Letters

To the Editor: Both the American Diabetes Association (ADA)1 and the World Health Organization (WHO)2 have lowered the fasting plasma glucose (FPG) level for the diagnosis of diabetes from 7.8 mmol/L to 7.0 mmol/L. The Australian Diabetes Society (ADS) has also adopted the lower level.3 However, these organisations differ in the procedure for diagnosis they recommend. The recent article by Hilton and colleagues4 compared these procedures, with particular attention to including an oral glucose tolerance test (OGTT). We, on the other hand, have investigated the impact of lowering the diagnostic FPG level to 7.0 mmol/L. Data were obtained by the Geelong Osteoporosis Study from an age-stratified sample of women randomly selected from electoral rolls for the Barwon Statistical Division5 and adjusted to match the national age profile. Venous FPG level was determined after an overnight fast, together with blood pressure (BP, seated) and anthropometric measurements, in 944 women aged 20–91 years (mean age, 47.5 years; SD, 17.8 years). History of diabetes was ascertained by questionnaire. The prevalence of self-reported diabetes and diabetes defined by an FPG level of 7.0 mmol/L or higher was 4.3% (95% CI, 3.0%–5.6%; 41 women), whereas using an FPG level of 7.8 mmol/L or higher gave a prevalence of 3.8% (95% CI, 2.6%–5.0%; 36 women). With the lower cut-off level, 29% of women (12) were unaware of their diabetes, compared with 19% (7) using the higher FPG level cut-off point. Characteristics of those identified using the lower cut-off FPG level are shown in the Table. After age-matching all patients with diabetes with control participants, diabetes was significantly associated with obesity (body mass index, > 30; odds ratio [OR], 4.2; 95% CI, 1.5–11.6) and central body fat distribution (waist/hip ratio, > 0.8; OR, 8.0; 95% CI, 2.3–25.9); and non-significantly associated with higher blood pressure (systolic, > 140 mmHg; diastolic, > 85 mmHg; OR, 2.0; 95% CI, 0.8–4.9). The new criterion for diagnosing diabetes identifies a subgroup of the population with a high proportion of obesity and android habitus, with a tendency to higher blood pressure. The recommendation of lowering the diagnostic FPG level increases the prevalence of diabetes by an apparently small proportion, but would diagnose diabetes in an additional 34 000 women in Australia. Characteristics (mean ± SD) of diabetic women (FPG ≥ 7.0 mmol/L) and controls (FPG < 7.0 mmol/L). Characteristic Diabetics (n = 41) Controls (n = 903) P* Age (years) 65.1 ± 11.1 46.7 ± 17.7 < 0.0001 Weight (kg) 74.5 ± 16.2 68.6 ± 14.4 0.03 Height (cm) 158.6 ± 5.6 161.9 ± 6.5 0.0007 BMI (kg/m2) 29.6 ± 6.1 26.2 ± 5.3 0.001 Waist/hip ratio 0.88 ± 0.06 0.80 ± 0.07 < 0.0001 Systolic BP (mmHg) 139 ± 21 121 ± 21 < 0.0001 Diastolic BP (mmHg) 83 ± 16 76 ± 12 0.007 * t test. FPG = fasting plasma glucose; BMI = body mass index; BP = blood pressure.

Julie A Pasco PhD · Mark A Kotowicz MB BS, FRACP · Margaret J Henry PhD · Geoffrey C Nicholson PhD, FRACP, FRCP

To the Editor: The recent letter by McElduff and Hitchman1 has some very practical implications. They were able to show that pregnant women having a glucose challenge test (GCT) in the afternoon were nearly twice as likely to have a positive result as women tested in the morning, so that more women tested in the afternoon were diagnosed with gestational diabetes mellitus (GDM). If the function of the GCT is to aid in the diagnosis of GDM, then either all women should be tested in the afternoon or the glucose "cut-point" for the morning test should be reduced. But does the GCT now have any relevance? In the United States, where testing for GDM often still involves a three-hour glucose tolerance test (GTT) using a 100 g glucose load and four blood samples, the GCT was introduced to reduce the number of women who had to have this long and, because of the higher dose of glucose, relatively unpleasant procedure. In Australia, where a two-hour, 75 g GTT is used (requiring two blood samples), it is not as important to offer a simpler initial test. With the use of an initial GCT, about a quarter of women will need to have a GTT for confirmation, and the definitive diagnosis of GDM will be delayed. Further, the GCT is not specific and some women who may have GDM will not have a GTT. In addition, there will inevitably be some women who are GCT-positive, some of whom will have GDM, who do not return for the definitive GTT. Thus, while a GCT may be convenient for a busy hospital clinic with space limitations, it may not necessarily be in the best interests of the patient. Whether a GCT is ultimately helpful or possibly a hindrance requires further evaluation.

Robert G Moses MD

To the Editor: We conducted a joinpoint analysis of death certificate data on prostate cancer from the Australian Bureau of Statistics. Between 1979 and 1994, mortality rates increased by 2.1% per year (95% CI, 1.6% to 2.5%). However, between 1994 and 1999, mortality decreased by 4.2% per year (95% CI, – 5.8% to – 2.4%). The total decrease in mortality rates for the five years to 1999 (the most recent year for which data were available) was 22.6% (95% CI, – 32.9% to – 12.7%). Joinpoint analysis is a statistical method that measures changing trends over time. It chooses the best-fitting points (called joinpoints) at which the rate of increase or decrease changes significantly.1 We did not look at the data and then choose 1994 as the start of the decreasing trend in prostate cancer mortality. The significant decrease since 1994 (and the consistent increase between 1979 and 1994) were identified by the joinpoint analysis. Whether early diagnosis and treatment of prostate cancer subsequent to screening with prostate-specific antigen (PSA) tests can save lives is still an open question that is best answered by randomised-controlled, long term trials. Nevertheless, it is important that we try to understand the recent decrease in population-based mortality. The mortality decline started in 1995, about five years after PSA testing became widely available in Australia. The use of PSA testing increased dramatically, reaching a peak in most States in 1994 and 1995.2 In 1996, the Australian Health Technology Advisory Committee reviewed the evidence and recommended against screening.3 Since then, the number of PSA tests has decreased.2 Recent mortality declines have also been observed in the United States and the United Kingdom following increases in the use of PSA testing. The mortality decline in the US has been greater than that in the UK, coinciding with more PSA testing in the US than the UK.4 At least three questions arise from these observations: Is it plausible that the decrease could be due to some factor other than PSA testing, such as better treatment? How do we explain these results to men who want to make an informed choice about whether to be tested? Is this type of evidence strong enough to warrant a change in the current recommendations on PSA testing?

Michael Coory MB BS, PhD · Peter Baade BSc, PhD

Comment: Using joinpoint analysis, Coory and Bade have identified a significant trend towards lower mortality rates from prostate cancer since 1994, and relate this chronologically to the increased use of prostate-specific antigen (PSA) testing in Australia. The authors raise three reasonable questions regarding this finding: Is it plausible that the decrease could be due to some factor other than PSA testing, such as better treatment? Ecological data such as those presented by Coory and Baade are subject to the pitfalls of ecological fallacy and confounding.1 Ecological fallacy exists when there is an apparent association (eg, more men in the population screened, fewer men in the population dying), but there is no association at an individual level. Confounding would occur when there are other factors that account for the outcome, or distort the relationship between PSA testing and mortality. Treatment is one such factor, but changes in diet and other factors could also play a role (eg, lycopene from tomato-based foods2). How do we explain these results to men who want to make an informed choice about whether to be tested? If these data were presented to men, then it would have to be explained that the data do not provide evidence of the benefits of screening and that many other factors could explain the relationship. Further, men need to be informed of the risks of false negative and false positive results associated with screening, and the risks of complications associated with treatment. It is important to assist men to make a balanced decision. Interestingly, studies in which men have been involved in such informed decisions show that information can reduce the probability that men will choose to be screened.3,4 Is this type of evidence strong enough to warrant a change in the current recommendations on PSA testing? Ecological associations generate hypotheses that are worthy of further investigation. Randomised controlled trial evidence at best, or case–control studies at least, would be required to provide evidence of the benefit of screening using PSA. Some such studies are currently in progress, including the prostate, lung, colorectal and ovarian cancer screening trial of the National Cancer Institute,5 and the European Screening Study for prostate cancer.6 Those in favour of PSA testing argue that it is the only means of diagnosing prostate cancer at an early and potentially curable stage.7 Against this is the argument that many prostate cancers are not clinically significant, as they are slow growing and will not spread, and there is a danger of diagnosis and treatment adding substantially to men's psychological and physical morbidity without benefits in terms of survival or quality of life.8 In the absence of evidence from randomised controlled trials or case–control studies, screening for prostate cancer using PSA testing remains controversial.9

Julie B Byles

To the Editor: A previously well 21-month-old boy presented with burns on his right thigh. His mother had been making popcorn in a domestic popcorn machine while holding the baby about a metre from the machine to watch it churning out the popcorn through a chute into a bowl. Soon after the corn started to pop, the baby screamed, pointing to his right thigh. The mother immediately removed his nappy and found an unpopped corn kernel lodged between the nappy and the thigh. General examination revealed two 4 mm burns on the child's right thigh (Figure). Kenacomb cream (Bristol-Myers Squibb) was applied four times daily, and the lesions healed after five days, leaving two depigmented scars. Burns to the right thigh caused by a popcorn maker. Although popcorn machines are gaining increasing popularity there are few reports of injury. One report describes corneal burns in three adults caused by steam from microwave popcorn.1 We believe that this is the first reported case of paediatric injury, highlighting the potential hazards of hot-air popcorn machines. Corn's ability to pop lies in the fact that the kernels contain a small amount of water (14% of weight) stored in a circle of soft starch inside the hard outer casing.2 When heated to about 232ºC,3 the water expands, creating pressure within, until eventually the casing gives way. The kernels explode and pop, allowing the water to escape as steam, turning the kernels inside out. It is likely that the temperature of kernels that do not pop (known to popcorn connoisseurs as "old maids")2 can reach as high as 200ºC. Because of their increased density compared with popped kernels, old maids can, as occurred in our case, be thrown further than popped kernels. It is also worrisome that some popcorn machines are designed so that they can be used by children.4 The information booklet that accompanies the popcorn machine in this case does state "close supervision is necessary when this appliance is being used by or near children", and has a warning that as "some hot unpopped corn may be thrown from the machine during the popping process be sure to place the [machine] facing away from you or do not stand directly in front of the machine whilst it is in operation". The mother of the child in our case did not note the warning in the instruction booklet. She was not standing directly in front of the machine. Further, she was of the impression that the fluffy popped corns are not hot. With the increased use of popcorn machines, consumers must be aware of the potential dangers of injuries, especially to young children. Our case emphasises the importance of keeping children at a safe distance while preparing hot food, including apparently harmless, fluffy popcorn.

T H H Guan Koh MA, MB BChir, FRCPCH, FRACP

To the Editor: The report by Mangion et al1 draws attention to a serious risk in the environment. The general public has been increasingly protected against the risk of harm from domestic products by a combination of legal liability actions and government regulation. Thus, the continuing availability to the general public of hydrofluoric acid (HF) in concentrations that are hazardous is something of an anachronism. While we applaud Mangion and colleagues for raising the issue of HF burns, we feel that their article is deficient in failing to mention a number of important points. Topical calcium gluconate has been shown to be more effective in treating HF burns if the preparation contains dimethyl sulfoxide (DMSO).2 There is a great risk of blindness with ocular exposure to HF. Slow local injection with 10% calcium gluconate using fine needles, titrating its effect against the patient's pain, is a well described technique. This is another treatment option that could have been tried. Nail removal, described by Mangion et al as an "extreme measure", is, unfortunately, often required. It is less likely to be required with the application of DMSO/calcium gluconate solution and retrograde ischaemic intravenous injection of calcium. Local excision of contaminated tissue may be required after exposure to concentrated solutions. Management should be a team effort from the first moment, involving an intensivist/toxicologist and surgeon, as burns surgeons are trained in the care of HF exposure, and surgery is often needed. The availability, packaging, and labelling of preparations containing HF have recently been changed. Since 1 December 2001 it has no longer been possible for the general public to purchase any HF preparation stronger than 1%. All preparations now carry prominent labelling drawing attention to the risk of blindness if even dilute solutions of HF get into the eyes. Containers are now less easy to open by children. These changes have been introduced by the National Drugs and Poisoning Committee of the Therapeutic Goods Administration as a result of an independent review and lobbying by the Australian and New Zealand Burn Association (ANZBA). The ANZBA guidelines for referral to a specialised burns unit include chemical burns. The peculiar challenge posed by HF burns emphasises the need for the guidelines to be more widely disseminated. Currently, the New South Wales Department of Health has adopted the guidelines, so this policy is official throughout New South Wales.

Hugh C O Martin · Michael J Muller

To the Editor: A recent editorial in the Journal attempted to define a role for angiotensin II receptor (AIIR) antagonists in patients with type 2 diabetes.1 This was in the light of recent trial evidence that these agents reduce progression to renal failure in patients with type 2 diabetes and diabetic renal disease. Unfortunately, the guidelines provided were somewhat confusing and fragmented. I believe that a simpler treatment guide can be constructed, particularly when it is emphasised that the aim in diabetes is to use agents that prevent not only renal failure but also cardiovascular events. Substantial evidence already supports a role for the angiotensin-converting enzyme (ACE) inhibitor ramipril in treatment of diabetes. The HOPE study included 3577 people with diabetes and another risk factor for cardiovascular disease who were randomised to either placebo or ramipril (10 mg) for 4.5 years.2 This group had a mean baseline blood pressure of 142/80 mmHg and no clinical proteinuria. Ramipril lowered the risk of the combined primary outcome of myocardial infarction, stroke and cardiovascular death by 25% (P ≤ 0.001), and this effect was independent of blood-pressure lowering. Furthermore, ramipril reduced progression to overt nephropathy by 22%, with a reduction evident in patients with or without proteinuria.2 This protective effect of ramipril on renal function is consistent with previous evidence that ACE inhibitors slow progression of chronic renal failure in diabetes,3,4 and that ramipril slows progression of chronic renal failure in non-diabetic nephropathy.5 Whether other ACE inhibitors have the same effect as ramipril on cardiovascular events, and what doses obtain such an effect, remains speculative. Similarly, while there is now evidence that AIIR antagonists also have renal-protective effects, there is no definitive evidence that they protect against cardiovascular events. In view of this, and in the absence of comparative trials, ramipril (and not other ACE inhibitors or AIIR antagonists) is currently the first choice for treatment for diabetic patients with hypertension, with normotension and microalbuminuria, or with hypertension and micro- or macroalbuminuria. As most patients with diabetes and hypertension require multiple agents to achieve a target blood pressure less than 130/80 mmHg, additional therapy with a β-blocker, diuretic or calcium-channel blocker is often necessary. The significance of the newly available trial data is that AIIR antagonists provide an alternative to ramipril for reducing progression to chronic renal failure in patients with diabetic nephropathy who cannot tolerate ACE inhibitors because of the side effect of cough.

Roger E Peverill PhD, FRACP

In reply: Peverill raises the question of how to integrate the new data on renal protection by angiotensin II receptor (AIIR) antagonists with existing data on cardiovascular protection by angiotensin-converting enzyme (ACE) inhibitors in patients with type 2 diabetes. An AIIR antagonist would be favoured for renal protection for a diabetic patient with hypertension and evidence of early or overt nephropathy. With regard to patients with microalbuminuria, the HOPE and MICRO-HOPE studies showed that therapy with the ACE inhibitor ramipril (10 mg/day) was associated with a 24% relative risk reduction for the development of overt nephropathy over 4.5 years.1 In contrast, treatment of similar patients with the AIIR antagonist irbesartan (300 mg/day) for 2.6 years resulted in a 70% risk reduction for the development of overt nephropathy.2 Use of an AIIR antagonist in patients with overt nephropathy has also been shown to slow progression to end-stage renal failure.3,4 As the HOPE and MICRO-HOPE studies specifically excluded such patients, evidence supporting use of an ACE inhibitor in this context is lacking. An ACE inhibitor could be used if a diabetic patient has microalbuminuria and a history of coronary heart disease or additional risk factors for cardiovascular disease, especially if normotensive. However, almost all patients with microalbuminuria require antihypertensive therapy as well as therapy to protect target organs. The relative importance of blood-pressure-lowering versus non-lowering effects of antihypertensive therapy on reducing risk of cardiovascular disease remains uncertain.5 Furthermore, a recent meta-analysis of cardiovascular protection in 62 605 patients with hypertension (including the HOPE and United Kingdom Prospective Diabetes studies) did not find that ACE inhibitors affected cardiovascular prognosis beyond their antihypertensive effects.6 Finally, it is important to note that neither ACE inhibitors nor AIIR antagonists provide total protection from cardiovascular and renal events, and that further improvements are needed for both microvascular and macrovascular protection in patients with type 2 diabetes.

George Jerums · Mark E Cooper · Richard E Gilbert · Robert C Atkins

To the Editor: The review by Brodaty and colleagues1 on drug treatment of Alzheimer's disease provides a good, concise and balanced overview. However, Pfizer takes issue with some of the referenced safety data. In Box 3 of that article (Profiles of cholinesterase inhibitors), in reference to adverse effects of donepezil (Aricept, Pfizer), it is stated that "At 10 mg/day, nausea (17% of patients), diarrhoea (17%) and vomiting (10%) may occur.49" Reference 49 at this point appears to be an incorrect citation. The figures of 17%, 17% and 10% for nausea, diarrhoea and vomiting, respectively, appear to have been taken from an article by Rogers and Friedhoff.2 It is important to note that this was a non-comparative, open-label extension study, and these incidences of gastrointestinal adverse events are inconsistent with data presented in the Australian Product Information for donepezil,3 which quote rates of nausea, diarrhoea and vomiting of 11%, 10% and 5%, respectively. These incidences are derived from a patient cohort of 1102 patents who participated in appropriately designed comparative (active and placebo) pre-registration studies of donepezil. These data have recently been confirmed in a one-year, randomised, placebo-controlled study of donepezil in patients with mild to moderate Alzheimer's disease.4 Incidences of 11.3%, 7.0% and less than 5% for nausea, diarrhoea and vomiting, respectively, are quoted in that study. We contend that, while the citation referenced by Brodaty et al was incorrect, the figures quoted for the gastrointestinal safety incidences for donepezil are also inconsistent with the Product Information and current published data.

William Lam MB ChB, PhD

In reply: We thank Lam for pointing out an error in the referencing in Box 3 of our article1 regarding the figures for adverse events for donepezil. The correct reference was number 48 in our list, not 49, and was to Rogers, Farlow, Doody et al,2 not to Rogers and Friedhoff,3 as suggested by Lam. The other references in Box 3 were given as 20, 21, 23 and 48, but should have been listed as 19, 20, 21 and 47, respectively. Secondly, issue is taken with the rates of 17%, 17% and 10% for nausea, diarrhoea and vomiting, respectively, in people taking donepezil. We agree with the overall tenor of this letter that rates of side effects are generally lower in everyday practice. The figures we quoted for adverse events are higher than the 11%, 10% and 5% cited in the Australian Product Information for donepezil,4 as the article by Rogers and colleagues2 refers to rates of adverse events experienced by those on the 10 mg dose, after a forced titration after only one week on 5 mg. We presented data for adverse events at the 10 mg dose, as this was the dose recommended for donepezil given the findings of greater benefit on the higher dose. The Australian Product Information does not indicate whether the rates of adverse events refer to the 5 mg or 10 mg dose. Usual clinical practice, which is to start with 5 mg daily and increase to 10 mg after 4–6 weeks, results in fewer adverse events. The figures of 11.3% for nausea, 7% for diarrhoea and less than 5% for vomiting presented in the Nordic study,5 in which over 80% of patients were taking 10 mg of donepezil daily, with a more flexible titration schedule, appear to be more realistic.

Henry Brodaty AO, MB BS(SYD), MD(NSW), FRACP, FRANZCP · Jane R Hecker MB BS(Hons), FRACP · John A Snowdon MPhil, MD, FRCPsych, FRACP, FRANZCP · David J Ames BA, MD, FRCPsych, FRANZCP

To the Editor: I read with interest the recent article by Olynyk's group at Fremantle on the prevalence of coeliac disease in rural Western Australia.1 It is now increasingly realised that coeliac disease is underdiagnosed in adults because it may be clinically silent — but not necessarily asymptomatic. The symptoms, however, may be non-specific and not those traditionally associated with coeliac disease. In a recently reported small study from suburban Melbourne,2 I demonstrated that about 5% of patients (5/97) undergoing gastroscopy had coeliac disease based on small-bowel biopsy results. In only one of the patients was the disease suspected clinically. I have used these figures to argue the case for routine duodenal biopsy at the time of gastroscopy, regardless of the indication. It is important to note that in my study none of the patients presenting with diarrhoea, and only one of six with anaemia, had coeliac disease — so that restricting biopsy to this group would have missed most patients with coeliac disease. Timely diagnosis is important, as symptoms may be alleviated, presymptomatic nutritional deficiencies corrected, and the risk of cancer reduced by instituting a gluten-free diet. People presenting for gastroscopy represent a high-yield group for histological screening for coeliac disease in Australia.

Jeremy Ryan FRACP

In reply: We agree with Ryan that coeliac disease is common in the Australian community, with a prevalence of 1 in 250.1 Furthermore, all individuals positive for antiendomysial antibody who undergo small-bowel biopsy have typical features of coeliac disease.1 Clearly, there is a need to increase awareness relating to coeliac disease and determine appropriate screening strategies for our population. Ryan suggests that patients presenting for upper gastrointestinal endoscopy represent a group in whom a high diagnostic yield of coeliac disease is expected. However, clinical expression of the disease is variable.1 In this setting, we believe that it is important to determine the cost-effectiveness of the various screening strategies before introducing broad-based screening.3 There is no doubt that treatment of symptomatic patients who present with coeliac disease is appropriate, but there are limited data on outcomes for asymptomatic patients who are discovered in population-based screening programs. As we stated in our article, we recommend screening by serology and small-bowel biopsy if the clinical suspicion is high or the patient is in a high-risk group.

John K Olynyk BMedSc, MD, FRACP · Digby J E Cullen MB BS, FRACP · Guy Vautier BM, MRCP · Judith A Collett MB ChB, FRACP · Dominic F Mallon MB BS, FRACP · Chris J Hovell MRCP, DM

To the Editor: We are writing in response to the editorial of Ben-Tovim et al.1 Although we agree that more research into treatment efficacy in eating disorders is needed, we believe that the study to which reference is made2 is seriously flawed. The study should not be presumed to provide evidence about the effect of treatment on outcome, particularly as the majority of patients studied received no treatment. The high death rate (3/95 [3.2%] among patients with anorexia nervosa and 2/37 [5.4%] among patients with "eating disorders not otherwise specified") in such mildly ill patients (few of whom would have warranted hospitalisation on the basis of their weight) approximates that of seriously emaciated patients in longer-term studies of treatment outcome3,4 and could more properly be said to illustrate the results of having no treatment or inadequate treatment. Exactly what constituted specialised treatment is never actually described in the original article,2 in which "extended inpatient treatment" is defined as treatment lasting more than two weeks and "extended outpatient treatment" as three or more visits. Thus, the so-called "resource intensive treatment" the authors refer to would not necessarily represent even adequate management of these conditions. In our own 6–10-year outcome study5,6 cited by the authors, 61 emaciated patients with anorexia nervosa received, on average, 11 weeks of inpatient treatment consisting of nutritional rehabilitation and psychotherapy. Only one patient died (of suicide) and, of the patients fully assessed, 41/50 (82%) had a good or intermediate outcome. The degree of weight restoration achieved by the end of treatment correlated with the degree of osteoporosis 10 years later.7 In other studies, duration of illness and early intervention have been shown to significantly influence outcome.4 This contrasts with the findings of Ben-Tovim et al,2 which may have been skewed by an unusual level of chronicity in the study group. A recent study of 69 patients with eating disorders treated in our own multidisciplinary program showed that, on 12–18-month follow-up, 48/69 (70%) had improved and 34/69 (49%) no longer had an eating disorder diagnosis. Mean levels of all but one of the major behavioural and psychological features rated by the EEE-C (Eating and Exercise Examination by Computer) instrument8 were significantly reduced. The advice given by Ben-Tovim and colleagues to the parents of the hypothetical 15-year-old girl with anorexia nervosa is regrettably nihilistic and, if based on their Lancet study,2 not founded on sound or generalisable evidence. Parents should be referred to a program for which good outcomes have been demonstrated, treatment accords with published guidelines, the clinicians are suitably experienced, and in which early intervention is the aim.4

Janice D Russell MD FRACP FRANZCP · Suzanne F Abraham PhD

In reply: The commitment of Russell and Abraham to their own program has distracted them from accurate reporting and sound epidemiological principles. They say that the majority of patients that we studied received no treatment. Not so. We clearly stated that only 34 of the 220 patients studied received no treatment.1 They then draw a range of inferences from the fact that "3/95 (3.2%)" patients with anorexia nervosa died. In fact, only 2 of 95 patients with anorexia nervosa died as a consequence of that disorder during the five years of our study. At 2.1%, this is similar to the crude death rate of 1/61 (1.6%) that they describe in their own study. However, it is only acceptable to use a rare outcome as a measure of the efficacy of a treatment program if the clinical characteristics that put patients at risk for such an outcome are known and accounted for. We do not know the specific factors that put people at particular risk of dying from anorexia nervosa. Without such knowledge, small differences in crude death rates can not of themselves inform us whether treatment programs diminish or accentuate such risks. Unfortunately, there are no other studies against which to compare the outcomes of the patients with "eating disorders not otherwise specified" in our study. We have dealt with issues such as the representative nature of our study elsewhere.2 I stand by our work and the conclusions we draw from it.

David I Ben-Tovim PhD FRANZCP

To the Editor: A 27-year-old hospital medical officer received a penetrating needlestick injury to her left hand, drawing blood, after using a 22-gauge needle to deroof a vesicle for diagnosis in a two-year-old patient with orolabial herpes simplex virus type 1 (HSV-1). The medical officer had no significant medical history, took no regular medication, and had no previous history of oral or genital herpes. On Day 4, a vesicle appeared at the site of inoculation, with surrounding erythema. The medical officer first presented on Day 6, by which time the vesicle was crusting over, with several satellite lesions (see Figure). She described mild pain in her left axilla, but no fevers or sweats. A 10-day course of oral famciclovir (250 mg, three times daily) was prescribed and she was restricted from work until the lesions had completely healed (Day 16). During 12 months of follow-up there has been no clinical recurrence. Specimens from the two-year-old child were positive for HSV-1 by direct immunofluorescence, and HSV-1 DNA was detected by polymerase chain reaction (PCR). Specimens from the medical officer on Day 6 were negative for HSV-1 by direct immunofluorescence, but positive by PCR. There was no evidence of HSV-2 or varicella zoster virus in either specimen. To our knowledge this is the first reported transmission of HSV-1 after needlestick injury. Herpetic whitlow (HSV of the hands), a well-recognised occupational hazard for dentists and anaesthetists, is frequently misdiagnosed, resulting in unnecessary surgical procedures and delayed healing. In healthcare workers, pain and also work restrictions to limit cross-infection reduce productivity. Horizontal transmission can occur in the absence of clinical lesions, but latex gloves are an effective barrier.1 There are few guidelines available for postexposure prophylaxis for HSV-1, and no controlled clinical trials in humans. The short incubation period and early establishment of latency in HSV infection remain obstacles for effective delivery of postexposure prophylaxis. HSV can establish latent infection of neurones in the absence of peripheral replication.2 In an animal model, postexposure treatment with famciclovir or valaciclovir inhibited peripheral replication of HSV, reducing latent infection but not preventing it altogether.2 In one case report, a patient who started taking famciclovir within one hour of a needlestick injury did not develop whitlow and remained seronegative for HSV.4 Famciclovir and valaciclovir have high oral bioavailability, minimal toxicity and proven efficacy in treating HSV. Available data suggest that treatment with these drugs after documented exposure to HSV reduces the severity of acute disease, limits the number of neurones infected and may reduce the frequency of subsequent recurrences. If started early enough, postexposure prophylaxis may prevent latent infection altogether.

Mark W Douglas FRACP, BScMed(Hons) · Jane L Walters MB BS(Hons), MSc, DTM · Bart J Currie FRACP

To the Editor: We report two episodes of serotonin toxicity (or serotonin syndrome) caused by drug interaction in one individual chronically treated with dexamphetamine. The interacting drugs were venlafaxine, then later citalopram. We are not aware of any previous reports of serotonin toxicity caused by dexamphetamine in combination with either venlafaxine or any selective serotonin reuptake inhibitor (SSRI). A 32-year-old man presented after two days of marked agitation, anxiety, shivering and tremor. He was being treated with dexamphetamine, 5 mg three times daily, for adult attention deficit hyperactivity disorder. He had started venlafaxine (75 mg daily) two weeks previously, and this had been increased to 150 mg daily after a week. On examination, he was alert and oriented, but diaphoretic, shivering and had fine motor tremor. His heart rate was 140 bpm, blood pressure was 142/93 mmHg and temperature 37.3°C. Pupils were 3 mm diameter and reactive, with no nystagmus or ocular clonus. There was generalised hypertonia, hyperreflexia, 1–2 beats of inducible ankle clonus, frequent myoclonic jerking and tonic spasm of the right side of his orbicularis oris muscle. His abdomen was tense, but non-tender, with normal bowel sounds. An electrocardiogram showed sinus tachycardia with a baseline tremor, but no other abnormality. Therapy with dexamphetamine and venlafaxine was ceased, and cyproheptadine (8 mg doses up to a total of 32 mg over three hours) was given. The patient had a stepwise reduction in heart rate, with complete resolution of his symptoms, and was discharged the next morning. Dexamphetamine therapy was restarted three days later and citalopram therapy was commenced one week after discharge. Two weeks after discharge, he reported similar symptoms, and ceased citalopram. Three days later he was still agitated, with nausea, diarrhoea and teeth clenching. There was no rigidity, tremor or diaphoresis, and his heart rate was 76 bpm. He was given two 8 mg doses of cyproheptadine and was asymptomatic two days later. Some of this patient's symptoms could be attributed to noradrenaline excess. However, the combination of neuromuscular and autonomic features is more consistent with serotonin toxicity. The fact the symptoms resolved after administration of cyproheptadine (a 5-HT2-receptor antagonist) supports this hypothesis. There is no theoretical reason why the interaction of citalopram (a pure SSRI) and dexamphetamine should cause catecholamine excess, and again the more likely explanation is serotonin toxicity. Dexamphetamine causes psychostimulation and increased peripheral sympathomimetic activity. Centrally it causes presynaptic release of serotonin,1 and dopamine and catecholamine release.2 Venlafaxine and its metabolite, O-desmethylvenlafaxine, inhibit both neuronal 5-HT reuptake and noradrenaline reuptake,3 whereas citalopram, an SSRI, has little effect on noradrenaline reuptake.4 The combination of serotonin reuptake blockade and either presynaptic release of serotonin or monoamine oxidase inhibition by dexamphetamine will cause increased serotonin levels in the central nervous system, and is the likely mechanism of toxicity in this patient. This is consistent with the mechanism for other reports of serotonin toxicity.5 Increased awareness and cautious monitoring is advised when using a combination of dexamphetamine and either venlafaxine or an SSRI. This is particularly important in people using amphetamines recreationally and in children taking dexamphetamine for attention deficit hyperactivity disorder.

Felicity H Prior BPharm, GradDipEpi · Geoffrey K Isbister BSc, MB BS · Andrew H Dawson MB BS, FRCP, FRACP · Ian M Whyte MB BS, FRACP FRCP

To the Editor: We report a case of bilateral acute angle closure glaucoma associated with venlafaxine. A 45-year-old woman with a history of bipolar affective disorder and borderline personality traits was admitted with increasing depression and suicidal ideation. She was taking sodium valproate (1500 mg/day) and slow-release lithium (450 mg/day). She had also taken dothiepin (50 mg nightly) for seven days, but this therapy was ceased on admission. Her only previous ophthalmological history was hypermetropia, and she had been taking low-potency neuroleptic medications and selective serotonin reuptake inhibitors in the past with no significant adverse effects. She was treated with chlorpromazine (up to 150 mg daily), and venlafaxine therapy (extended release, 75 mg/day) was commenced. After three days of taking venlafaxine, she developed left retro-orbital pain associated with nausea and vomiting, with subsequent swelling and drooping of the left upper lid and a dilated and fixed pupil. The eye was congested and visual acuity was reduced to counting fingers. She was diagnosed with acute angle closure glaucoma, treated with timolol, and transferred to a tertiary referral hospital. During this period, she sustained an injury to her right eye following an assault by a third party. A computed tomography scan revealed a right blow-out fracture with inferior rectus muscle entrapment. On admission, intraocular pressures were 16 mmHg in the right and 50 mmHg in the left eye, and gonioscopy revealed closed angles (grade 1–2). Ninety minutes after being given intravenous mannitol, topical apraclonidine hydrochloride, latanoprost and pilocarpine eye drops, the intraocular pressure dropped to 35 mmHg in the left eye. Initial laser iridotomy was unsuccessful because of a hazy cornea. Laser iridotomy was repeated several times after topical steroid therapy until it was successful. The right orbital floor fracture was repaired with a Medpor implant (Porex Surgical Products Group, Atlanta, Georgia). Eight days after starting venlafaxine therapy, she developed similar symptoms in her right eye, despite prophylactic treatment with pilocarpine eye drops four times a day. Venlafaxine was discontinued, and three days later a successful right laser iridotomy was performed. Her visual acuity was 6/5 in her right and 6/18 in her left eye. After eight weeks she was receiving no ophthalmic treatment and her intraocular pressures were well controlled. In a MEDLINE search, we found no published reports of venlafaxine associated with acute angle closure glaucoma. The manufacturers report it as a rare adverse event (fewer than 1/1000; data on file; Wyeth-Ayerst Laboratories). There has been one previous report of increased intraocular pressures in two patients with known narrow-angle glaucoma who began taking venlafaxine.2 Glaucoma has also been reported with paroxetine.3-4 Our patient had no associated family history of glaucoma, but her eyes were predisposed to angle closure glaucoma owing to hypermetropia. Patients with acute angle closure glaucoma usually have a structural defect that produces a narrow drainage angle, and thus moderate dilation of the pupil may precipitate an attack. Drugs like tricyclic antidepressants cause mydriasis and may cause the narrow angles to close as a result of anticholinergic effects. Venlafaxine, however, is a serotonin and noradrenaline reuptake inhibitor without anticholinergic activity. This fact and the time course suggest that a combination drug interaction may have occurred in this patient, perhaps by the hepatic inhibition of chlorpromazine metabolism by venlafaxine, increasing anticholinergic activity, or by a direct effect of venlafaxine on the eye unrelated to mydriasis.

Bradley Ng MB ChB · G Mark C Sanbrook MB BS, FRANZCP · Anthony J Malouf · Smita A Agarwal

To the Editor: Akathisia is a clinical syndrome that manifests as the subjective sense of unease or restlessness, or observable motor manifestations such as shuffling or tramping movements of the legs and feet, or both.1 The marked distress associated with akathisia can lead to impulsive suicide attempts.2 It is commonly associated with antipsychotic medications, as well as various antidepressants, including tricyclics and selective serotonin reuptake inhibitors (SSRIs). Mirtazapine is a novel antidepressant. It acts centrally to increase both noradrenergic and serotonergic neurotransmission. Common side effects include sedation, weight gain and increased appetite. Tremor is listed as an adverse reaction, but this does not represent akathisia. A MEDLINE database search (up to September 2001), using the words "akathisia" and "mirtazapine", did not reveal any reports of an association. However, as of mid-November 2001, the Adverse Drug Reactions Advisory Committee (ADRAC) had received five reports of "hyperkinesia (probably equivalent to akathisia)" associated with mirtazapine. We would like to report two cases of acute akathisia associated with mirtazapine. A 52-year-old man was referred by his psychiatrist for inpatient management of his depressive illness. He had previously tried multiple antidepressants, including various tricyclics and SSRIs. However, because of the sexual side effect anorgasmia, adherence to antidepressant treatment was poor. He was prescribed mirtazapine (30 mg at night). Within an hour of taking the first dose, he complained of feeling restless and unable to keep his legs still. He was given 1 mg of clonazepam, which settled his symptoms after 30 minutes. He was also observed to jiggle his legs and feet while at rest. His symptoms recurred the next day, necessitating further successful treatment with clonazepam. Mirtazapine therapy was continued, and the patient's depression improved significantly over the next few days, and the akathisia gradually resolved with regular use of clonazepam. A 73-year-old woman with chronic depression was admitted after an overdose. Her medications on admission were omeprazole, amiodarone, bendrofluazide and fluvoxamine (50 mg). The fluvoxamine was changed to mirtazapine (15 mg/day initially, increased to 30 mg/day after three days). After the first 30 mg dose, she described intense restlessness in her legs lasting up to two hours. The distress necessitated reintroducing the fluvoxamine in place of the mirtazapine. Within three weeks the patient was readmitted with depressed mood and suicidal ideation. Mirtazapine (30 mg at night) was re-introduced, with consequent acute return of restless legs. The patient's akathisia settled when the mirtazapine was reduced to 15 mg at night. No additional treatment was required. The neurobiological basis for akathisia remains unclear. Involvement of central serotonergic and adrenergic neurotransmitter systems has been postulated. One of mirtazapine's main actions is blockade of α2-adrenoreceptors. Clonidine, an α2-agonist, is effective in treating akathisia.3 We suggest that mirtazapine's adrenoceptor action might be the basis for the occurrence of akathisia in these patients.

Boregowda G Girishchandra MB BS, DPM, DipNB · Liana Johnson BPharm, MPS · Rebecca M Cresp MB BS · Kenneth G D Orr MB BS, FRANZCP

To the Editor: In a recent editorial, Mason and Diamond state that ergocalciferol (vitamin D2) is bioequivalent to cholecalciferol (vitamin D3) and that 1000 IU/day of ergocalciferol is sufficient for the treatment of vitamin D deficiency.1 Both statements are contentious. Although ergocalciferol (vitamin D2) is the only single prohormonal form of vitamin D available on prescription in Australia, there are three reasons to be cautious about the use and dose equivalence of ergocalciferol (vitamin D2) compared with cholecalciferol (vitamin D3). Cholecalciferol (and not ergocalciferol) has been shown in two randomised trials to reduce fracture rates when administered concomitantly with calcium to elderly patients.2,3 Furthermore, all recently studied agents for treating postmenopausal osteoporosis (alendronate, risedronate, raloxifene and parathyroid hormone 1-34 [the first 34 amino acids of the hormone]) were shown to lower fracture rates, but study participants were routinely given supplementary calcium and vitamin D when deficiency was established. At least two studies specified the use of cholecalciferol. Vitamin D2 (ergocalciferol) and vitamin D3 (cholecalciferol) are probably not bioequivalent.4,5 Ergocalciferol administration to vitamin-D-replete premenopausal women reduced the amount of circulating 25-hydroxyvitamin D3 (25OHD3) while only modestly increasing 25OHD2 levels, with a resultant marginal effect on the total 25OHD level.4 In contrast, the equivalent dose of cholecalciferol increased the circulating level of 25OHD3 significantly.4 Lastly, while radioimmunoassays (RIAs), such as the INCSTAR/DioSorin assay (Stillwater, Minnesota, USA), used in both studies of vitamin D levels published recently in the MJA6,7 are able to measure 25OHD levels, they are incapable of differentiating between 25OHD2 and 25OHD3. Furthermore, neither of the commercially available RIAs (the other one is made by IDS Ltd, Tyne and Wear, UK) is able to measure both vitamin D metabolites with equivalent accuracy. In a study comparing RIAs for the measurement of 25OHD against high performance liquid chromatography (the gold standard method) both assays did not recognise 25OHD2 as well as 25OHD3, with r2 of 0.74 and 0.58, respectively, for 25OHD2.8 Until further research is available, using more patients and a greater number with vitamin D deficiency, caution must be exercised in the interpretation of 25OHD levels measured with RIAs. This applies especially to individuals taking ergocalciferol (vitamin D2) for the treatment of vitamin D deficiency. Thus, it would appear that cholecalciferol (vitamin D3 ) has a role to play in the reduction of osteoporotic fractures, but only when administered with calcium. If administering ergocalciferol (vitamin D2), a far greater dose than 1000 IU/day may be needed, and the use of commercial RIAs to determine the therapeutic response may be misleading.

Paul Glendenning PhD, FRACP · Rebecca S Mason · Terrence H Diamond

In reply: Glendenning raises a number of interesting points, which require some clarification. Are ergocalciferol (vitamin D2) and cholecalciferol (vitamin D3) biologically equivalent? The statement that ergocalciferol and cholecalciferol are bioequivalent in humans is made by most authoritative textbooks, based mainly on evidence from early studies of the antirachitic efficacy of ergocalciferol and cholecalciferol compounds, and contrasts with reduced efficacy of ergocalciferol in birds and monkeys.1 Recent studies using more precise measurements have raised some doubts as to the absolute equivalency of ergo- and cholecalciferol, but the differences are marginal2 and not universally found.3 There are few recent data on relevant biological endpoints. Serum concentrations of the active hormone, 1,25-dihydroxyvitamin D, were not different after administration of ergo- or cholecalciferol,2 and increases in bone mineral density were greater after ergocalciferol therapy than after cholecalciferol in patients taking anticonvulsants.1 In short, on current evidence, differences in biological activity between ergocalciferol and cholecalciferol are likely to be relatively minor. Are there problems monitoring therapy? 25-Hydroxyvitamin D values may be used for monitoring treatment. The possibilities of impaired detection of the 25-hydroxy metabolite of ergocalciferol by some assays,2 and perhaps a smaller rise in total 25-hydroxyvitamin D concentrations after low doses of ergocalciferol, should be borne in mind when monitoring therapy. What is the current recommendation for vitamin D supplementation? While the availability of larger dose sizes and/or cholecalciferol preparations would be helpful, 800 IU of ergocalciferol and 1 g of calcium for six months was shown to reduce secondary hyperparathyroidism in older patients,1 and 600 IU/day (same for ergo- and cholecalciferol) is the new recommended adequate intake for older patients with limited sun exposure.1

Paul Glendenning

To the Editor: A recent editorial in the Journal focused on the "blurring of research ideals and corporate interests".1 But there are other funding and commissioning bodies, including government, whose wants or needs also have the potential to blur research ideals and exert control over what can be published. In recent times, those who pay the piper increasingly want to call the tune. Understandably, this is also an issue for research into Aboriginal ill health.2,3 Van Der Weyden's plea for the development of national guidelines on institutional conflict of interest should therefore be broadened to include all funding bodies. One suggestion for inclusion in these guidelines, to enhance public interest in research, is an obligation for authors to state not only their sources of funding, but "the origin of the research question they are attempting to answer"4 and the person or group who initiated the funding of the project.

Max Kamien

To the Editor: We report the appearance of erythromycin and inducible clindamycin resistance in the south-west Pacific strain of non-multiresistant methicillin-resistant Staphylococcus aureus, which has recently appeared in eastern Australia. Infections occur predominantly in Polynesian people and are usually community-acquired. Most strains belong to Western Samoan phage patterns (WSPP1 or WSPP2) and pulsotype A when typed by pulsed-field gel electrophoresis.1,2 These strains are resistant to all β-lactams, but are usually susceptible to erythromycin, clindamycin, gentamicin, tetracycline, trimethoprim–sulfamethoxazole and ciprofloxacin. Although most of these antibiotics would not be recommended for therapy,3 clindamycin has been recommended for non-parenteral treatment of soft-tissue and bone infections, as it is efficacious in treating similar infections caused by methicillin-susceptible S. aureus.4 Twenty isolates of community-acquired, non-multiresistant pulsotype A MRSA were collected from patients from southern Brisbane and Logan in 1997 and 1998.2 A further 16 isolates were obtained from Ipswich patients between December 1998 and February 2001. We found that all 36 isolates were susceptible in vitro to gentamicin, tetracycline, trimethoprim–sulfamethoxazole, ciprofloxacin, rifampicin, fusidic acid and vancomycin. However, two of the Ipswich isolates had erythromycin and inducible clindamycin resistance. When susceptibility testing was performed using standard disc methods, both isolates appeared resistant to erythromycin but susceptible to clindamycin. However, on testing for inducible macrolide resistance (MLSB phenotype) using a disc-approximation method, both showed inducible clindamycin resistance.5 These isolates were from superficial abscesses in Polynesian people with community-acquired infection. They were indistinguishable by pulsed-field gel electrophoresis, but there were no epidemiological links. As yet, we have found no community-acquired pulsotype A strains of MRSA with erythromycin resistance and constitutive (ie, non-inducible) clindamycin resistance. Two other recent Australian studies found erythromycin resistance in 13 of 153 and three of 29 isolates of non-multiresistant MRSA, respectively.3,6 These studies did not report clindamycin susceptibilities, and it was not clear what proportion of the isolates belonged to phage patterns WSPP1 or WSPP2, or were community-acquired. As clindamycin has been recommended as a therapeutic option for soft-tissue and bone infections caused by non-multiresistant MRSA, this finding of inducible clindamycin resistance has important implications. Microbiology laboratories should screen for inducible clindamycin resistance in erythromycin-resistant strains, and, if found, an alternative antibiotic should be used for treatment.7 Alternatively, rather than assessing inducible clindamycin resistance with a disc-approximation test, some laboratories may prefer to report all erythromycin-resistant strains as clindamycin-resistant. Also, given the increasing incidence of community-acquired MRSA infection in Australia, all suspected staphylococcal infections that are not responding to empiric therapy with β-lactam antibiotics should be swabbed for culture.

Wendy J Munckhof MB BS, FRACP FRCPA, PhD · Jacqueline Harper BSci (Hons), PhD · Jacqueline Schooneveldt BAppSci, MAppSci, GCM · Graeme R Nimmo MB BS, MSc MPH, FRCPA

To the Editor: We strongly support the proposal1 that it is time to create a special grant category for curiosity-based research proposals. Having been in biomedical research for over 50 years, we have experienced a period when most research was curiosity based. We can thus compare with the present situation — research aiming for a rapid, practical and commercial outcome has become almost a necessity for survival because of the increasingly severe reduction in government funding for universities and research institutes. We would like to illustrate the value of curiosity-based research with a few Australian examples from our own fields. In 1946, one of us (F F) was working on the experimental epidemiology of the causative agent of infectious ectromelia of mice (related to vaccinia virus). A chance observation — that the mice which survived the infection developed a skin rash — led to further study of the virus as a model for smallpox, measles and chickenpox infections. Thus, unexpected discoveries were made about the way the virus spreads through the body during the incubation period of these diseases.2 In 1951, myxomatosis spread in rabbits in the Murray–Darling basin of south-eastern Australia. The virus was initially extremely virulent (99% fatal), but the rabbits slowly developed genetic resistance. One of us (F F) studied the virus for 15 years, and this work was acknowledged as the best example of the co-evolution of viral virulence and host resistance.3 In 1957, Macfarlane Burnet proposed the clonal selection theory of antibody formation — that individual B lymphocytes made antibody of a single specificity.4 This was one of the most original concepts ever proposed in biology and it took 10 years to be widely accepted. It has since led to the production of monoclonal antibodies, which are used as basic reagents in research and diagnostic laboratories, and are now being used in immunotherapy. In the 1970s, Peter Doherty and Rolf Zinkernagel studied the role of the newly discovered cytotoxic T cells (which could lyse virus-infected cells) to find out how T cells recognised the infected cells. They showed that killing was restricted by the major histocompatibility complex (MHC) and that its role was to signal "altered self" to the T cell.5 These studies led to the award of the Nobel Prize in 1996. Cytotoxic T cell activity has since been shown to be the main immune mechanism for controlling and clearing many intracellular infections. Induction of a strong cytotoxic T cell response is the mechanism of candidate vaccines currently being trialled against HIV-1. In the late 1960s, one of us (G A), together with Chris Parish, showed that a bacterial protein, flagellin, induced antibody tolerance over a wide dose range. Parish was the first to show the inverse relationship between antibody and cell-mediated immune responses, which led others to describe two classes of helper T lymphocytes.6 These have been shown to be important in the development of allergy in infants, and offer the opportunity for immunotherapy to reduce the later incidence of allergy. None of these research programs was initiated with a commercial goal in mind. Benjamin Franklin, when asked about the importance of some research, replied "Of what use is a baby?".

Gordon L Ada AO, DSc, FAA · Frank Fenner MB BS, MD

To the Editor: There is much conflicting evidence that increased intake of vitamin C enhances the natural protective mechanisms of the body and decreases both the incidence and severity of the common cold.1 It is regrettable that the study by Audera and colleagues failed to show a significant therapeutic effect of megadose vitamin C in treatment of the common cold.2 The groups compared had, on average, similar composition after randomisation. However, the viral infections that cause the common cold and its progression to ill health, as evidenced by multiple symptoms, are affected by many factors, while symptom severity is well known to vary greatly. Therefore, the study's reliance on respondents' self-diagnosis of symptom severity and onset is a significant weakness in design. Randomisation of participants to the treatment groups may have been insufficient to override this design deficit, thereby significantly biasing the outcome. A better design might have combined patient self-report of symptom severity with physical examination, thus allowing independent and professional assessment of severity. Also, proper assessment of previous history of severity of cold symptoms is crucial for proper randomisation to treatment groups. If Audera and colleagues' study failed to control for this history, then randomisation may have also failed to balance its effect equally between treatment groups, significantly compromising the study's validity to detect any therapeutic benefit of vitamin C. Cold symptoms also vary diurnally, while severity varies with alcohol use and smoking status,3,4 which also affect vitamin C absorption.5,6 No information was provided on study participants' alcohol consumption and smoking status. Finally, the study did not assess stress, which may constitute a further, important uncontrolled bias. A recent cohort study of stress and the common cold concluded that all four dimensions of stress investigated — stressful life events, negative affects, positive affects and perceived stress — were significantly related to occurrence of the common cold.7 Stress may also have significantly affected symptom severity and participants' perception of their symptoms. Certainly, the trend observed in the placebo group of shorter duration of some symptoms and lower mean severity could have been due to less severe symptom history, compounded by a lower degree of overall stress.

Luis Vitetta · Avni Sali · Bill Paspaliaris · Nicola J Reavley

In reply: Precisely because of the temporal variation in symptom severity described by Vitetta and colleagues, we judged that medical professionals are not as well able, in a variably timed interview, to quantify patients' cold symptoms as the patients themselves can do on a continuing basis. Therefore, we consider that our study1 would have been no more valid if the detailed symptom severity cards had been supplemented by one or more physical examinations. In that respect, we are in good company with others who have studied the common cold over many years.2 We agree that double-blind randomisation does not necessarily distribute all relevant variables equally. That is why, in Box 2 of our study report, we presented four variables — age, sex, mean number of colds in the previous year, and mean number of days unwell with colds in the previous year.1 The likelihood that stress, smoking and alcohol status would have been sufficiently maldistributed in this large group to mask a significantly beneficial effect in even one of the three groups which received high-dose vitamin C seems vanishingly small. Nevertheless, we acknowledge that the study would have been stronger if we could have reported the distribution of these three potential confounders. We contest the view of Vitetta and colleagues that the evidence from randomised controlled trials of vitamin C in treating the common cold conflicts significantly (see Box 1 of our article1). The overview finding — that mega-doses of vitamin C for prophylaxis produce a relatively trivial reduction in cold severity but no reduction in incidence3 — was the stimulus for our own study. No community studies of this issue have been flawless, but the mounting collective evidence suggests that we should look elsewhere for a cold panacea.

Carmen Audera · Roger V Patulny · Beate H Sander · Robert M Douglas