patients) receive the same health care as those who have private insurance is unjustified. Most studies to date have not considered other schemes such as private insurance, ignoring the potential impact they may have on health outcomes. We recognise that there may be demographic differences between patients who are involved with other schemes such as private health insurance, and these factors may contribute to outcomes. In noting the inherent limitation in this approach, we nonetheless argue that there are also limitations to including these patients, and therefore an analysis that excludes privately insured patients is a valid addition to the literature on compensation. In response to the point raised by Gabbe and colleagues, we note that our analysis showing that “stressful interaction with the compensation agency” accounted for variance in anxiety scores was designed to investigate potential mechanisms that may explain why anxiety was higher in the MVA-compensable group. We did not conduct the analysis to argue that this group was not more anxious than the non-compensable group. They were more anxious. Glozier and Large were concerned that differential attrition may affect comparisons between the two groups. To clarify, there were no significant baseline differences between completers and non-completers on any measure. Their second issue relates to the removal of patients with private health insurance from the analyses. To clarify, we removed anyone who indicated at 24 months that they had accessed private or other forms of compensation, regardless of their original compensation classification. Studdert and colleagues were concerned that we used univariate differences to identify control variables. We adopted this process to replicate the statistical methodology used by Gabbe et al,1 in an attempt to replicate their findings. In conclusion, our study illustrates the complexity of compensation research and the importance of carefully defining populations — a point that has not yet been adequately addressed. Indeed, a recent review of the literature argues that most compensation research is methodologically limited.2 We agree that there are limitations to our methodology, as there are in previous studies, and recognise that conducting this kind of research is inherently difficult. We welcome the establishment of the Institute for Safety, Compensation and Recovery Research, noted by Collie and Ellis, and its support of this challenging and complex research.

Meaghan L O’Donnell · Mark C Creamer · Richard A Bryant · Alexander C McFarlane · Derrick Silove

To the Editor: I would like to clarify some of the statements made about the National Coroners Information System (NCIS) in the article by De Leo and colleagues.1 The article cites a study conducted by the NCIS concerning the presence of intentional self-harm determinations in coronial findings. Unfortunately, De Leo and colleagues did not note that this study was an internal and informal review of a small random sample of findings conducted by the NCIS that examined disparities between codes assigned on the NCIS and coronial findings. On the basis of this limited study, I would not endorse the statement (which was attributed to me) that “nationally, 29% of coroners omit reference to intent”. Further, it is important to note that NCIS staff do not assign intent codes on entries in the NCIS, and that this coding is performed by clerks in each of the coroners’ offices. The statement in the article that “the NCIS judged 111 (39%) as involving intentional self-harm” is therefore misleading, and should instead have indicated that the coding on the NCIS showed 111 deaths (39%) as involving intentional self-harm. I acknowledge that De Leo and colleagues did not intend to mislead readers as to the work of the NCIS.

Jessica D Pearse

To the Editor: I applaud the Medical Journal of Australia’s recent attempt to increase the evidence base of complementary medicine.1 However, it is disappointing that the Journal’s idea of doing so seems to be to import the same dogmatic and misinformed debate currently occurring in the United Kingdom. Ernst makes little secret of his antihomeopathic agenda and engages in some “cherry picking” of his own, neglecting, for example, to mention the substantial methodological criticisms of some of the references he chooses to use to support his points.2 Further, expert testimony at the British House of Commons Science and Technology Committee’s evidence check on homeopathy identified 24 condition-based systematic reviews and meta-analyses on homeopathy, of which nine were positive, five were negative and 10 were inconclusive.3 As a system of medicine, this compares more closely with the evidence base for conventional medicine than many would care to admit.4 It has also long been observed that the complex and individuated nature of complementary therapies — and many conventional therapies, for that matter (including many surgical and psychological interventions) — makes clinically relevant evaluation with a placebo-controlled trial difficult.5 Cochrane reviews may certainly be “the best” at reviewing the trials, but this means little if those trials were not an appropriate evaluation tool in the first place. Rarely do these trials reflect the real-world settings in which patients, medicines and practitioners exist. The challenge is not simply to be better than placebo, but to produce the largest clinical effect possible in a real-world setting. In his article,1 Ernst himself seems to acknowledge the potential broader real-world benefits that patients receive from homeopathic treatment, as confirmed by observational data,6 yet seems inclined to focus only on reductionist approaches to evaluation that are well known to be ill suited to homeopathic research, or focuses on the implausible nature of the medicine itself. We need to take a different approach and work out why it is that patients who choose to use homeopathy get better (as they quite often do). To do this, we need not just more basic and clinical research, but more health services and public health research on homeopathy — reviewing the reviews adds little if there is simply not enough to review in the first place. Throwing out the baby with the bathwater helps no-one, least of all the patient. And the patient, not ideology, is what it should be all about.

Jon L Wardle

In reply: Wardle’s letter raises several points that deserve comment. Wardle calls me dogmatic, misinformed and antihomeopathic. Such ad hominem attacks hardly promote a rational debate. When I started my job of scrutinising homeopathy 17 years ago, I was pro-homeopathy1 — I once worked in a German homeopathic hospital — and became more sceptical as the evidence base for homeopathy became more clearly negative.2 This, it seems to me, is the opposite of dogmatic. Wardle cites the report by the House of Commons3 in the United Kingdom and claims that it “identified 24 condition-based systematic reviews and meta-analyses on homeopathy, of which nine were positive . . .”. In truth, it was a submission from homeopaths to the House of Commons that made this statement. The report itself found no positive evidence for homeopathy and even criticised how the homeopaths tried to mislead the inquiry.3 Wardle also thinks that clinical trials are “ill suited” to evaluate homeopathy because homeopathy is “complex and individuated” and clinical trials “rarely . . . reflect the real world”. The notion here is that, if the scientific method does not support our belief, it must be the former rather than the latter which is at fault. Adopting this attitude would take us right back into the Dark Ages. After discussing these issues for 17 years, I have the impression that most homeopaths are in favour of rigorous, reductionist science — insofar as it generates the results they want. Whenever this is not the case, they point to observational studies that are wide open to bias and confounding, and therefore show us precious little. Finally, Wardle seems to imply that homeopathy works because patients like it and that this is what truly helps patients. The truth is that medicine has made huge advances only since we buried this attitude. It is time now that proponents of homeopathy do the same — not to conform with a dogma, but because patients would live longer and healthier lives.

Edzard Ernst

To the Editor: At the risk of appearing self-serving, we refer to our recently published article in which we recommended that there should be “greater uniformity in [food] packaging design, colour and descriptions”.1 In light of the Australian Government’s recent mandate on plain packaging for tobacco products,2 we can see no reason why this should not be extended to processed foods possessing no redeeming nutritional qualities. This would include soft drinks, potato chips, a great many of the so-called foods offered as replacements for fruit in children’s school lunches, biscuits and sweets. We challenge the Prime Minister and Minister for Health to do this, or explain why they won’t.

Bebe Loff · Brad R Crammond

To the Editor: Infections caused by multiresistant gram-negative organisms are difficult to treat. Carbapenems are often used as a last resort but even these are under threat with the emergence of acquired metallo-b-lactamases worldwide, including Australia,1,2 India, China and Europe. We report the first case of a Providencia rettgeri producing the blaNDM-1 (New Delhi metallo-b-lactamase) type of metallo-b-lactamase in Australia. A man from Canberra, aged in his mid 50s, had elective plastic surgery in India in September 2009. This was complicated by a hypoxic brain injury, after which the patient spent 4 weeks in an intensive care unit. He was subsequently transferred to Canberra for ongoing hospital care. A urinary catheter specimen collected on admission in November 2009 showed a heavy growth of multidrug-resistant P. rettgeri and Pseudomonas aeruginosa. The P. rettgeri was resistant to all b-lactam antibiotics, including meropenem, as well as to all aminoglycosides, ciprofloxacin, tigecycline and colistin. The P. aeruginosa was resistant to all antipseudomonal antibiotics except for colistin (tigecycline was not tested as it has low or no antipseudomonal action). The patient was not given antibiotic therapy but the indwelling urinary catheter was changed and contact precautions were put in place. Both organisms were sent for molecular testing, which showed that the P. rettgeri had 100% homology with blaNDM-1.3 The patient cleared the organisms after 2 months, and since then has received ongoing inpatient care in the rehabilitation unit. The first NDM-1 type of metallo-β-lactamase was found in Klebsiella pneumoniae isolated from a Swedish patient who had recent medical contact in India.3 Data from the United Kingdom’s Antibiotic Resistance Monitoring and Reference Laboratory suggest that isolates with the NDM-1 enzyme have recently been repeatedly imported to the UK from the Indian subcontinent. There may now be circulation of these resistant isolates in the UK because some infected patients have no identifiable overseas links. Hospitals have been urged to be vigilant for multiresistant gram-negative bacteria in patients with recent hospital contact in the Indian subcontinent as well as the Eastern Mediterranean.4 Identification of an Enterobacteriaceae organism carrying blaNDM-1 is very concerning. No antibiotic may be available to treat patients who develop serious infection with such organisms, and there is the added concern regarding cross-infection in health care facilities. The plasmid carrying blaNDM-1 also contains genes that confer resistance to several other antibiotics.3 It appears likely that, in the near future, the NDM-1 enzyme will become a very successful metallo-b-lactamase globally. Patients infected with multiresistant gram-negative bacteria have entered the “post-antibiotic era”.

Geethanie A T P Fernando · Peter J Collignon · Jan M Bell

To the Editor: The arrival of national registration of health professionals on 1 July 2010, and consequential amendments to state and territory legislation, overcomes registration and recognition complexities currently facing Australian medical practitioners who wish to practise in multiple jurisdictions. However, national registration stops short of removing current inconsistencies among state and territory laws in various areas of medicine — including those that regulate prescription of Schedule 8 (S8) medicines. Despite recognition of S8 restrictions in every part of the country, medical practitioners potentially need to have a working knowledge of up to eight separate sets of controlled substances laws when prescribing S8 medicines. Imagine a general practitioner providing a morphine prescription with repeats to a couple caravanning around Australia for 3 months. That prescription may need to comply with the laws of each jurisdiction as to what details need to be included on the prescription form. Getting it wrong could result in significant delay and inconvenience for the holidaymakers if the pharmacist refused to dispense the prescription because it didn’t comply with local laws. For example, in South Australia a prescription for a drug of dependence must not be dispensed by a pharmacist if the patient’s date of birth is not included on the prescription form1 — yet there is no similar requirement in New South Wales, Tasmania, Victoria or the Australian Capital Territory.2-5 Conceivably, through unawareness or habit, a prescriber in one of the latter jurisdictions might omit a patient’s date of birth from a relevant prescription form, leading to problems for the patient when travelling in other states. Granted, such a situation could be corrected with a few phone calls or faxes, but in a busy practice, who has time to be repeating tasks? National regulation of health practice (achievable through referral of powers or harmonising state and territory laws) — in this case, the prescription of S8 medicines — would logically complement national registration and help maximise anticipated benefits after 1 July 2010.

Colin M Brown

To the Editor: We report the case of a 19-year-old female veterinary student who presented with a 2-week history of a rapidly growing mass on the lateral side of the deltoid area of her left arm. On examination, the mass was soft, freely movable, slightly warm and non-tender. There was no regional lymphadenopathy. The patient was afebrile, with no signs of systemic illness. Ultrasound showed a 3.2 × 2.3 × 1 cm, low-echogenic, lobulated lesion in the subcutaneous fat, with surrounding increased echogenicity suggestive of inflammation. Magnetic resonance imaging showed a poorly defined lesion throughout the deep and superficial fascia, with infiltration into the underlying deltoid muscle (Box 1). A provisional diagnosis of sarcoma was made, but an ultrasound-guided core biopsy sample showed non-necrotising epithelioid granulomas (Box 2). Two weeks after presentation, the patient noticed a smaller raised lump, about 1 cm in diameter, on the volar aspect of her left forearm. Four months before presentation, following negative results of both Q fever serological testing and a Q fever skin test administered on the volar aspect of her left forearm (at the site of the smaller lump), the patient had received a Q-VAX (CSL, Melbourne, Vic) vaccination in the left deltoid (at the site corresponding to the larger mass). Three months after this, and about 3 weeks before the deltoid mass first appeared, the patient received a Fluvax (CSL, Melbourne, Vic) influenza vaccination at the same left-deltoid site as the Q fever vaccination. This raises the possibility that the influenza vaccination may have been associated with the subsequent Q fever granuloma reaction. Q fever, a disease caused by the zoonotic rickettsial organism Coxiella burnetii, is an occupational hazard for many Australians in animal-related professions. The disease is characterised by an acute, self-limiting febrile illness, with pneumonia and hepatitis being infrequent complications. In 2001, an Australian national Q fever vaccination program was initiated, which led to a 50% decline in the incidence of Q fever.1 Side effects are normally rare and minor; during 2001–2004, only 86 adverse reactions were reported from about 49 000 vaccinations.2 Development of non-necrotising granulomas following Q fever vaccination is uncommon, with only six cases previously described.3-5 To explore the possibility of a causal association between influenza vaccination, Q fever vaccination, and development of a granuloma, we traced four of these six patients with Q fever granuloma by contacting the authors of the previous reports. One patient had been vaccinated against influenza 2 months before the time of Q fever vaccination, and another had influenza vaccination afterwards (as in our case). Of the five cases (including our case) for which clinical history was available, three had a temporal association between influenza vaccination and Q fever vaccination, followed by the development of the non-necrotising granuloma. The indurated lesion at the separate Q fever skin-test site on the volar forearm, found in two other patients5 as well as ours, supports the notion of a systemic immune reaction rather than simply a local reaction at the vaccination site. The Naranjo score in this case was 7, indicating a “probable” adverse drug reaction. Although the granuloma was self-limited in all known cases, this case shows that there is significant risk of misdiagnosis on clinical grounds. It seems prudent to be aware of the possible association between these two vaccinations. Magnetic resonance images of lesion Axial (top) and coronal T1-weighted fat-saturated post-contrast (bottom) images demonstrate a poorly defined enhancing lesion (red arrows) in the subcutaneous fat, which superficially infiltrates and extends in a plaque-like manner in relation to the underlying deltoid muscle (white arrows). 2 Ultrasound-guided core biopsy sample of lesion The core biopsy of fibroadipose tissue shows numerous well formed non-necrotising epithelioid granulomas (arrows). The granulomas are composed of epithelioid histiocytes and a few multinucleated giant cells, surrounded by numerous lymphocytes (a mixture of B and T cells, with a greater proportion of T cells). (Original magnification × 100; haematoxylin–eosin stain.)

Deborah Burnett · Leslie Burnett

To the Editor: When I looked at the cover of the 5 April issue of the MJA, I feared finding another article focused on discrediting the prostate-specific antigen (PSA) test. Instead, I congratulate the authors, and the Journal, for presenting one of the rare balanced articles on this topic.1 Denham and colleagues called for an end to taking sides in the debate over PSA testing, and focused instead on helpful guidance. The problem is not whether PSA helps us find prostate cancer, but that we lack clinical tools for deciding which patients would benefit from aggressive treatment. However, the authors point out that there are two important tools that can help with this decision: low free to total PSA ratios, and rapid PSA doubling time (< 3 years).1 The PSA test is now one of the most sensitive, precise and highly standardised immunoassays in the clinical laboratory. The difficulty does not lie with the measurement but with its application. The Royal College of Pathologists of Australasia, together with the Urological Society of Australia and New Zealand, recently produced a monograph that discussed the appropriate use of the PSA test.2 It emphasised using age-related cut-offs for PSA levels, the free to total PSA ratio, and the calculation of PSA doubling time as modern tools to achieve optimal benefit from the test. The Australian Medicare Benefits Schedule (MBS) was changed in May 2009 (following a suggestion from the Urological Society of Australia and New Zealand) to improve utilisation of free to total PSA ratios. The Box indicates the per capita request rates of PSA testing (MBS item number 66655) and free to total PSA ratios (MBS item number 66659) from May 2009 to February 2010. Consistent with Denham et al’s observation that there are regional differences in the attitudes to prostate cancer,1 there is a twofold variation in PSA requesting and a sevenfold difference in free to total PSA ratio requesting across the Australian states. Furthermore, the relationship between the two tests is, if anything, inverse, suggesting that increased use of PSA testing is less commonly followed up by modern tools such as free to total PSA ratio. As a chemical pathologist, the appropriate clinical use of the PSA test has been a career-long concern of mine.3 Even though the discoverer of PSA has labelled the test a public health disaster,4 recent “case–controlled” studies using PSA in an outdated approach (without free to total PSA ratios or doubling times) have shown a marginal benefit for screening.5,6 The indiscriminate use of PSA testing can be helpful to some but disastrous for others. Modern PSA tools may significantly improve management beyond these marginal effects. As always, the value of medical investigations lies in how intelligently we use them. Average per capita request rates of PSA testing* and free to total PSA ratio,† May 2009 – February 2010 PSA = prostate-specific antigen. ACT = Australian Capital Territory. NSW = New South Wales. NT = Northern Territory. Qld = Queensland. SA = South Australia. Tas = Tasmania. Vic = Victoria. WA = Western Australia. * Medicare Benefits Schedule (MBS) item number 66655. † MBS item number 66659.

Kenneth A Sikaris

To the Editor: Many of the arguments put forward by Bain and colleagues about modelling hospital occupancy1 are true in an academic sense. We agree that many acute care models use simplified inputs and outputs, without accounting for rapid daily fluctuations in occupancy. Occupancy definitions are often misleading and subject to gaming. We agree with the work by Bain, Taylor and others that highlights as a problem “the inability to move patients from the [emergency department] to a ward” and that hospitals should “engage procedures to free inpatient beds well in advance” of access block occurring.2 The capacity of a hospital must have flexibility to deal with demand fluctuations.1 However, we do not agree that the “85% occupancy” figure for optimal efficiency is a candidate for myth status. This threshold is used in various systems around the world. Efficiency is a well recognised concept in queuing theory and depends on setting the utilisation (occupancy) rate at a level where the costs of “underutilised” resources (beds, staff not fully used, etc) are shown to acceptably match the “costs” from delayed care (mortality, morbidity, economic, political, prolonged stay, etc). In particular, quoted occupancy figures often look better than reality because of unopened, unusable beds or data manipulation to improve reported occupancy. The more complex and variable the inputs into a system with multiple competing queues (such as hospitals, where acute admissions may be > 70% of the workload), the greater the need for additional capacity to avoid deferred service. Major acute care hospitals show dramatic daily fluctuations in bed use, with changes between low and high occupancy rates in the order of 15%–20%.3 As inputs and variation increase, the likelihood of marked performance deterioration increases exponentially. Real life (or “clinical modelling”) shows that reported occupancies around or above 85% routinely lead to loss of access to care. No benefits of emergency department or hospital overcrowding have been reported in the medical literature, only harm.4 Relying on developing complex mathematical models before accepting that we are already beyond the acceptable “efficient” occupancy of our current health systems is not a realistic way forward, when patients are dying due to the lack of available appropriate beds.4 Garling, in the overview of his report referred to by Bain and colleagues, states “my recommendations should make more beds available and reduce access block”, but that even with more efficient bed usage “the addition of 350 beds each year” will be required.5 Action is required now and shouldn’t be delayed or subverted by searching for perfect mathematical models. Our hospitals have demonstrably inadequate capacity, resulting in overcrowding with devastating effects. As we have repeatedly stated, it’s all about available beds.

David Mountain · Daniel Fatovich · Sally McCarthy

To the Editor: In her editorial, Christiansen states that the 2010 edition of Therapeutic guidelines: antibiotic (version 14) is very likely to recommend 24 hours of antibiotic prophylaxis for cardiac surgery,1 rather than the present regimen, which recommends that patients having routine cardiac surgery be given a large dose of cephazolin at induction, with a second dose if the operation is prolonged for more than 3 hours, and no doses after surgery.2 The only evidence presented in support of this proposed change is a single randomised controlled trial that purported to show a higher rate of surgical site infections after a single dose of cephazolin, than after a prolonged, multidose regimen.3 The study was flawed, for two reasons. First, cephazolin was not given until 20–30 minutes after induction of anaesthesia, which is likely to have been too late, because, as Christiansen points out, β-lactams should be given 30–60 minutes before incision.1 Second, the trial was analysed on a per-protocol, rather than an intention-to-treat basis, and 189 of the 1027 participants (18%) were excluded, so the findings may be seriously biased.4 Three other trials have compared one or two doses of a cephalosporin with multiple doses of the same antibiotic in patients having cardiac surgery; none found that multiple doses were superior, although all three were small studies with faults in their design.5-7 In 1998, McDonald and colleagues published a detailed review of single versus multiple doses of antimicrobial prophylaxis for major surgery. The analysis was in response to a suggestion by Christiansen and others that single-dose antibiotic prophylaxis may be inadequate for patients undergoing vascular surgery.8 McDonald and colleagues pointed out that the recommendation for single-dose surgical prophylaxis in Therapeutic guidelines: antibiotic (version 13),2 is based on microbiological first principles, published studies reporting efficacy, convenience of administration, reduced antibiotic resistance and toxicity, and relatively low cost. Their careful analysis of 28 randomised trials, in which the same antimicrobial was used in each arm, showed no advantage from the administration of multiple doses; the odds ratio for infection was 1.06 (95% CI, 0.89–1.25). There is no microbiological reason to suppose that the crucial interaction between contaminating bacteria and the prophylactic antibiotic in the heart is any different from that in the lung, biliary tree, uterus, bowel, prostate or bone.8 In the absence of such evidence, there is no sound reason to change the current, long-standing Therapeutic guidelines: antibiotic2 recommendation.

Frank Shann

In reply: Professor Shann raises three issues regarding the recommendation for 24 hours’ prophylaxis for cardiac surgery. First, he states that the trial1 on which this recommendation was made was flawed, because “cephazolin was not given until 20–30 minutes after induction of anaesthesia, which is likely to have been too late”. The trial included patients having coronary artery surgery and/or cardiac valve replacement, and, for these procedures, the time between induction of anaesthesia and surgical incision is about 60–75 minutes, as patients require the placement of intravenous lines and preparation for coronary artery bypass surgery. Antibiotic administration is recommended 30–60 minutes before incision, thus, administration 30 minutes after induction provides optimal serum concentrations at incision, in patients undergoing cardiac surgery. Second, the per-protocol analysis is perhaps less than ideal, but the demographics, clinical characteristics and operative data were comparable for the patients included in the analysis. Third, as stated by Professor Shann, the three earlier studies2-4 were either very small or flawed in design. The McDonald systematic review5 included 28 studies, only two of which were on cardiac surgery, both being the earlier flawed studies2,3 quoted above. The Therapeutic guidelines: antibiotic review process involves a rigorous, evidence-based assessment with input from experts in the field. The medical community of Australia can have every confidence that the recommendations made are current and evidence-based.

Keryn J Christiansen

To the Editor: I agree with the claim by Richardson that “the overlap between [primary care and emergency department (ED)] services is not as important as many have claimed” and that “‘primary care patients’ and ‘ED [Australasian Triage Scale] category 4 and 5’ patients are not interchangeable”.1 A review of the literature — especially from New Zealand — would show there are considerable differences between patients who attend the two types of services. For example, a comparison of patients with asthma attending either a Wellington after-hours medical centre or an ED service located only 800 metres away2 found that the after-hours medical centre was more likely to see younger patients who live further from the service, are given repeat medications, and are referred back to their general practitioner. In contrast, the ED patients were less likely to be referred by a GP and more likely to be admitted to hospital with asthma than patients attending the after-hours centre. Thus, the two services differed in terms of their clinical policies (repeat prescribing and referral) and patients’ demographic characteristics (age, place of residence). I applaud Richardson for highlighting the powerful effects of hospital policies on the behaviour of people outside hospital walls by saying, “it is not the so-called primary care patients who are blocking ambulances from offloading — it is the ‘access block’ patients waiting for beds on the inpatient wards who are inappropriately occupying ED space and staff time”. This claim has nothing to do with the kind of patients who attend primary care services, but more to do with the influence of management policies arising from within hospitals on patient flow from primary care. It confirms research in New Zealand demonstrating how hospital policies (on advertising their services) can have powerful contradictory effects on attendance at EDs. In some cases, people have been subjected to hospitals advertising the clear message that people should attend the ED when they should be seen in primary care instead; and in other cases, people are dissuaded from attending the ED when they are subjected to advertisements about the poor choices people make to attend a hospital. In each case, it is the hospital policy that determines the direction of flow, not the patients in primary care.3-5

Marjan Kljakovic

To the Editor: A recent commentary from the Editor of the Journal1 raises the health-funding dilemma facing current and future Australian governments. All stakeholders in the health industry need to dispassionately scrutinise the role of current models in perpetuating inefficient or socially discriminatory patterns of care. Health care economics is indivisible from the tendency of the broader economy to sustain growth or create disparities, and comparative analysis of systemic economic policies informs the divergent evolution of health systems. The United States, epitomising the free market paradigm, combines a high gross national product with a high poverty rate and significantly unequal income distribution.2 Per capita health care expenditure and its annual rate of increase are comparatively high.3 This is juxtaposed with one of the highest infant mortality rates in the developed world,3 as well as significant racially related variations in health indices. The Scandinavian societies, particularly Sweden, epitomise the benefits of a social welfare model that maintains low unemployment, relatively low income disparity, and advanced technology, while maintaining a healthy private sector.2 The Swedish health care system, which maintains best practice health indices, is characterised by administrative devolution, combined taxation and insurance-based funding, high equity of access, guaranteed maximum primary-care waiting times, and annually capped out-of-pocket expenses.4 Although it had one of the highest per capita health expenditures in the 1980s, its annual rate of increase is one of the lowest in the OECD.3 Since 1975, Australian economic and social policy has increasingly shifted towards a free-market orientated system, comparable with that of the US and United Kingdom rather than the more mixed economies of most European countries or the social welfare economies of Scandinavia. In this context, Medicare is an anomalous relic, which, because of inadequate funding, has struggled to contain patients’ out-of-pocket expenses and maintain equity. The most successful health care systems have a relatively small private sector, limited fee-for-service provisions, and smaller income disparities between health care workers. In contrast, Australia’s hybrid model has a significant private/entrepreneurial component based on a fee-for-service structure that is driven by market forces rather than needs analysis. As Medicare increasingly withers due to neglect, out-of-pocket expenses will continue to rise and health equity will diminish, but sections of private medicine, fuelled by demand from the more affluent, will continue to thrive. As we confront the unpalatable likelihood that the market-driven private sector is a major cause of increasing per capita expenditure, we shall have to justify the viability and equity of the dominant fee-for-service model.

Jeremy W Butler

To the Editor: Observing jugular venous pressure (JVP) is central to cardiovascular examination. Lewis, in 1930,1 was the first to report the use of the external jugular vein as a manometer for recording pressure in the right atrium. Unfortunately, some textbooks on clinical examination and many clinical teachers incorrectly state that the external jugular is unreliable for measuring JVP and that only the internal jugular should be used. The problem with this is that the internal jugular vein is located deep within the neck, where it is covered by the sternomastoid muscle and is therefore not usually visible. Lewis used the sternal angle as a reference point, presuming that it lay 5 cm above the centre of the right atrium in all positions of the patient between lying and sitting. A recent study using computed tomography to examine 160 patients noted that the median vertical distance between the sternal angle and the mid right atrium was 5.4 cm, thus confirming that Lewis’s estimate of the sternal angle in relation to the right atrium was correct (bearing in mind that adults are taller than they were in the 1920s).2 The mean right atrial pressure is the mean of the peak and trough of the external jugular wave above the sternal angle expressed in cm H2O. Over the past few decades, several studies have confirmed that the original findings of Lewis were correct: there is no significant difference in JVP whether it is measured using the internal or external jugular vein, and the external jugular pulse accurately reflects directly measured right atrial pressure.3 In a study of 52 patients with chronic congestive heart failure who had right heart catheterisation, elevation of the JVP showed 57% sensitivity for a raised pulmonary capillary wedge pressure (≥ 18 mmHg) and 93% specificity for non-elevation of JVP, corresponding with a capillary wedge pressure of ≤ 18 mmHg. If elevated JVP was inducible as well, sensitivity increased to 81% and specificity dropped to 80%, with a predicted accuracy of 81%.4 Generations of frustrated medical students and doctors who have stared intently at their patients’ necks awaiting that elusive flicker of the internal jugular pulse have been overlooking an accurate source of clinical information — namely the pulse in the external jugular vein. Lewis was right 80 years ago: measuring the external JVP is a valuable clinical tool and should be practised frequently.

David M Colquhoun · Glenn Jenkins

To the Editor: We describe here some of the difficulties in managing and investigating outbreaks of viral respiratory infection in aged care facilities (ACFs) in the context of an influenza pandemic. This adds to the previous report on logistics in a hospital setting.1 On 12 June 2009, NSW Health received a call from a surveillance officer in a remote town regarding a possible pandemic (H1N1) 2009 influenza outbreak in an ACF. On 9 June, a 77-year-old female resident had become unwell, without specific symptoms of influenza-like illness. From 7 to 10 June, nine of the other 27 residents developed influenza-like illness. On 10 June, nasal swabs were taken from the 10 unwell residents by the local general practitioner for influenza nucleic acid testing (NAT). On 12 June, the index case tested positive for pandemic influenza, while the other residents tested negative. Due to concern that there might be a pandemic influenza outbreak in the facility, the index case and the nine residents with influenza-like illness were given oseltamivir (75 mg twice a day for 5 days) from 13 June; the other 18 residents and the 27 staff were given oseltamivir prophylaxis (75 mg daily for 10 days). A formal outbreak investigation and further laboratory testing (NAT, serological testing) revealed a dual outbreak dominated by rhinovirus (10 cases), with two cases of pandemic influenza and one case of untyped influenza A. All 28 residents and 26 of the 27 staff had received seasonal influenza vaccine in early 2009. This outbreak illustrates that more than one respiratory virus may co-circulate in ACFs during winter outbreaks of respiratory infection. We followed Department of Health and Ageing policy guidelines for oseltamivir use in ACFs2 and the facility was closed to visitors from 12 to 18 June. However, as all residents had received seasonal influenza vaccination, and given that older people are generally at lower risk of pandemic (H1N1) 2009 influenza,3 we could have had a higher threshold for oseltamivir use. The total estimated cost of treatment and prophylaxis was $2750 (55 residents and staff at $50/person) for oseltamivir alone. Co-infection with respiratory viruses may be more common than thought in ACFs; a recent Canadian study found two and three different pathogens in 15% and 4% of respiratory infection outbreaks, respectively, from a total of 83 outbreaks (of which 91% occurred in long-term care facilities).4 If many ACF outbreaks have more than one respiratory virus involved, laboratory investigations should take a multiplex approach that covers common respiratory viruses. As many patients as practical (at least five) should be swabbed and tested to guide treatment, prophylaxis and other investigations. Community influenza surveillance should ideally include information on sensitivity to oseltamivir, and on other circulating respiratory viruses.

Gulam Khandaker · Bridget Doyle · Dominic E Dwyer · Robert Booy

To the Editor: It is a naïve public health physician who predicts ahead of time how many people will die in a disease outbreak. Such doctors have short careers. What Collignon calls the “wrong and exaggerated” expert predictions1 of mortality from the recent influenza pandemic are based on the numbers that the World Health Organization advised governments to use in planning for pandemics.2 They are derived from a sensible calculation: plan for a situation considerably better than the 1918–1919 pandemic but somewhat worse than the 1957 or 1968 pandemics. The problem in Australia is not so much the pandemic plans produced through the time-honoured process of ad-hoc, temporary federal government committees for implementation by multiple, variously organised state and territory authorities. The real problem is producing a consistent, flexible response to any developing national infectious disease emergency. No other nation tries to do that without having a national authority, made up of full-time professionals with a fair degree of independence from the political process. The United States has its Centers for Disease Control and Prevention (http://www.cdc.gov); the United Kingdom its Health Protection Agency (http://www.hpa.org.uk); and, perhaps the most pertinent example, Canada has its Public Health Agency (http://www.phac-aspc.gc.ca), established in the aftermath of the SARS (severe acute respiratory syndrome) outbreak. The European Union has set up a supranational European Centre for Disease Prevention and Control (http://www.ecdc.europa.eu). A plan can only ever hope to put in place all the resources needed for a response, but a flexible, consistent, science-based and targeted national response to infectious and other health emergencies requires a professional national authority.

Rodney C Givney

To the Editor: The recent clinical update by Chakera and colleagues highlights the problems and adverse patient outcomes that occur when current and prior diagnostic images are not accessible during the clinical care process.1 While the article describes a locally crafted, tactical, information technology (IT) solution, it fails to mention that much work has been done internationally to create a standards-based, scalable architecture for image and document exchange. This work has been done by Integrating the Healthcare Enterprise (IHE) (http://www.ihe.net), a global collaboration between health care equipment suppliers, IT experts and clinicians. The aim of the collaboration is “to improve the way computer systems in healthcare share information”. The profile for cross-enterprise document and image sharing is known as XDS-I. XDS-I defines how to use established health care and IT standards (eg, the Digital Imaging and Communications in Medicine [DICOM] and Health Level 7 [HL7] standards2,3) to facilitate secure exchange of health care information, including images, between health care institutions. Information exchange is independent of the hardware and software in place at the participating institutions, and system integration using XDS-I supports user needs, including security and privacy, while streamlining workflow. Using a single technical approach, implemented at a regional or state level, diagnostic images can be exchanged, along with documents such as radiology and laboratory reports, discharge summaries, and even general practitioner care plans. Providers can, with patient permission at the time of care, access such documents via secure internet connections. The solution developed by Chakera and colleagues covers Western Australian public hospitals and parts of the private sector. XDS-I is a platform that also allows image sharing between the public and private sectors, regardless of the picture archiving and communication system adopted by participating practices or hospitals. Many of the operational problems identified in the article by Chakera and colleagues (consent, staff time costs and manual processes) have been addressed in the IHE XDS-I profile. While projects such as the pilot program by Chakera and colleagues are useful learning exercises, locally crafted single-vendor solutions (even those using industry standards) are rarely scalable to broader usage. We strongly commend the IHE XDS-I model to everyone considering image exchange systems in Australia.

Nicholas J Ferris · Philip J Dubois · Christopher Lindop · Vincent B McCauley · Peter A MacIsaac

To the Editor: Alarm about the dangers of computed tomography (CT) scans1,2 is unjustified. The only hard facts about bio-harm from ionising radiation come from the 1945 atomic bomb explosions, which emitted very large amounts of radiation. Bio-harm from low-dose medical radiation has never been confirmed; the claim is based on backward extrapolation of data on radiation doses from the Japanese atomic bombs, which were orders of magnitude greater than doses in diagnostic radiation. The resultant linear no-threshold theory, which postulates that there is no safe radiation dose, remains unproven. Radiation protection authorities use this model because it is expedient, if unverified and overly conservative. Those who treat it as dogma forget that it remains a theory, and any derived calculations are subject to large uncertainties. Radiation scientists question its validity,3 and many regard the estimated risks as grossly exaggerated or negligible.4 The Health Physics Society has stated that the risks to health from radiation doses below 100 mSv are either too small to be observed or non-existent.5 The theory is also challenged by evidence that low-dose radiation is actually beneficial and protects against the effects of large-dose radiation by inducing DNA repair enzymes. A study of 407 000 nuclear shipyard workers and another of 7800 Russians exposed to low-level radiation from the 1957 Mayak nuclear facility accident showed that the exposed groups developed significantly less cancer than their unexposed controls.4 The lifespan of British radiologists over the past century has exceeded that of any other control group.6 Figures quoted in the media for cancer attributable to medical radiation are theoretical calculations and have never actually been observed. They are as “real” as estimated cancer rates due to mobile phones and power lines. Newspaper claims such as “More than 400 new cases of cancer a year in Australia are attributable to diagnostic radiology”2 are alarmist and misleading — they disguise the fact that their figures derive from an unproven theory, not from observations. Ironically, concern about the dangers of CT is rising even as the actual radiation doses involved are falling. A 2010 CT scanner emits 1/20th the radiation of its 5-year-old predecessor. CT coronary angiography can be accomplished today with a dose of less than 1 mSv — equivalent to six chest x-rays or 6 months of background radiation. Patients for whom a CT scan is medically indicated should not be denied one of modern medicine’s greatest benefits because of unfounded fears. The risks of delayed or missed diagnosis or wrong treatment far outweigh the theoretical risk of harm from a CT scan.

Carl M Blecher

To the Editor: We present a case of fatal cocaine overdose in which the drug was contaminated with levamisole, a therapeutic agent known to cause reversible agranulocytosis. The deceased, a previously well woman in her early 20s, was found dead in circumstances suspicious of a drug overdose. The death was reported to the coroner and the autopsy findings were unremarkable, with no evidence of injury or significant natural disease processes. Toxicological sampling of blood revealed a cocaine level in the blood of 4.9 mg/L, as well as the cocaine metabolite benzoylecgonine at a level of 3.4 mg/L. These are lethal levels for cocaine and benzoylecgonine.1 Cocaine was also detected in a nasal swab, and levamisole was detected in the nasal swab and in the blood, as well as in a quantity of white powder found near the woman’s body. The cause of death was given as cocaine toxicity. Illicit cocaine in Australia is generally diluted (“cut”) with a range of innocuous substances, including fructose and sucrose, and less commonly with other drugs, such as lignocaine.2 In this case, levamisole was detected as a contaminant. Levamisole is primarily used as a veterinary anthelmintic, and used uncommonly in humans for rheumatoid arthritis, and as adjuvant therapy to fluorouracil in the treatment of a variety of cancers.3,4 Agranulocytosis is a significant side effect of levamisole, and this has limited its use in humans. The clinical presentation of agranulocytosis is protean, presenting with a spectrum of abnormalities, ranging from a flu-like illness to leukopenia with a potentially fatal outcome.3-5 The mechanism whereby levamisole induces agranulocytosis is unknown, although a strong link with the human leukocyte antigen HLA-B27 and rheumatoid factor positivity suggests a likely genetic predisposition.4 Recent reports from the United States and Canada have highlighted the presence of levamisole in seized illicit cocaine entering these countries, with up to 69% of seized cocaine lots containing levamisole. There has been a subsequent clustering of fatal and nonfatal cases of agranulocytosis in a number of disparate locations.3-5 The reason for contaminating cocaine with levamisole is unknown, although there are indications that levamisole may promote the effects of cocaine by interfering with its reuptake at a synaptic level.4 Although there was no evidence of agranulocytosis in the present case, we highlight the apparent recent appearance of this contaminant in the cocaine supply in Australia, because it has the potential to induce reversible agranulocytosis in people not otherwise obviously at risk for this condition.

Johan A Duflou · Issabella G Brouwer · Shane Darke

To the Editor: The clinical justification for a medical intervention depends on absolute prevalence (p) of a disease or condition in a population and the relative risk reduction (R) that would result from the intervention. These variables determine the “number needed to treat” (NNT) to prevent one occurrence of a disease, according to the formula NNT = [100 ÷ (P × R)], where P and R are expressed as percentages. This principle applies to thromboprophylaxis in medical patients. However, the prevalence of venous thromboembolism (VTE) in hospitalised patients is uncertain. The main justification for medical thromboprophylaxis given on the National Institute of Clinical Studies (NICS) website1 is an unpublished report prepared by the University of Western Australia (UWA) on behalf of the NICS.2 The report noted that 40.8% of all hospital cases of VTE were “medical” rather than “surgical” or “idiopathic” (primary) cases, but the prevalence of VTE (overall or in each subgroup) was not stated. We obtained coded separation data for all multiday admissions to Royal Perth Hospital (RPH) for the most recent 2-year period with complete data (2005–2007). We searched for VTE events with a principal or secondary coding, and classified them as medical or surgical VTE cases according to the definitions used in the UWA report (for medical VTE, “admissions in which a diagnosis of VTE was recorded as a complication or in a diagnostic field other than the principal diagnosis OR admissions with VTE as the principal diagnosis within 3 months of a non-surgical [medical] admission”).2 We acknowledge that the use of prophylaxis during the index admission or any previous admission within 3 months (which we did not measure) may have meant that we underestimated the prevalence of VTE. Prevalence was calculated by dividing the event number by the total number of admissions or by the number of medical or surgical admissions, as required (expressed as a percentage). At RPH over the 2-year period, 805 VTE events (574 in medical and 231 in surgical patients) were observed in 72 991 medical and 29 177 surgical admissions (total, 102 168 admissions). These included 357 pulmonary emboli (44.3%), 207 of which were primary events. Of the 805 VTE events, 312 (38.8%) were medical, 209 (26.0%) surgical and 284 (35.3%) idiopathic (the corresponding proportions in the UWA report were 40.8%, 37.7% and 21.5%, respectively). The overall VTE rate in the medical patient population was 0.79%, but for “medical VTE” as defined in the UWA report,2 the rate was 0.43%. In surgical patients, the rate of VTE was 0.79% in our study, or 0.72% based on the UWA definition. Hence, according to the definitions used by the UWA report, “surgical VTE” is actually more frequent than “medical VTE”. Our study confirms that about 40% of VTE is in medical patients, but that the absolute prevalence is low (0.43%). This is similar to the rate of 0.4% reported in the PREVENT (Prevention of Recurrent Venous Thromboembolism) study.3 The 40% figure is not relevant for consideration of prophylaxis, as it depends on the number of non-medical events. The low prevalence is directly relevant, and weakens the case, as previously argued, for routine thromboprophylaxis.4 In summary, the NICS support for medical thromboprophylaxis may be biased by its reliance on the UWA report. The risk is of overuse of drugs that cause bleeding, and hence of doing more harm than good.5

J Alasdair Millar · Glenda E Lee · Rinaldo Ienco