| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
|
|
|
|||||||
|
|||||||||
TRACK to:
|
|
Electronic letters published:
![[Read Comment]](/icons/misc/sectionDOWN.gif){kind=icon}
Some thoughts
Clinical diagnosis of bacterial rhinosinusitis
Slay the dragon too?
Length of symptoms
duration of illness for sinusitis
Power problems
The quest for the Holy Grail
Answer to "Van Duijn work in 1992"
Van Duijin work in 1992
|
|
|||
|
Shelley Chang, Cleveland, Ohio Student, Case Western Reserve University, School of Medicine.
Send response to journal:
|
Thank you for all your insights. Our journal club had the pleasure of reading the De Sutter article for our last meeting. Here are some of my personal thoughts while reading the De Sutter article and some noted from jounal club discussion : Being a secondary analysis of data from a clinical study in which patients were randomized to be comparable between treatment groups, there is a need to ensure that the select sampling (based on symptoms of acute sinusitis and other loss to follow ups) is still comparable between main exposure groups. A comparison of patients who were excluded or included was reported. However, basic characteristics of patients from each arm of the main exposure (ie. placebo vs treatment group, or other main signs and symptoms under consideration) were not included in the article. The article also did not report the numbers of those who discontinued for various reasons according to treatment and symptoms. In the original study, 7 (3.4%) patients from placebo group and 1 (0.5%) patient from amoxicillin group withdrew before day 10 due to exacerbation of symptoms (RR 0.25, 95% CI 0.04-1.56, P=0.07). It is difficult to conclude whether differential loss to follow up is a problem and whether baseline measure of general feeling of wellness may be associated with treatment arms. Table 3 in the original article reported a greater mean reduction of symptoms in 10 days between treatment groups for thick nasal discharge. There is also a marginally significant reduction of waking up tired (P=0.09). Other factors reported included a reduction in sadness (P=0.18), reduced concentration (P=0.19), and reduced pain in upper teeth when chewing (P=0.17). On the other hand, the change in level of fatigue (P=0.38)and reduced productivity (P=0.29) were less significant. It may be interesting to analyze the present study using the other signs and symptoms for which treatment effect was more significantly observed. The outcome in the present study is defined as the patient indicated in his or her diary feeling generally ¡§well¡¨ again as noted in the diary. This outcome differs from the outcomes studied in the original clinical trial which used the disappearance of symptoms that the patient identified as most greatly affecting their health at baseline. It may also be valuable to analyze the present study using the 2 other outcomes considered in the larger clinical study. These include all physical signs disappeared at day 10 and all symptoms indicated as most important item affecting health at baseline disappearing by day 10 which may potentially have better precision than measure of general wellness based on diary. It may also be appropriate to consider the change in general feeling of wellness or reduction in signs and symptoms. I thought it would helpful for interpreting the data if some raw and summary data of the major exposures and outcomes of interest may be reported in addition to complex statistical tests. In multivariate analysis, there may be grounds to include variables based on clinical relevance. In particular, it would be interesting to see whether adjusting for treatment, reduced productivity, and general feeling of illness would change the hazards of other factors either qualitatively or quantitatively. The baseline exposure of reduced productivity and general feeling of illness are both subjective. One may argue that the outcome measured in this case, feeling ¡§well¡¨ again, is also subjective. Therefore, to detect for the true treatment effect, it may be necessary to adjust for general feeling of wellness at baseline since they may be potential confounders of the treatment effect even if they may not interact with treatment. The time dependent nature of the illness may be problematic for studies. Over the course of illness, symptoms may disappear while new ones emerge within a few days. A variety of symptoms may present sequentially or in waves in a time dependent manner. Future studies might also consider the emergence of new signs and symptoms or exacerbation of existing ones in addition to the reduction of the symptoms observed at recruitment. It may be clinically important to adjust or stratify by the duration of illness at randomization or perhaps a comprehensive assessment of the history of illness and include time dependent variables. The authors included the variable ¡§complaints greater or equal to 7 days¡¨ in univariate analysis, but this variable was not adjusted for during multivariate analysis. An alternative method may be to use delayed entry time-to-event analysis if the exact number of days since illness began is available as a continuous variable. One of the main goals of this study was to determine whether treatment effect varies according to patients presenting with particular signs and symptoms. If differences in treatment effects across strata are reported, it will be interesting to note the magnitude and direction of differences even if they may not be all statistically significant due to low power to detect interaction. From a social and public health perspective, those who are sent home without antibiotics should be convinced to return to the doctor immediately if conditions do worsen. If a physician tells the patients that it is probably viral and sends them home with medications treating symptoms, the patient may potentially be less likely to return since they have just been assured by the doctor that it is viral. Treating for symptoms alone may mask more serious complications. For some patients, repeated visits to the physician may also be very costly in time and money. One alternative may be to give patients a delayed prescription of antibiotics. It seems there may be no magic bullets in predicting who needs a prescription. Hopefully, in the near future, an algorithm tree incorporating signs and symptoms, detailed history, biomarkers, and culture can be developed to make use of all these tools in a cost-effective manner. Competing interests: None declared |
|||
|
|
|||
|
Jay C Smith, Pembroke, NH M.D., New Hampshire Hospital
Send response to journal:
|
De Sutter’s study reinforces the clinical difficulty in differentiating viral from bacterial infection. It clearly shows that >1 week of purulent drainage doesn’t have predictive value for response to antibiotics. Unfortunately, it didn’t study, as a distinct group, the 3 or more signs and symptoms that together have been shown to have predictive accuracy. Their study included many with just 2 of these, increasing the proportion of viral rhinosinusitis episodes in their sample. Using >5 mm of mucosal thickening on X-ray to define greater likelihood of bacterial infection actually also finds a group more likely to have viral rather than bacterial infection. Plain X-rays in 98 adults with self-defined colds found 39% had >5mm thickening at day 7 and almost 90% were normal two weeks later without use of antibiotics [1]. With respect to a more predictive “3 of 4” (unilateral pain predominance, unilaterally predominant purulent rhinitis by history, bilateral purulent rhinitis by history, purulence on exam of nose) [2], the study would lack their stated estimated power of 0.80 at an alpha of 0.05 to find a 15% recovery rate difference. Not analyzing this subset is puzzling, having cited this ENT study as a trial of high accuracy along with a review [3] that included general practice studies that refer to similar groups of criteria but looked at “double sickening” in the selection of patients (but didn’t differentiate unilateral and bilateral purulence by history). Those studies also found that measuring ESR or CRP was more predictive than individual signs and symptoms. It is an important insight that these groups of clinical data can predict which patients are likely to have maxillary sinus opacification or air-fluid level, X-ray findings that puncture studies have shown to be associated with bacterial sinusitis (sensitivity 73%, specificity 80% for bacteria in sinus aspirate) [4]. The predictive power relative to the value of antibiotic treatment still needs to be directly tested but a guideline based on this seems appropriate to me to guide antibiotic use for acute rhinosinusitis (until further studies are done). I use presence of at least 3 of the following 4 clinical data: A) history of purulent rhinorrhea, B) purulent secretions in the nasal cavity, C) unilateral predominance of pain and D) “double sickening”. If the patient has only has 2 of these 4, ESR elevation can be helpful in predicting bacterial infection as well. I base this on a conflation of results from the general practice and ENT trials referenced. It is also based on the belief that most patients would be happier if they did not need to have, and wait for the results of, a blood test. Further study is definitely needed as my guideline may overuse antibiotics for some viral and minor bacterial infections that would improve just as fast without them. 1. Puhakka T, Makela MJ, Alanen A, et al. Sinusitis in the common cold. J Allergy ClinImmunol,1998.102(3): p. 403-8. 2. Berg O, Carenfelt C. Analysis of symptoms and clinical signs in the maxillary sinus empyema. Acta Otolaryngol, 1988. 105(3-4): p. 343-9. 3. Lindbæk M, Hjortdahl P. The clinical diagnosis of acute purulent sinusitis in general practice — a review, British Journal of General Practice, 2002, 52, 491-495. 4. Lau J, Zucker D, Engels E, et al. Diagnosis and Tretment of Acute Bacterial Rhinosinusitis, Evidence Report/ Technology Assessment (Agency for Healthcare Policy and Research), 1999, March(No. 9): p. 1-249. Competing interests: None declared |
|||
|
|
|||
|
Jim Young, Basel, Switzerland Institute for Clinical Epidemiology and Department of Internal Medicine, University Hospital Basel
Send response to journal:
|
John - thanks for your comments and for your support. This collaborative approach really grew out of Dan Merenstein’s trial. This trial was designed to test the 7 days of symptoms criterion but funding constraints restricted patient recruitment [1]. So our initial idea was to consider support for this 7-day rule across all participating trials. Later we decided to extend our analyses to consider support for other signs and symptoms. A related symptom is a prior common cold or a two- phase illness. Not all trials collected information of this sort, but we hope to get data from at least four trials so we can consider this symptom too. Your comments raise an important point. The choice of 7 days is quite arbitrary. This means we should look to see if benefit from antibiotics increases as prior duration of illness increases. In statistical terms, we need to represent prior duration of illness not just as a binary covariate (more or less than 7 days) but as a continuous covariate as well. This second method is typically the more powerful. I’ll make sure we do this. Because such subgroup analyses are always exploratory, any significant result should be interpreted with caution. If we are able to identify a patient subgroup that appears to benefit from antibiotics, logically the next step is to try to confirm this finding in a multi- centre RCT. So I think your reward is safe for some time to come. On the other hand, I don’t think we should give up just yet. 1. Merenstein D, Whittaker C, Chadwell T et al. Are antibiotics beneficial for patients with sinusitis complaints? A randomized double- blind clinical trial. J Fam Pract 2005;54:144-51. Competing interests: None declared |
|||
|
|
|||
|
Alan L. McGaughran, MD, Saltsburg and Latrobe, PA, USA Family Physician, Latrobe Area Hospital, Excela Health
Send response to journal:
|
In the analysis of severity of symptoms, length of symptoms was "bundled" with other symptoms rather than separated out. The time to recovery was measured from the time of physician contact and not from the time of onset of symptoms. (Obviously, the former is more objective and the latter more subjective.) However, if there was a subset of patients with a longer length of symptoms before presentation that might have benefited from antibiotics, it would have not been detected in this evaluation, since the time to recovery was measured from the time of physician contact. (In other words, the total length of symptoms could have been reduced, even if the time to recovery beginning from the physician contact was not significantly changed.) Competing interests: None declared |
|||
|
|
|||
|
John Hickner, Chicago, IL family physician, U of Chicago
Send response to journal:
|
I should qualify my statement about duration of illness predicting bacterial sinusitis, and explain the reason the 7-day duration of illness was included in the CDC sinusitis management guidelines. There is no solid research evidence that sinusitis-like illness of duration greater than 7 days is more likely to be bacterial sinusitis. The panel of physicians that developed the CDC guidelines, however, felt that the 7 day guideline would help to limit unnecessary antibiotics for sinusitis because many patients would be improving spontaneously by that time. For that reason, in the final writing I did agree to including duration of illness in the treatment guidelines. The problem with this approach, besides the lack of experimental evidence, is that many viral upper respiratory infections last for 14 to 21 days. This is the reason I urge you to include duration of illness in your analysis. Competing interests: None declared |
|||
|
|
|||
|
John Hickner, Chicago, IL family physician, U of Chicago
Send response to journal:
|
Jim, thanks for your insightful comments about the power issue. I am delighted that you are joining forces to tease through this issue. Mort Lindbaek and I considered such a study a few years ago, but for a different purpose; to determine if length of illness had anything to do with antibiotic responsiveness. The CDC guidelines for treatment of rhinosinusitis, which I authored, include a caveat about waiting for 7 days. It was against my wishes that this was included. I am not convinced that length of illness has anything to do with it. So, while you are at it, please consider length of illness so that we might slay that dragon, too. Guess I had better start a second retirement fund if you are getting close to finding the Holy Grail of sinusitis! (Just watched the Da Vinci code for the first time this weekend.) John PS Hi An; keep up the great work! Competing interests: None declared |
|||
|
|
|||
|
Jim Young, Basel, Switzerland Institute for Clinical Epidemiology and Department of Internal Medicine, University Hospital Basel, Beat Müller, Heiner C Bucher
Send response to journal:
|
We read De Sutter’s study [1] and Hickner’s latest editorial [2] with great interest. In an earlier editorial, Hickner concludes that either we need a new test that can differentiate bacterial from viral rhinosinusitis or we need to identify a set of signs or symptoms that can do this [3]. Now he concludes that this second strategy is unlikely to succeed given the results of De Sutter’s study. The trouble with this study and our study too [4] is that both lack the power to detect useful differences in the effect of antibiotic therapy between patient subgroups. Consider the confidence intervals in de Sutter’s Table 3. Take purulent rhinorrhea for example; long thought to be a classic symptom of bacterial infection. The 95% confidence interval (hazard ratio 0.67 to 1.84) is so wide that it cannot rule out say a ratio of 1.50. Such a ratio, if found, would suggest that antibiotics provide more than trivial benefit for those patients with this symptom. All 15 variables in Table 3 could have a hazard ratio of 1.5; 5 variables could have a hazard ratio of 2.0. This lack of power is to be expected in secondary analyses using data from clinical trials that were powered to detect a difference between two treatment groups. A possible solution is a meta-analysis using individual patient data. De Sutter’s group and ours have joined forces together with a number of other groups that have run similar trials where patients with clinically diagnosed sinusitis were randomised to either an antibiotic or placebo. Collectively we will have at least 1500 patient from 7 trials. These numbers are no guarantee of success. First the increase in patient numbers in a meta-analysis does not necessarily even compensate for between trial variability. Second, as De Sutter writes, a clear-cut sign or symptom may simply not exist. Nevertheless, this collaborative approach will provide the best evidence yet for whether Hickner’s second strategy has any chance of success. A new test is the other strategy. In this context, procalcitonin shows promise as a marker for bacterial infection [5] and we have just completed a trial in primary care where procalcitonin was used to guide antibiotic therapy in patients with acute respiratory tract infections [6]. So the search for Hickner’s Holy Grail continues [3] and the one million dollar reward he now offers should give added incentive to those trying to develop an affordable near-patient test for procalcitonin. 1. De Sutter A, Lemiengre M, Van Maele G et al. Predicting prognosis and effect of antibiotic treatment in rhinosinusitis. Ann Fam Med. 2006;4:486-493. 2. Hickner JM. A new look at an old problem: inappropriate antibiotics for acute respiratory infections. Ann Fam Med. 2006;4:484-485. 3. Hickner JM. Acute sinusitis, antibiotics, and the Holy Grail. J Fam Pract. 2005;54:152-153. 4. Young J, Bucher H, Tschudi P et al. The clinical diagnosis of acute bacterial rhinosinusitis in general practice and its therapeutic consequences. J Clin Epidemiol. 2003;56:377-384. 5. Christ-Crain M, Jaccard-Stolz D, Bingisser R et al. Effect of procalcitonin-guided treatment on antibiotic use and outcome in lower respiratory tract infections: cluster-randomised, single-blinded intervention trial. Lancet. 2004;363:600-607. 6. Briel M, Christ-Crain M, Young J et al. Procalcitonin-guided antibiotic use versus a standard approach for acute respiratory tract infections in primary care: study protocol for a randomised controlled trial and baseline characteristics of participating general practitioners [ISRCTN73182671]. BMC Fam Pract. 2005;6:34. Competing interests: Beat Müller has received payment from BRAHMS (a manufacturer of procalcitonin assays) for travel expenses, speaking engagements, and research. |
|||
|
|
|||
|
An De Sutter, Belgium Department of General Practice and Primary Health Care
Send response to journal:
|
The purpose of our study was different from the study of van Duyn. van Duyn looked for clinical symptoms predicting the presence of fluid in the maxillary sinuses on ultrasonography. We looked for clinical symptoms predicting the prognosis or effect of antibiotics. The presence of fluid in the sinuses does not necessarily mean that there is a bacterial infection that will respond to antibiotics or influence the prognosis. Actually, I think we must be careful in using the five symptoms of van Duyn to diagnose acute sinusitis because ultrasonography is not a good golden standard for fluid in the sinuses (Engels EA et al. J of Clin Epidemiology 2000;53:852-862.) and the 5 symptoms have not been confirmed in subsequent diagnostic trials. Competing interests: None declared |
|||
|
|
|||
|
James Alan Cave, Newbury, Berkshire General Practitioner
Send response to journal:
|
I am surprised at the findings of these authors given van Duijin, Brower and Lamberts work published in 1992 and Frank Dobbs and Douglas Flemings subsequent letter in the BMJ. There work seems to provide 5 useful symptoms predictive for maxillary sinusitis. Van Duijin NP BMJ 1992;305:684-7 Dobbs F & Fleming D (letter) BMJ 1992;305 1435 Competing interests: None declared |
|||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH |
TRACK: