Letters

In reply: We thank Lavoipierre for his response to our article.1 He brings up two main points. First, he suggests that our statement that newer computed tomography (CT) scanners emit more radiation is incorrect.2 We acknowledge that actual absorbed doses for exactly the same imaging are lower with newer scanners, but the total exposure of the population may be increasing. In the United States, between 1993 and 2006, the annual number of CT scans increased by more than 10% per year, while the population increased less than 1% per year.3 If some scans are unnecessary for safe clinical care, then unsafe radiation exposure is a consequence. His second point is that we should use common sense and not sensationalise our findings out of concern that some patients may not undergo scans that are clinically indicated. We are surprised by his interpretation of the risk of CT imaging, since we quoted a widely accepted estimate.4 Certainly, for an individual, the absolute risk increase is small and may be outweighed by the benefit the scan can provide. However, patients frequently undergo CT scans on the basis of a questionable clinical indication, or when safer imaging modalities would be superior. These unnecessary scans could be avoided. To that end, we advocate common sense, rational test ordering and appropriate discussion of risks, especially where the expected benefit of CT scanning is limited.

Gerben B Keijzers · Charles J Britton

To the Editor: Douglas and other senior epidemiologists have described the imminent demise of the Master of Applied Epidemiology (MAE) program.1 This program has produced around 160 graduates, at least 104 of whom (65%) are working in the broad area of infectious diseases epidemiology. It may be thought, therefore, that there is no further scope for employing MAE graduates. My experience indicates otherwise and I would be very surprised if it were unique. In a small epidemiology unit of a state reference laboratory, two of the six staff members are MAE graduates. One past employee, also an MAE graduate, is a close collaborator. Last year, on very short notice, this person was able to take over coordination of the infectious diseases epidemiology course in the Master of Public Health program (MPH) at the University of Melbourne after the previous coordinator, also an MAE graduate, accepted a position as Head of the School of Health and Social Development at Deakin University. The MAE graduates from the epidemiology unit contributed to the MPH teaching program as guest lecturers. The University of Melbourne was unable to attract a full-time infectious diseases epidemiologist to the coordinating position, suggesting a lack of suitably trained and experienced people in this discipline. There is no doubt that the MAE program produces suitably trained and experienced people, as attested by Douglas et al1 and Kelly et al in the Journal.2 It also fits with my experience as an employer of MAE graduates and a collaborator with other graduates and staff from the program. Continued funding of the MAE program was recommended by independent reviewers.3 There are many who believe that, not only should the MAE program continue, it should continue as part of a national infectious diseases prevention and control program.2 It is appropriate that similar Australian programs already exist for heart disease, cancer and diabetes. In a letter to the Journal, Givney eloquently described the arcane processes of interlocking acronym-rich, mandate-limited committees that constitute Australia’s approach to the national coordination of infectious diseases.4 This description would be recognised by many, who would agree with Givney and Kelly et al that the time has come for Australia to establish its own independent expert body for monitoring and control of infectious diseases.

Heath A Kelly

To the Editor: Trachoma control guidelines from the World Health Organization1 and Communicable Diseases Network Australia (CDNA)2 recommend the “SAFE” strategy that includes surgery for trichiasis, antibiotic treatment, facial cleanliness and environmental improvement. Lack of access to antibiotics in isolated areas should not be a major contributor to the persistence of trachoma in Australia because special Pharmaceutical Benefits Scheme arrangements (SPBSA) under section 100 of the National Health Act 1953 for the supply of medicines to remote-area Indigenous health services should enable a ready availability of azithromycin. The 1999–00 to 2001–02 evaluation of the SPBSA suggested that the program had led to major increases in the supply of medicines in remote areas, but that the supply of azithromycin did not change as a result of the introduction of the program.3 We assessed the relationship between reported azithromycin treatment of people with trachoma, their household contacts, and community members; levels of treatment recommended by WHO and CDNA guidelines; and the total number of courses of azithromycin available through the SPBSA. The National Trachoma Surveillance and Reporting Unit (NTSRU) provided 2008 data on treatment with azithromycin and trachoma prevalence in the Northern Territory, South Australia and Western Australia (refer to the report for limitations of the data).4 Data for 2007–08 on the supply of azithromycin to health services under SPBSA were obtained from Medicare Australia. The WHO guidelines recommend treatment of an entire community if the prevalence of active trachoma among children is above 10%.1,5 The CDNA guidelines2 recommend that contacts (> 6 months of age) of infected children within a household be treated. Our estimate of trachoma treatment according to CDNA guidelines is based on multiplying the number of infected children (from the NTSRU data) by the average number of members in remote Indigenous households.5 The impact of shared and multiple residence on estimates of household contacts could not be taken into account. The Box shows the numbers of azithromycin courses available through the SPBSA to remote-area Aboriginal and Torres Strait Islander health services in the three states. In the NT, the reported number of courses given (3069) fell well below the level recommended by both WHO (by 34%) and CDNA (by 41%) guidelines. In SA, reported courses of azithromycin given (7) fell well below the 45 suggested by the CDNA guidelines. The prevalence of trachoma did not exceed 10% in any SA community, so no treatment was required under WHO guidelines. In WA, the reported number of courses given (2917) also fell below recommended levels, although the deficit (WHO, 35%; CDNA, 81%) varied substantially depending on which guidelines were used. Reported treatment with azithromycin was below levels recommended by the CDNA and the WHO despite health services having sufficient courses available to them to mostly meet these targets. All aspects of the SAFE strategy are important in the eradication of trachoma. However, improving the supply and distribution of azithromycin should be relatively easy to implement, fund and monitor. Azithromycin courses* available, 2007–08;† courses given, 2008;‡ and WHO- and CDNA-recommended courses,§ in three Australian states with remote-area Indigenous health services WHO = World Health Organization. CDNA = Communicable Diseases Network Australia. * Azithromycin courses are given to those with active trachoma, their household contacts, and community members. † Under special Pharmaceutical Benefits Scheme arrangements and reported by Medicare Australia. ‡ Reported by the National Trachoma Surveillance and Reporting Unit (NTSRU). § Based on NTSRU data and using community population estimates.

Margaret Kelaher · Angeline S Ferdinand · Hugh R Taylor

To the Editor: Girgis and Seibel1 are to be congratulated as endocrinologists for raising the important issue of bisphosphonate therapy and femoral fractures. This is a relatively new phenomenon that even a couple of years ago, although already documented, did not seem to be on many endocrinologists’ radar. However, I question the authors’ statement that such fractures are rare. In my small outpost of the orthopaedic world (Northeast Health Wangaratta), three cases have been seen in 12 months. All patients characteristically had a spontaneous non-traumatic fracture, a short oblique or transverse fracture in the subtrochanteric area, and had been on alendronate for more than 5 years. If a journeyman orthopaedic surgeon is seeing a cluster of cases, I suggest that these fractures are not rare. On the other hand, a review of our audit figures in Wangaratta in north-eastern Victoria (where we have a stable rural population of about 17 000) provides an interesting statistic. Over 15 years (1993–2008), Northeast Health Wangaratta admissions for fractured neck of femur fell steadily from 74 to 35 a year. This trend continues. We should not throw out the baby with the bathwater — osteoporosis treatment is likely partly responsible. Also, a deeper level of community aged care support and much better comorbidity management may have helped to reduce falls. The real question is whether all patients who have been on alendronate (or any bisphosphonate) for 5 years need to have a mandatory holiday from the drug. In line with Wolff’s law (bone will adapt to loads under which it is placed), the precursor femoral stress lesions will heal by remodelling, as long as bisphosphonate therapy is suspended. I suggest that the suspension should be for 2 years, which is the time required for full fracture remodelling. I agree with Girgis and Seibel that these drugs have been effective in reducing fracture incidence; however, their long-term use needs further study.

Michael P Falkenberg

In reply: We thank Falkenberg for his comments in reference to our article in the Journal.1 Two recent large-scale population-based studies have suggested that subtrochanteric femur fractures are rare both in the general population and among bisphosphonate users.2,3 While personal experience may often suggest otherwise, a cluster of atypical fractures cannot be used as an indicator of true incidence in the absence of data on the frequency of bisphosphonate use in a particular population. It is certainly safe to say that atypical fractures occur much less frequently than osteoporotic hip fractures. In a 5-year retrospective analysis of femur fractures at our centre, osteoporotic hip fractures outnumbered atypical fractures by a factor of greater than 60.4 Given the body of high-quality evidence on the antifracture efficacy of bisphosphonates, we agree that discarding an effective class of drugs because of a presumed association with an uncommon fracture pattern would be like throwing out the baby with the bathwater. Should we routinely advise patients to take a drug holiday after, say, 5 years of bisphosphonate therapy? There is no good evidence for that either. Although bisphosphonates bind to bone for extended periods, severely suppressed bone turnover or signs of mechanical failure (microcracks) are rarely, if at all, seen in patients chronically treated with bisphosphonates. Similarly, the few bone biopsy studies in patients with atypical fractures do not uniformly support the hypothesis of severely suppressed bone turnover as a cause of these fractures.5,6 Instead of being based on the theoretical assumption of an uncertain risk, the decision for a drug holiday should be made on a case-by-case basis, guided by factors such as the patient’s on-drug fracture history, the presence of other relevant risk factors for osteoporosis, and changes in bone density and bone turnover. Many questions remain unanswered regarding atypical femoral fractures and their biomechanical evolution. Until further research is conducted, the fear of the unknown, namely the impact of bisphosphonates on bone remodelling and microfracture accumulation, should not replace strong evidence in support of their antifracture efficacy.

Christian M Girgis · Markus J Seibel

To the Editor: Point prevalence studies have been used for many years as markers of antimicrobial consumption,1,2 but they have suffered from a lack of standardisation with regard to the populations studied and the data collected. These deficiencies make it difficult to generalise the data outside the study populations. In recent years, the European Surveillance of Antimicrobial Consumption (ESAC) has sought to overcome these weaknesses by implementing a web-based point prevalence survey in 20 European countries using standardised definitions for site of infection, indication and quality indicators, such as whether the indication for prescription is documented in the case notes.3 As Australian studies of this type have not been published, we undertook a pilot study to assess whether this point prevalence tool is feasible and useful in an Australian setting. All surgical inpatients (excluding intensive care patients) in Sydney’s Royal North Shore Hospital (a 541-bed tertiary referral hospital) at 8 am on a single day in December 2009 were included. Current prescriptions for antimicrobials were captured from the medication prescription charts, and data on duration of therapy, dose and route of administration were collected for that point in time to give a “snapshot” of prescribing on that day. We reviewed patients’ medical records to establish site of infection, indication for treatment, and whether the reason for the antimicrobial was documented. ESAC codes for site of infection and indication were used. Of 178 patients, 95 (53%) had been prescribed 140 antimicrobials (ESAC mean, 30% of patients; range, 19%–59%).3 Cephazolin was the most commonly prescribed antimicrobial (32/140; 23%), and 115 antimicrobials (82%) were administered intravenously. Intra-abdominal sepsis (31/140; 22%) and cellulitis or wound infection (21/140; 15%) were the most common anatomical sites requiring treatment. Of the 140 prescriptions, 48 (34%) were for community-acquired infections, 29 (21%) were for postoperative infections, and 35 (25%) were for surgical prophylaxis. Of the 35 prescriptions for surgical prophylaxis, 14 (40%) were administered for more than 1 day (ESAC, 57%).3 The reason for initiation of the antimicrobial was documented in the patient’s medical notes for 95 of 140 prescriptions (68%) (ESAC, 64%).3 Use of the ESAC methodology allowed us to benchmark our results against published international reports. However, it gave no information on the appropriateness of the prescriptions or whether they adhered to antimicrobial guidelines. The ESAC point prevalence tool is easily applicable to the Australian health care system and provides useful information on antimicrobial consumption and quality indicators at an institutional level. Addition of questions regarding appropriateness of prescribing could be incorporated to give further relevant information.4 Use of the same methodology in other Australian centres would enable comparison across institutions and, potentially, national and international collaboration.

Jennifer A Kieran · Rosaleen G O’Doherty · Bernard J Hudson

To the Editor: It is regrettable that research published by Studdert and colleagues suggests that barriers remain to the open disclosure of medical error in Australia.1 Although all states have enacted legislation to protect defendants who apologise to plaintiffs, such protection is variable and inconsistent across Australia.2 Most significantly, the definition of what constitutes an apology varies greatly from jurisdiction to jurisdiction, as do the various protections afforded to the apology. Unfortunately, this only serves to complicate what should be an open and candid discussion of the circumstances surrounding a medical error. However, the fear of legal action is not the only barrier preventing doctors from disclosing harm-causing medical errors to patients and their families. Acknowledging that an error has caused serious harm to a patient is extremely distressing to doctors, most of whom enter the medical profession with the aim of relieving the suffering of others. When their actions inadvertently result in harm to patients, the impact can be devastating.3 The emotional reaction to a medical error is usually one of intense anxiety and concern for the patient’s welfare, then deep reflection on how the error occurred and how the outcome might be mitigated. This may be followed closely by the doctor’s anxiety about his or her own welfare and the professional and legal consequences of the mistake, including the potential loss of reputation or even job. Consequently, doctors are sometimes tempted to rationalise away their role in causing harm or minimise their responsibility for disclosing mistakes and failures.4,5 However, an apology is a powerful tool to facilitate healing of the emotional scars of a patient’s injury. Furthermore, from the perspective of a medical indemnity insurer, while open and transparent disclosure of medical error would seem an improbable risk-management strategy, there is no doubt that a truthful and compassionate explanation of some errors that cause harm may actually reduce the risk of litigation.6 So, as Studdert and colleagues conclude, doctors should be and are supported and encouraged to enter into these difficult discussions by their insurer, without a presupposed fear that the actual process of open disclosure might contribute to the risk of litigation. One hopes that stakeholders will continue to work towards removing barriers to this process, with a greater focus on clinical risk management and less on recrimination and blame within the Australian health care system in years to come.

Julian L Rait · Elizabeth H Van Ekert

To the Editor: Readers of the Journal’s editorial on the importance of clinical patient-outcome monitoring asking “Where are the clinical data?”1 will be pleased to know that the state of Victoria collects considerable data on surgical outcomes. Since 2001, the Department of Health’s Victorian Surgical Consultative Council (VSCC; http://www.health.vic.gov.au/vscc/) has monitored surgical outcomes in the state’s public hospitals. Monitoring of both morbidity and mortality outcomes is combined with voluntary and mandatory case reporting and a high level of participation of hospitals and surgeons. It is compulsory for hospitals to report a range of sentinel adverse events and undertake corrective strategies. Data about surgical inpatients are obtained from discharge coding in medical records statewide, and the chief executive officers and directors of surgery of health services with “outlying performance” are invited to analyse the case records and provide their findings to the VSCC. A VSCC subcommittee, the Surgical Outcomes Information Initiative, promulgates the conclusions in a de-identified manner to hospitals, surgeons and trainees, with a view to improving safety, systems and surgical outcomes. Since 2008, deaths of public hospital patients that occur under surgical care are monitored by the Victorian Audit of Surgical Mortality (VASM, derived from the VSCC), which covers most surgical specialties, and is soon to embrace private hospitals and, hopefully, gynaecological surgery as well. De-identified educative information from autopsies and case analyses is offered to surgeons and trainees, whose participation in case reporting and assessing is now a professional requirement of the Royal Australasian College of Surgeons (RACS). Surgical mortality as monitored by the VASM is now bi-nationally compared through the RACS’s overarching Australian and New Zealand Audit of Surgical Mortality. Victoria’s Department of Health has longstanding collections of data for anaesthetic, obstetric and perinatal outcomes, with similar consultative councils. Several surgical specialties have for over a decade collected and promulgated their morbidity and mortality information. The Melbourne Vascular Surgical Association requires its members to participate in clinical outcome audits, as does the Australian and New Zealand Society for Vascular Surgery. Other specialties that audit outcomes are orthopaedic surgery (the bone and joint registry), transplantation surgery and cardiac surgery. Surgeons’ general experience of clinical patient-outcome monitoring is that of enthusiasm for its professional and community benefits, including the benefit of knowing that Australia’s overall surgical standards are comparable with the world’s best. The incentive remains to improve data capture, patient safety and eternal clinical vigilance, and these endeavours deserve support.

Peter L Field

To the Editor: We would like to report the deaths of three young children in Australia as a result of hanging from bicycle helmets. Our aim is to draw attention to this rare but entirely preventable cause of childhood death. Helmets are required to be worn when bicycles are ridden, and have been the subject of mandatory standards since 1989.1 A number of accidental deaths have, however, been reported in the United States, Scandinavia and Canada as a result of young children becoming suspended by their bicycle helmets while playing on playground equipment. This has led to a series of warnings about not allowing children to wear helmets in playgrounds.2,3 The National Coroners Information System (NCIS)4 is an electronic database containing information on coronial cases from all Australian states and territories since 2001. We undertook a review of the NCIS for all deaths of children in Australia that were associated with bicycle helmets from 2001 to 2009. Three cases of deaths due to hanging were identified; these involved a 2-year-old boy who was suspended by his helmet strap between a bunk bed and a wall (in 2003), a 3-year-old boy who was suspended by his helmet strap when he tried to climb out of a home window (in 2007), and a 5-year-old boy who was suspended from an overhead clothesline while jumping on a trampoline (in 2009). These cases show that accidental hanging is still occurring among young children who wear bicycle helmets while engaging in activities other than bicycle riding. Importantly, hanging from bicycle helmets can occur in places other than playgrounds, sometimes by quite unusual mechanisms. Although such deaths are rare,5 it is important for parents and child carers to ensure that bicycle helmets are only worn by children for their intended purpose, and not during other activities.

Roger W Byard · Allan Cala · Donald Ritchey · Noel Woodford

To the Editor: We read with interest the letter by Baschera et al1 regarding the retrieval of non-permanent inferior vena cava filters (IVCFs) in trauma patients. With the advent of retrievable devices, there has been a renewed interest in the use of prophylactic IVCFs after trauma. Retrievable IVCFs are a particularly attractive option for trauma patients, who are often young, with only a transient predisposition to develop venous thromboembolism (VTE). They theoretically provide prophylaxis against pulmonary embolism (PE) in high-risk patients for whom chemoprophylaxis is contraindicated, while avoiding long-term complications associated with permanent IVCFs. However, we share the authors’ concern about low retrieval rates of IVCFs due to loss of patients to follow-up.2,3 The decision to place an IVCF in a patient who has undergone trauma is based on a risk–benefit ratio. If such decisions are made on the presumption that IVCFs will be retrieved, while in practice many IVCFs are not retrieved, the increasing popularisation of retrievable filters may be under false pretences. At our institution, a level 1 trauma centre, all IVCFs are inserted by interventional radiologists under fluoroscopic guidance. At the time of insertion, patient and device details are entered into a purpose-built IVCF database that is run and maintained by the Department of Radiology. At this time, a follow-up appointment is also booked for the patient to attend the IVCF outpatient clinic run by the interventional radiologists. Here, ongoing VTE risk factors are evaluated and the timing of IVCF retrieval is determined. The timing of IVCF retrieval after trauma is controversial, and the occurrence of PE after retrieval is well documented. The risk period for PE after trauma is difficult to quantify, but likely to extend beyond the discharge date for many patients, especially if ongoing surgery is scheduled. For this reason, our practice is to retrieve IVCFs in the outpatient setting 3–4 months after injury. Prolonged IVCF dwell times must be balanced against the increasing difficulty of retrieval because of IVCF endothelialisation, but we feel that a timeframe of 3–4 months does not compromise retrieval rates. Successful retrieval of filters has been reported up to 317 days after insertion.4 As in the case described by Baschera et al,1 it is not uncommon to find a clot in the filter at the time of attempted retrieval, for which multiple causes have been proposed.5 In this situation, it is our practice to rebook patients for a repeat cavogram and second retrieval attempt after therapeutic anticoagulation for 1–2 months.

Lachlan M Batty · Jim Koukounaras · Stuart M Lyon

To the Editor: Significant morbidity and mortality are associated with excessive alcohol use, which, for Aboriginal Australians, generally occurs within a context of disadvantage. During 2007–2009, we assessed cognitive and psychological factors (using CogState1 and Strong Souls2 [CogState Ltd, Melbourne, Vic]) of 21 men and 11 women on admission to a 2-month Aboriginal residential treatment program in the Northern Territory. Participants’ mean age was 32 years (SD, 8.7 years) and the mean length of time for which they had used alcohol was 13.3 years (SD, 7.7 years). To determine the effect of age, number of years of drinking and other factors on continued alcohol use, we reinterviewed and reassessed participants in their home community with the same cognitive and psychological measures used at the initial assessment after a mean period of 11 months (SD, 4.4 months). At both baseline and follow-up, the number of participants for whom data were available varied for some characteristics. The Human Research Ethics Committee of the Northern Territory Department of Health and Community Services and Menzies School of Health Research (including the Aboriginal Ethics Sub Committee) approved the study. At baseline, 14 of 23 alcohol users reported drinking every day or most days, and 26 of 31 drank more than 10 standard drinks on each occasion. At follow-up in the community, 23 had resumed drinking at the same level, and nine had reduced their use (six had stopped using alcohol, and three had resumed drinking at lower levels). Compared with users who reduced their alcohol intake, users who did not showed poorer paired associate learning at the time of admission for treatment, and poorer performance at follow-up in visual attention, learning and executive function, visual learning and recall, and paired associate learning tasks (Box). This suggests that while subtle cognitive impairment may be a risk factor for continued heavy alcohol use after treatment, heavy alcohol use is also a likely cause of additional cognitive deficits.3 While reduced alcohol use may be associated with improvements in cognitive function, continued use may lead to further cognitive decline. Alcohol users who resumed drinking at the same level were significantly more likely to experience the psychological symptom “worry” after treatment (4/6; Fisher exact test, P < 0.05) than were users who reduced their alcohol use (0/6), which suggests that alcohol may have been used for self-medication or that excessive alcohol use may mask underlying psychological problems. Interestingly, a greater proportion of alcohol users who resumed drinking at the same level (10/16) were also using cannabis at follow-up, compared with those who reduced their use (1/9; Fisher exact test, P < 0.05). Cannabis use has been independently associated with psychological symptoms in other Australian studies, but with no impact on cognition.2,4 Our data indicate that there is a need to treat mental health problems concurrently with alcohol misuse problems among alcohol users undergoing treatment. Alcohol users who resumed drinking at the same level were less likely to return to remote communities with restricted alcohol availability (11/23), compared with those who reduced their alcohol use (9/9; Fisher exact test, P < 0.01), lending some support to the effectiveness of alcohol restrictions. Overall, our data show that cognitive problems and alcohol availability may be underlying factors in ongoing alcohol misuse by Aboriginal Australians. Charactersitics of alcohol users who resumed drinking at the same level and those who reduced their alcohol use after a 2-month residential treatment program, at baseline and at follow-up (n = 32) Characteristic Unchanged alcohol use, median Reduced alcohol use, median Z Significance No. of alcohol users 23 9 Age at baseline, years 31.3 29.0 − 0.15 ns Years of drinking, at baseline 13.0 11.6 − 0.59 ns Visual attention, speed (log transformed)* Baseline 2.81 2.76 − 1.67 ns Follow-up 2.79 2.71 − 2.10 P = 0.04 Working memory, accuracy (arcsine transformed)† Baseline 0.70 0.70 − 0.19 ns Follow-up 0.80 0.74 − 0.53 ns Psychomotor speed, moves per second† Baseline 0.77 0.95 − 0.35 ns Follow-up 1.17 1.37 − 0.75 ns Learning and executive function, moves per second† Baseline 0.44 0.47 − 0.39 ns Follow-up 0.58 0.76 − 2.32 P = 0.02 Visual learning and recall, moves per second† Baseline 0.48 0.46 − 0.21 ns Follow-up 0.73 0.84 − 2.20 P = 0.03 Paired associate learning, duration (seconds)* Baseline 307.81 214.11 − 2.52 P = 0.01 Follow-up 286.51 168.48 − 2.67 P = 0.008 ns = not significant; P > 0.07. * Higher values indicate poorer performance. † Higher values indicate better performance.

Kylie M Dingwall · Paul Maruff · Sheree Cairney

To the Editor: The article by Philip Sambrook et al on the impact of adverse news media on prescriptions for osteoporosis1 contains a number of serious errors and ignores more recent data. Their article quotes the estimated incidence of osteonecrosis of the jaw (ONJ) after oral bisphosphonate treatment for osteoporosis to be between 1 in 10 000 and 1 in 100 000 patient treatment-years. This is in fact a gross underestimate.2 The results of an independent study funded and conducted by the United States Food and Drug Administration (FDA) found an incidence of between 1 in 537 and 1 in 1537.3 This result is similar to that of an independent Australian study, which found an incidence of 1 in 296 to 1 in 1130.4 Thus, there is evidence that bisphosphonate-associated ONJ is much more common in patients treated with oral bisphosphonates for osteoporosis than the authors claim. Bisphosphonate-associated ONJ is also much more serious than presented. Patients with ONJ can be affected for years. The condition causes considerable morbidity, with greater interference in a patient’s life than the condition of osteoporosis and vertebral fractures. It should be noted that the FDA study3 and the Australian study4 were based on documented cases of ONJ, whereas the study by Sambrook et al1 was not based on actual fractures. It is unfortunate that, despite the wording of the Pharmaceutical Benefits Scheme (PBS) regulations, the Pharmaceutical Benefits Advisory Committee has let it be known that osteoporosis diagnosed by bone mineral density is not a requirement for PBS-supported bisphosphonate therapy — all that is required is a minimal trauma fracture, although it is well known that most patients with minimal trauma fracture do not have osteoporosis (however that is defined).5 The issue of adverse effects of bisphosphonate treatment for osteoporosis is currently being tested by a US class action. In the bellwether case of Boles v Merck & Co,6 the issue being tested is whether it was appropriate to prescribe alendronate for a patient with osteopenia but no fractures, and how much this treatment contributed to the patient’s end-stage ONJ. The plaintiff had exposed bone, constant pain, a pathological fracture, and pus dripping from her chin. The New York Supreme Court found in her favour. Despite the claim by Sambrook and colleagues that unbalanced media coverage “has the potential to do more harm than good”,1 the media do have an important role to play in exposing to the public the risks of pharmaceutical products and the actions of some pharmaceutical companies. It should be noted that Professor Sambrook was not only involved in The 7:30 Report current affairs program, but was also subsequently granted the unusual right of reply on that program. Following these two programs, the public made its decision based on facts about the incidence and morbidity of ONJ that have been confirmed by recent independent studies.3,4,6

Paul J Sambrook · B E Christopher Nordin · Alastair N Goss

In reply: Readers will probably deduce that Paul J Sambrook is no relation of Philip N Sambrook. The Australian Broadcasting Corporation, which aired The 7:30 Report that Paul Sambrook and colleagues refer to, subsequently acknowledged in writing that there were a number of factual errors in the original program. Under these circumstances, a right of reply was entirely appropriate. There is ongoing debate about the incidence of osteonecrosis of the jaw (ONJ), but as the modelling used in our study1 did not involve any assumptions about this, it is irrelevant to our findings. The studies by Lo et al2 and Mavrokkoki et al3 did not report incidence of ONJ — they reported prevalence. Moreover, one of the coauthors of the letter by Paul Sambrook et al (above) has published that the estimates by Mavrokkoki et al, being from a retrospective postal survey, must be viewed with caution.4 The Lo et al study was approved by an institutional review board of the United States Food and Drug Administration (FDA), not “funded and conducted” by the FDA, as Sambrook and colleagues state in their letter. There have been numerous studies of the effects of vertebral fractures and osteoporosis on quality of life. The authors claim that ONJ causes greater interference to a patient’s life than osteoporosis or vertebral fractures, but cite no references to justify this bald assertion. The US court case of Boles v Merck & Co5 was not an “independent study” (as implied in the last paragraph of the letter by Sambrook et al), but a legal proceeding that is being appealed. One of the coauthors of the letter was a paid expert for the plaintiff. ONJ is a serious complication of bisphosphonate therapy in the small proportion of affected individuals. We agree that patients need to be informed of the risks of therapy, but they also need to be informed of the consequences of not taking therapy (ie, the benefits forgone) to make a really informed decision. Our article was intended to let patients understand what those consequences might be. The courts are certainly not the place for informed debate. And the media should appreciate that unbalanced reporting can have significant consequences.

Philip N Sambrook · Jiang S Chen · Judy M Simpson · Lyn M March

To the Editor: Clarke and Fitzgerald showed that substantial savings could arise from the implementation of alternative pricing arrangements for off-patent statins that provide incentives to reduce prices and increase generic substitution.1 This is exemplified by comparing statin prices between Australia and England. Europe offers some more lessons with respect to savings based on generic medicine usage and how they can be attained. The size of savings reported by Clarke and Fitzgerald needs to be interpreted with caution. A scenario of 100% generic substitution is proposed. Such a scenario has not been observed in any European country and, for therapeutic reasons, is probably not desirable.2 Furthermore, the comparator country matters: for instance, generic medicine prices in England are lower than in France, the Netherlands and Germany, but are higher than those in Scandinavian countries.3 Finally, the implementation of a tendering system for statins may create unintended effects, such as a switch in prescribing behaviour. For example, the Belgian tendering system for simvastatin reduced expenditure on off-patent medicines containing simvastatin by 30%, but increased expenditure for patented medicines containing atorvastatin or rosuvastatin by 16% and 40%, respectively.4 Clarke and Fitzgerald’s article does not go into detail on how savings can be attained through use of generic medicines. The majority of European countries regulate generic medicine prices by pricing rules or reference pricing. For instance, the implementation of a minimum price difference between originator and generic medicines is the driver of savings arising from generic substitution in some countries, including France, Portugal and Spain. The reference pricing system in Norway stimulated generic competition to a greater extent and led to lower prices than regulation that imposed maximum prices.4 However, price regulation may constitute a barrier for further price competition: no additional price reductions may occur beyond those imposed by regulation.4 The European experience also indicates that the ability of the generic medicine industry to deliver competitive prices can be achieved if it is assured a high volume of the pharmaceutical market. High volume is dependent on demand-side measures that create incentives for physicians, pharmacists and patients to use generic medicines. For example, a European study showed that savings as a result of price competition are higher in countries that have a higher market share of generic medicines.5 Therefore, demand-side measures are critical to increase the generic substitution rate and to maximise the effect of competition based on generic medicine prices.

Steven R A Simoens

In reply: The main purpose of our recent article1 was to quantify estimates of pharmaceutical expenditure over the next decade using various assumptions regarding the price and use of generic statins. We report estimates of billions of dollars in potential savings associated with various scenarios that increase the current off-patent statin use in Australia from around 25% of prescriptions to between 50% and 100%. Simoens misinterprets our conclusions as recommending only using generic statins in Australia. We do not advocate any particular level of generic substitution, but argue that the optimal mix of patented and generic statins should be determined by using cost-effectiveness analysis. Simoens questions whether our results would change if we had compared statin prices with a country other than England. To address this issue we have compiled a comparison of current or recent wholesale price of 40 mg simvastatin across 13 countries in the Organisation for Economic Co-operation and Development (Box). Although there is some variation between countries, the main difference is with Australia, which has the highest wholesale price — about five times greater than the average price across all comparator countries. This price ratio is similar to the one used in our original study. Also, Simoens highlights several issues relating to alternative pricing arrangements for statins and other generic drugs in European countries. We agree that Australia may be able to learn from overseas experience when reforming its system of pricing generic pharmaceuticals. The impact on expenditure of the tendering system for supply of generic pharmaceuticals may have been counteracted by changes in prescribing behaviour in Belgium, but it has been successfully used in the Netherlands to cut the price of simvastatin and other major generics by over 80%. This has been estimated to save around 310 million euros annually.2 However, tenders are not the only way to reduce the price of generic pharmaceuticals. In Canada, the Ontario Ministry of Health and Long-Term Care has recently introduced a policy which sets the subsidy for generics at 25% of the original price under patent. Generic 40 mg atorvastatin, whose patent in Canada recently expired, now costs just A$17 per month.3 In contrast, under the recent Memorandum of Understanding4 between the Australian Government and Medicines Australia, the current wholesale price of A$61 for 40 mg atorvastatin will decline by only 16% after the patent expires in Australia in 2012, and there will be no further downward adjustment until at least 2014. Wholesale price of simvastatin 40 mg in 13 countries in the OECD* OECD = Organisation for Economic Co-operation and Development. * Comparator prices were converted to Australian dollars using the average exchange rate over the past 3 years, and 1 month supply was assumed to be equivalent to 30 tablets.

Philip M Clarke · Edmund M Fitzgerald

To the Editor: I am writing about important errors contained in a letter of reply by Hamilton-Craig.1 In his response to a letter by Montgomery,2 he states: The Simvastatin and Ezetimibe in Aortic Stenosis (SEAS) trial (in which patients with aortic stenosis were treated for 52.2 months with statin plus ezetimibe or statin plus placebo) showed a 4.7% reduction in ischaemic [cardiovascular disease] events in the ezetimibe group (P = 0.02; number needed to treat, 23), driven by a reduced need for coronary artery bypass grafting. However, in the SEAS trial, patients were treated with statin plus ezetimibe or with placebo, not with statin plus placebo. Therefore, the reduction in ischaemic events in the statin/ezetimibe group cannot be attributed to ezetimibe, as it could have been caused by the effect of simvastatin alone (or by the combined effect of the two drugs). I note also that the absolute reduction in ischaemic events over the course of the trial was 4.4% (15.7% v 20.1%), not 4.7%.3 With respect to the ENHANCE (Ezetimibe and Simvastatin in Hypercholesterolemia Enhances Atherosclerosis Regression) trial, Hamilton-Craig states: As no placebo group was included, neither lack of benefit nor harm from ezetimibe therapy can be inferred.1 In the ENHANCE trial, patients with familial hypercholesterolaemia were treated with simvastatin plus ezetimibe or simvastatin plus placebo. There was no significant difference in progression of mean carotid intima media thickness (CIMT) between the two groups (0.0058 mm in the simvastatin/placebo group v 0.0111 mm in the simvastatin/ezetimibe group [P = 0.29]).4 Therefore, contrary to Hamilton-Craig’s statement, there was a placebo group, and a lack of benefit from ezetimibe was shown in the trial. Thus, the SEAS trial was unable to confirm a benefit of ezetimibe, as the active treatment arm included both a statin and ezetimibe, while the ENHANCE trial showed no additional benefit of ezetimibe on the surrogate endpoint of CIMT progression in patients taking a statin. Author’s note: The United States Securities and Exchange Commission disclaims responsibility for any private publication or statement of any Commission employee or Commissioner. This letter expresses my views and does not necessarily reflect those of the Commission, the Commissioners, or other members of the Commission staff.

Marilyn K Mann

To the Editor: I would like to endorse the letter by Montgomery1 questioning the efficacy of ezetimibe. As yet there are no data to support its use in clinical trials using carotid intima media thickness (CIMT) as a measure of treatment effectiveness, and there is also some evidence to suggest it could be harmful. A randomised trial conducted by Berneis et al2 suggested that ezetimibe may induce an unfavourable pro-atherogenic low-density lipoprotein (LDL) subfraction profile by increasing small, dense LDLs. The Simvastatin and Ezetimibe in Aortic Stenosis (SEAS) trial3 compared simvastatin/ezetimibe 40/10 mg daily with placebo, so any benefit from that treatment may have been from either the ezetimibe or the simvastatin. It tells us nothing about the effectiveness of ezetimibe alone. Until the results of clinical trials are available, I believe ezetimibe should be used with much reluctance and only considered as a last resort. I agree with Hamilton-Craig4 that it is a matter of concern that slow-release niacin, which is effective and safe, is not available under the Pharmaceutical Benefits Scheme in Australia. Searching for supplies of this drug in Australia or overseas seems the best option for treating patients whose levels of LDL cholesterol are inadequately controlled with statin therapy.

Brett H Forge

In reply: In the interests of scientific exactitude, I am indebted to Marilyn Mann for corrections regarding the Simvastatin and Ezetimibe in Aortic Stenosis (SEAS) trial. However, as the Ezetimibe and Simvastatin in Hypercholesterolemia Enhances Atherosclerosis Regression (ENHANCE) trial did not include the control group ezetimibe versus placebo, the effects of ezetimibe on carotid intima media thickness remain somewhat speculative.1

Ian R Hamilton-Craig

To the Editor: I note the letter in the 7 June issue of the Journal by Ojaimi and colleagues, in which they express their concerns about the lack of uniform screening guidelines for patients on hydroxychloroquine therapy.1 The most obvious problem was the apparent failure of the treating rheumatologist to ensure that the patient was adequately followed up for ocular side effects associated with this treatment. The potential for these side effects has been documented for many years and, while more pertinent in earlier times when chloroquine was used more frequently, isolated case reports about retinal toxicity related to hydroxychloroquine continue to be presented in the medical literature. All patients need to be warned that there is a risk to their eyesight before they start treatment. The pertinent questions are, however, what is the risk and what is the cost of screening? One study found a single case of retinal toxicity in a series of over 1200 patients being treated long-term with hydroxychloroquine.2 Using this information and the data quoted by Ojaimi and colleagues on hydroxychloroquine use,1 there would currently be about 20 000 patients in Australia using hydroxychloroquine, and 17 potential patients with toxicity in the whole nation. Screening guidelines vary, with most recommending a baseline screen and then, depending on the presence or absence of “high risk factors” (dosage > 6.5 mg/kg/day, treatment for > 5 years, renal or hepatic impairment, concomitant eye disease, age > 60 years), testing 2 years after the baseline test and then annually,3 or 5 years after the baseline test and then annually.4,5 The cost of screening the 20 000 Australian patients being treated with hydroxychloroquine according to these guidelines and assuming a standard initial specialist consultation (Medicare Benefits Schedule item 104) and field test (Medicare Benefits Schedule item 11222) was performed would be between $20 174 000 and $28 820 000 over a 10-year period. One really has to ask, can we afford it? The patient described by Ojaimi and colleagues1 falls clearly into the high-risk group in terms of her daily dose and duration of treatment. The slit lamp findings of vortex keratopathy also indicate she had a high risk of retinal toxicity. Rather than screening all patients on hydroxychloroquine therapy, doctors prescribing the medication should be more aware of the potential consequences, and monitor the dosage carefully so that patients in the high-risk group can be referred for screening. Multifocal electroretinography and fundus autofluorescence could potentially be used for screening, but these modalities are not available routinely and their utility in screening is yet to be established. Until then, patients at high risk of hydroxychloroquine retinopathy should be referred to an ophthalmologist for examination. The minimum examination would include measurement of visual acuity, slit lamp biomicroscopy, dilated fundus examination and automated perimetry, looking specifically for changes in the central 10 degrees of visual field.

Trevor J P Hodson

To the Editor: We share the concern of Fernando and colleagues about the recent emergence of multidrug-resistant bacteria via travellers returning to Australia and the resulting reduction of therapeutic options for such patients, who may have entered a “post-antibiotic era”.1 In countries such as India that lack an integrated laboratory network, the precise magnitude of the threat of plasmids encoding blaNDM1 type metallo-β-lactamases is unknown. Molecular testing for the NDM-1 type of β-lactamases might be available at only two or three national laboratories, and most hospitals will have no such testing facilities. During replication of multidrug-resistant bacteria carrying blaNDM1 type metallo-β-lactamases, susceptibility to some older antimicrobial agents might be retained. This was evident in two multidrug-resistant Enterobacteriaceae isolates that were identified at the Sant Parmanand Hospital (a 140-bed tertiary-care, multidisciplinary hospital that serves the population of Delhi and two adjoining townships), at which blaNDM1 metallo-β-lactamase testing is not available. From March to August 2009, 509 Enterobacteriaceae isolates were identified — Klebsiella pneumoniae (368), Escherichia coli (112), Salmonella enterica serotypes Typhi, Paratyphi A, and Paratyphi B (20) and Proteus species (nine). Bacteria were identified by morphological, biochemical and serological characteristics. Antimicrobial susceptibility was tested by the disk diffusion method, according to the United States Clinical and Laboratory Standards Institute guidelines. Two of the K. pneumoniae isolates were resistant to meropenem, piperacillin–tazobactam, cefepime, amoxicillin–clavulanic acid, amikacin, gentamicin, ciprofloxacin and tigecycline. One had been isolated from the pulmonary secretions of a 45-year-old woman in March 2009; it was susceptible to ofloxacin and chloramphenicol. The other had been isolated from the urine of a 55-year-old woman in July 2009; it was susceptible to chloramphenicol and nitrofurantoin. Among all the K. pneumoniae isolates, there were 53 resistant to meropenem, 113 resistant to ceftriaxone, 82 resistant to gentamicin and 111 resistant to ciprofloxacin. Among all the E. coli isolates, there were three resistant to meropenem, 38 resistant to ceftriaxone, nine resistant to gentamicin and 35 resistant to ciprofloxacin. The proportion of meropenem- and gentamicin-resistant isolates was significantly higher for K. pneumoniae compared with E. coli (Fisher exact test, P < 0.001). Older antimicrobial agents such as ofloxacin, chloramphenicol and nitrofurantoin, developed in the 1940s and 1950s, would not be the initial choice for today’s clinicians managing patients with severe multidrug-resistant infections. However, they should be considered before declaring that the post-antibiotic era has arrived.

Subhash C Arya · Nirmala Agarwal

To the Editor: The article by Hawthorne and Hamilton, International medical students and migration: the missing dimension in Australian workforce planning?,1 highlights the issue of international students wanting to stay in Australia for internships and beyond. While the ethical dilemma created by keeping much-needed future doctors from their own countries has been extensively debated, the reality is that Australians, especially those in rural and remote settings, will rely on overseas-trained doctors for health care until the “tsunami” of current Australian medical students complete their training (in about 2020). The health workforce initiative of the Department of Health and Ageing Rural Clinical Schools (RCSs) promotes rural careers by funding 25% of Australian students for a year of rural clinical training. Funding is not provided for international students because of cost and limitations of rural training capacity (supervisor and infrastructure shortages). While “most” international students are interested in metropolitan practice,1 our RCS, at the Melbourne Medical School’s Rural Health Academic Centre, has had repeated, passionate requests from international students to attend the RCS. And quality rural placements increase (international) student and trainee interest in rural practice.2,3 Even if about 80% of the 25% of students who undergo a year of rural training (20% overall) return to rural health care (an optimistic assumption), Australia will still have a shortage of Australian-trained doctors willing to work in rural Australia (about 20% of medical graduates for 29% of the population). We suggest that the cost and capacity to train an international graduate of an Australian medical school may be less than the cost of recruiting, acculturating and up-skilling an international medical graduate. And targeted training and retention of international students from less disadvantaged countries will decrease recruitment from countries with severe health worker shortages.4 Thus, it might be wiser, more ethical, and perhaps more cost-effective to allocate funding to truly interested international students to attend an RCS or an extended quality placement in rural Australia; and then to support them for postgraduate training positions in exchange for rural payback (“bonding”). Of course, the challenge will be to identify international students with a true interest in rural health care and a willingness to be part of the solution. But if such students were willing and could be selected, they might fill the estimated 10% gap between the 20% of Australian students intending to practice rurally and the 29% of the Australian population that lives rurally. We hope that a cost-effectiveness analysis of trade-offs discussed here will become a priority for Australian health-workforce researchers so that the feasibility of this strategy can be determined.

Dawn E DeWitt · William R Adam

To the Editor: The benefits of bicycle helmet use have been the subject of recent discussion, with calls from some experts to review laws mandating the wearing of helmets.1 The objective of this brief report is to summarise long-term trends in cyclist head injuries seen at an inner-city major trauma centre and determine the odds of any skull fracture or intracranial bleed associated with not wearing a helmet. This was a retrospective study conducted at the Royal Prince Alfred Hospital (RPAH, Sydney, New South Wales), covering several local government areas that have the highest bicycle-use rates in NSW,2 where the law for mandatory helmet wearing was enacted in 1991. Patient data were obtained through the hospital trauma registry, which contains data on all patients admitted to the hospital with trauma. These data included information on helmet use routinely abstracted from ambulance and medical notes. Inclusion criteria were cyclists admitted from 1991 to 2009, who were over 16 years of age and involved in an incident on a public road. We excluded patients transferred from other hospitals. Head Abbreviated Injury Scale (AIS) scores (AIS 1990, 1998 and 2005 versions3) were used, with a head AIS score ≥ 3 indicating severe head injury, such as significant intracranial bleeding or depressed or comminuted skull fracture. Injuries with an AIS score of 2 included isolated concussion and simple skull fractures. To investigate the association between helmet use and head injury, we reviewed the medical charts of all cyclists admitted with trauma from 2008 to June 2010. We compared mechanism of injury (fall off bike without collision versus collision with another vehicle or object), anatomical injury (skull fracture or intracranial bleed), helmet use and the type of road where the incident occurred (state or regional roads versus local roads), according to NSW Roads and Traffic Authority classifications. Data were analysed using Stata software, version 10.1 (StataCorp, College Station, Tex, USA). Percentages were calculated with 95% confidence intervals, and categorical data were compared using χ2 tests. Mean ages were compared using the Student t test, and a logistic regression model was used to obtain odds ratios for any skull fracture or intracranial bleed associated with not using a helmet, after adjusting for mechanism of injury and road type. The study was approved by the Sydney South West Area Health Service RPAH Ethics Review Committee (RPAH Zone). There were 979 patients who met our inclusion criteria. The long-term trend in the number of cyclists sustaining severe head injuries remained low (range, 0–3 per year) (Box 1). Cyclists as a percentage of total admissions for trauma increased from 1.3% in 2005 (29/2258 [95% CI, 0.9%–1.8%]) to 3.9% in 2009 (122/3104 [95% CI, 3.3%–4.7%]). Trends in helmet use and severe head injury are summarised in Box 2. Severe head injury rates as a percentage of total cyclists admitted decreased from 10.3% (3/29 [95% CI, 3.6%–26.4%]) in 2005 to 2.5% (3/122 [95% CI, 0.8%–7.0%]) in 2009, a relative reduction of 76%. Helmet use in admitted cyclists from 1991 to 2009 ranged from 85% to 100%. Information was available about the location of the fall and helmet use for 287 of the 313 cyclists identified from 2008–2010 (Box 3). Their mean age was 36 years (95% CI, 34–37 years) and 81% were men. Non-helmet wearers had five times higher odds of intracranial bleeding or skull fracture compared with helmet wearers after adjusting for road type and mechanism of injury (odds ratio, 5.3 [95% CI, 1.7–17.1]; P = 0.005). The increase in admissions for bicycle injury is consistent with recently reported population trends.4 In addition, the number of cyclists sustaining severe head injuries has remained consistently low over the long term, with an apparent decline in the rate of severe head injuries in admitted patients since 2005. The odds reduction for skull fractures and intracranial bleeds in those wearing helmets is within the range reported in a Cochrane review of helmet use.5 The benefits of helmet use need to be placed in the context of lifetime costs of severe traumatic brain injury, estimated to be around $4.8 million per incident case.6 It is the opinion of the trauma service at RPAH, based on these findings, that mandatory bicycle helmet laws be maintained, and enforced as part of overall road safety strategies. 1 Trends in cyclist admissions and head injuries in admitted cyclists, RPAH, Sydney, New South Wales, 1991–2009 AIS = Abbreviated Injury Scale. RPAH = Royal Prince Alfred Hospital. 2 Trends in bicycle helmet use and severe head injury as a percentage of total cyclist trauma admissions, RPAH, Sydney, New South Wales, 1991–2009 AIS = Abbreviated Injury Scale. RPAH = Royal Prince Alfred Hospital. 3 Head injury in helmet and non-helmet users among 287 cyclists admitted to Royal Prince Alfred Hospital with trauma, 2008 to June 2010 Helmet (n = 241) No helmet (n = 46) Significance† Age, years (95% CI) 36 (34–38 years) 33 (29–37 years) P = 0.14 Men (%; 95% CI) 196 (81%; 76%–86%) 39 (85%; 71%–92%) P = 0.60 Fall off bicycle* (%; 95% CI) 83 (34%; 29%–41%) 13 (28%; 17%–43%) P = 0.75 State/regional road (%; 95% CI) 63 (26%; 21%–32%) 11 (24%; 14%–38%) P = 0.75 Skull fracture or intracranial bleed (%; 95% CI) 8 (3%; 2%–6%) 6 (13%; 6%–36%) P = 0.005 * Without direct collision with another vehicle, object or person. † Two-tailed P < 0.05 significant.

Michael M Dinh · Susan Roncal · Timothy C Green · Elizabeth Leonard · Amanda Stack · Chris Byrne · Jeffrey Petchell

To the Editor: It is incorrect for Chong and colleagues to state that “during the financial year 2006–07, 32 777 people were admitted to hospital in Australia due to road crashes”. It is also wrong for them to claim “it is often not clear how many of these road crashes are traffic crashes, and how many are non-traffic crashes”.1 Henley and Harrison report that there were 52 066 people seriously (but not fatally) injured due to land transport injury in 2006–07, and 32 777 of these (63.0%) occurred in traffic (on-road) accidents.2 A further 13 639 (26.2%) land transport injury cases in that year were explicitly described as non-traffic (off-road) accidents. The National Injury Surveillance Unit of the Australian Institute of Health and Welfare regularly publishes transport injury-specific analyses, including reports on land transport injuries (both traffic and non-traffic), rail-related transport injuries and transport injuries involving Indigenous Australians. The most recent of these reports is Henley and Harrison’s.2 In addition to the statistics mentioned above, they also report that the national age-standardised rate of non-traffic transport injuries was 66.5 per 100 000 population. In the previous year, this rate was 67.2 per 100 000 population.3 The next report in this series, to be published shortly, will include analysis of national trends in the rate of non-traffic transport injuries over the period 2000–01 to 2007–08.

Clare E Bradley · James E Harrison · Geoffrey I Henley

In reply: We acknowledge our error in reporting Henley and Harrison’s findings,1 and commend Bradley and colleagues for providing information about traffic and non-traffic transport injuries separately. We are also pleased that the National Injury Surveillance Unit will soon publish a report including analysis of trends in non-traffic transport injuries, extending our analyses beyond New South Wales. This is consistent with our conclusion that more needs to be done to understand non-traffic crashes.2 We defend our claim that the statistics in many reports and articles often do not clearly distinguish between traffic and non-traffic crashes and injuries.

Shanley S S Chong · Wei Du · Julie Hatfield

To the Editor: An editorial by Stewart alludes to the problem of silent multiple comparisons when interpreting P values from cancer cluster investigations.1 Visible multiplicities such as occur with pre-specified subgroup analyses or sequential monitoring of trials are difficult enough, but at least in these circumstances we know how many multiple comparisons are under consideration. More difficult are silent multiplicities such as occur with cluster investigations (and also with publication bias2 or reporting bias3) where we do not know how many multiple comparisons should be considered. The P value is intended to be an objective measure of the play of chance, and this is (arguably) the case when applied to a pre-specified primary hypothesis in a randomised trial. But this is not the case for cluster investigations, in which the number of multiple comparisons can never be known with any certainty. Statisticians analysing data from a cluster could obtain any P value they wanted by calibrating it against an arbitrary number of multiple comparisons. Where does this leave scientific reasoning in cluster investigations? All cases of cancer have causes; the key question in a cluster investigation is whether the cases have a common cause related to the neighbourhood or workplace from which the cluster was reported. Only rarely is an obvious common cause identified, and a decision to take some action (eg, evacuate the workplace) needs to be based on expert opinion. For the ABC cluster, no obvious common cause was identified. However, the expert panel was concerned that the women with breast cancer were relatively young and were long-term employees at the site, suggesting that there might be an unidentified common cause related to the site.4 This concern, based on expert opinion, is (arguably) enough evidence to evacuate the site. Investigation of cancer clusters is a difficult task. If a common cause cannot be identified, then there is no objective evidence on which to obtain agreement among experts about the importance of the cluster. Specifically, we need to be very clear that, for cluster investigations, a P value (even when adjusted for multiple comparisons) does not provide an objective measure of whether the cluster is due to chance. In the end, an expert group has to make a decision in the presence of uncertainty. When communicating the results to the public, the uncertainty should be acknowledged — as should the fact that experts sometimes disagree.

Michael D Coory