Recently the Journal published ‘A simple index to predict prognosisindependent of axillary node information in breast cancer’ bySeshadri et al.1 The authors derived a prognostic index on aretrospective data set of 736 patients followed up for a median of65 months. They claim that this useful index separates patients intogroups with different outcomes.

In their introductory paragraph, Seshadri et al. discuss thewell-established Nottingham Prognostic Index (NPI),2,3 an indexderived from an initial study of 387 patients, tested prospec-tively and shown to be highly reproducible4 on 1168 patients andnow reported on over 3000 patients with a median follow-up ofover 10 years.5 The index has been confirmed by a number ofcentres, and its accuracy proven in a major study by the DanishCooperative Breast Group on over 9000 patients.6

Seshadri et al. criticized the NPI on several grounds:1. That the ‘scoring system is complex’.To compile the NPI, Nodal Stage (measured 1–3) is added to

Grade (1–3) and to size (cm× 0.2). Although this calculationadmittedly requires the ability to count up to three, to add threesimple numbers together and to multiply by two, we have had amultiplicity of Australasian surgeons working in our unit whohave managed it!

2. ‘Knowledge of lymph node status is required’.This is gained in Nottingham by a simple sampling procedure

that carries no long-term morbidity. Without this knowledgegroups likely to fare very badly cannot be differentiated. Thus whilethe NPI selects a group with 80% mortality by 7 years, the worstgroup selected by the Adelaide index is predicted to have 50% sur-vival by 7 years, even though significant numbers in the Adelaideseries have died by this time.

3. Elston/Ellis ‘Grading is complex’.Elston and Ellis have been able to establish the prognostic

value of grading in their prospective study.7 In multivariateanalysis this makes as powerful a contribution to prognostic dis-crimination as does lymph node (LN) stage.

Confirmation of the reproducibility of grading in this way hascome from the Danish series previously mentioned (without repro-ducible grade the NPI would not have been reproducible) and fromtwo international studies8,9 with the conclusion that grading isreproducible if pathologists take the trouble to learn the technique.

The requirements of prognostic index are:1. That it must separate out groups widely differing in prognosis,

from the very good to the very bad.Prediction of survival at 10 years by the Adelaide Index (AI)

ranges only between 42 and 75%. This is not as good as the rangeof prediction achieved by a single parameter, such as lymph nodepositivity (25% for positive; 65% for negative). The NPI range(Fig. 1) lies between 95% (excellent prognostic group) and 15%survival (very poor prognostic group) at 10 years.

2. An index must allocate reasonable percentages to each groupselected. For example tumour size alone appears to be a useful

prognostic factor until it is realized that only small percentages oftumours lie in the small and large groups; size accords the overallaverage prognosis to the great majority.

The AI does not satisfy this requirement, for in the four groupsselected the highest risk group has only 8% of patients. Thus 92% of patients lie within a range of predicted survival at 5 years ofbetween 70 and 90%, not usefully separating them from the overallaverage of around 75%.

3. The index must be tested prospectively.The index was derived by retrospective analysis of the data set.

Factors identified as of prognostic value were then re-applied tothe same data set in the AI and under this circumstance it isentirely predictable that any retrospectively derived index willappear to perform well.

4. The parameters must be simply measured in order that theindex may be used in other centres and its validity must be confirmedby other centres.

5. The index must be shown to hold up over a long period of time.The NPI has been demonstrated to be valid for estimating survivalover a 15-year period.3 At present the AI is derived from a median 65months of follow-up.

The AI does not appear to satisfy any of these criteria. It doesnot perform any better in discrimination of prognosis than do anumber of single factors (lymph node status, size, grade, c-erb-B2).

One of the factors used in the AI is oestrogen receptor (ER). In theabsence of hormone therapy ER does not emerge as an independentprognostic factor on multivariate analysis (if grade is included).However, if all women are placed on adjuvant Tamoxifen thenER achieves independent prognostic status. For example, in theBirmingham Prognostic Index,10 derived in lymph node positivetumours and in which study all women received adjuvant Tamoxifen,ER is a factor predictive of the lengthening of survival produced bythe adjuvant Tamoxifen.

Aust. N.Z. J. Surg. (1999) 69, 67–68

COMMENT

PREDICTION OF SURVIVAL IN THE INDIVIDUAL WITHBREAST CANCER

ROGERBLAMEY

Department of Surgery, Nottingham City Hospital, Nottingham, UK

Fig. 1. Separation of 3083 patients into groups by the NottinghamPrognostic Index. A wide separation is seen, with a reasonable per-centage of cases distributed to each group.

Prognostic factors give prediction of survival without consideringresponse to systemic therapies and thus identify those in need of suchtherapy. Predictive factors identify whether hormonal or cytotoxicagents may be effective in those identified.

Strict criteria must be applied before a prognostic index can be con-sidered validated. The AI is unconfirmed prospectively and nottherefore ready for clinical use. At this stage it is not promising inits ability to discriminate prognosis sufficiently well in comparisonwith other indices or even with single prognostic factors. Otherindices have been published with better separation than the AI,including one from an Australian centre.11 The NPI produces the bestseparation and is the only prospectively and inter-centre confirmedpublished index at present that fulfils the criteria for an indexcapable of clinical use.12

REFERENCES1. Seshadri R, Horsfall DJ, McCaul K et al. A simple index to

predict prognosis independent of axillary node information inbreast cancer. Aust. N.Z. J. Surg.1997; 67: 765–70.

2. Haybittle JL, Blamey RW, Elston CW, Johnson J, Doyle PJ,Campbell FC. A prognostic index in primary breast cancer. Br. J.Cancer1982; 45: 361–66.

3. Galea MH, Blamey RW, Elston CW, Ellis IO. The NottinghamPrognostic Index in primary breast cancer. Breast Cancer Res.Treat.1992; 22: 207–19.

4. Todd JH, Dowle C, Williams MR et al. Confirmation of a prog-

nostic index in primary breast cancer. Br. J. Cancer1987; 56:489–92.

5. Blamey RW. The design and clinical use of the NottinghamPrognostic Index in breast cancer. Breast1996; 5: 156–7.

6. Balslev I, Axelsson CK, Zedeler K, Rasmussen BB, Carstensen B,Mouridsen HT. The Nottingham Prognostic Index applied to9149 patients from the studies of the Danish Breast CancerCooperative Group (DBCG). Breast Cancer Res. Treat.1994;32: 281–90.

7. Elston CW, Ellis IO. Pathological prognostic factors in breastcancer. I. The value of histological grade in breast cancer: Experi-ence from a large study with long-term follow-up. Histopath.1991; 19: 403–10.

8. Robbins P, Pinder S, de Klerk N et al. Histological grading ofbreast carcinomas: A study of interobserved agreement. Hum.Pathol.1995; 26: 873–9.

9. Frierson HS, Wolber RA, Berean KW. Interobserver repro-ducibility of the Nottingham modification of the Bloom andRichardson histological grading scheme for infiltrating ductalcarcinoma. Am. J. Clin. Pathol.1995; 105: 195–8.

10. West Midlands Breast Group. Pathological and immunohisto-chemical prognostic factors in clinical stage I breast cancerpatients. Breast1997; 6 (Abstr.): 242.

11. Bryan RM, Mercer RJ, Bennett RC, Rennieg GC. Prognosticfactors in breast cancer and the development of a prognosticindex. Br. J. Surg.1986; 73: 267–71.

12. Clark GM. Integrating prognostic factors. Br. Cancer Res. Treat.1992; 22: 187–91.

68 COMMENT AND REPLY

Aust. N.Z. J. Surg. (1999) 69, 68–69

COMMENT AND REPLY

SINGLE- VERSUS MULTIPLE-DOSE ANTIMICROBIAL PROPHYLAXISFOR MAJOR SURGERY: COMMENT

J.C. HALL

Royal Perth Hospital, Perth, Western Australia, Australia

McDonald et al. have recently published a systematic review ofantimicrobial prophylaxis for surgery.1 Their conclusion reaf-firms the conventional wisdom that ‘continued use of single-doseantimicrobial prophylaxis for major surgery is recommended’.However, they have failed to define the limits of this generalization.A particular problem is that the authors have regarded all formsof surgery as being equivalent, when there is substantial evi-dence that patients may require more than a single dose of ashort-acting agent if they undergo lengthy procedures, bleedheavily, have poor local tissue perfusion, or acquire a prosthesis.

McDonald et al. comment that ‘more studies may be requiredbefore the conclusions of Hall et al. become the basis for pre-scribing guidelines in vascular surgery’. However, the clinicaltrial in question found in favour of a multi-dose regimen forpatients who undergo vascular surgery,2 which is in accord withconventional surgical practice. It seems that McDonald et al.are promoting short-term prophylaxis for vascular surgery on

the basis of the results of clinical trials that relate to other types ofsurgery. Incidentally, I am not aware of any reputable prescribingguidelines that strongly promote the use of short-term prophylaxisfor patients undergoing vascular surgery.

It seems that McDonald et al. have forsaken biological andclinical plausibility in their haste to climb aboard the data bargeof meta analyses and structured reviews. They might have presenteda more balanced review by collaborating with a surgeon—there is noevidence of this in the author list or the acknowledgements.

REFERENCES1. McDonald M, Grabsch E, Marshall C, Forbes A. Single- versus

multiple-dose antimicrobial prophylaxis for major surgery: Asystematic review. Aust. N.Z. J. Surg.1998; 68: 388–96.

2. Hall JC, Christiansen KJ, Goodman M et al. Duration of anti-microbial prophylaxis in vascular surgery. Am. J. Surg. 1998;175: 87–90.

Aust. N.Z. J. Surg. (1999) 69, 69

SINGLE- VERSUS MULTIPLE-DOSE ANTIMICROBIAL PROPHYLAXISFOR MAJOR SURGERY: REPLY

MALCOLM MCDONALD,* ELIZABETH GRABSCH† AND ANDREW FORBES‡

*Infectious Diseases Service, Geelong Hospital, †Department of Infectious Diseases and Clinical Epidemiology, MonashMedical Centre and ‡Department of Epidemiology and Preventative Medicine, Monash University, Victoria, Australia

Single-dose antimicrobial prophylaxis is recommended for mostsurgical procedures in the latest issue of Therapeutic Guidelines:Antibiotic,1 although the authors of this book make the point that theoptimal duration has not been established for vascular surgery.As Professor Hall points out, however, ‘conventional wisdom’ and‘conventional surgical practice’ may be entirely separate matters;longer courses are used in some settings despite guidelines.

The biological principle of single-dose prophylaxis was out-lined in the original laboratory work of the Harvard surgeon JohnBurke.2 Over the last 30 years it has gained clinical plausibilitythrough numerous published studies in a variety of surgical dis-ciplines. Using strict criteria for quality, we collated and analysedthe results of available published material and concluded thatthere was no evidence of clear superiority of either single- ormultiple-dose antimicrobial prophylaxis in the prevention of surgi-cal wound infection.3 In other words, based on clinical trial data,we could find no compelling reason to depart from the generalrecommendations of Therapeutic Guidelines: Antibiotic.We donot claim superiority of single-dose over multiple-dose regimens.If, as Professor Hall claims, we ‘failed to define the limits of thisgeneralisation’ in this endeavour, it is because our limits weredefined by the currently available literature.

Therapeutic Guidelines: Antibioticalso recommends repeatedintraoperative doses for prolonged procedures; this was specificallycovered in our inclusion criteria. As for the ‘substantial evi-dence’ regarding multiple-dose antimicrobial prophylaxis for

patients with ‘heavy bleeding’ or who ‘have poor local tissueperfusion, or acquire a prosthesis’, we were unable to locate it inour literature search. However, if it is close at hand we would bedelighted to review it.

The drug company-sponsored randomized controlled, but notdouble-blinded, trial quoted by Professor Hall4 was the only suchstudy available in vascular surgery that found in favour of amultiple-dose regimen. Moreover, the conclusion was made fromanalysis of data that was on the borderline of accepted statisticalsignificance. As to whether this is clinically significant, it is notentirely clear. The suggestion that confirmatory studies may berequired before the results of one study become the basis for pre-scribing guidelines is not unreasonable.

REFERENCES1. Antibiotic Writing Group. Therapeutic Guidelines: Antibiotic

10th edn. Melbourne: Therapeutic Guidelines Ltd, 1998.2. Burke JR. The effective period of preventive antibiotic action in

experimental incisions and dermal lesions. Surgery1961; 50:161–8.

3. McDonald M, Grabsch E, Marshall C, Forbes A. Single- versusmultiple-dose antimicrobial prophylaxis for major surgery: Asystematic review. Aust. N.Z. J. Surg.1998; 68: 388–96.

4. Hall JC, Christiansen KJ, Goodman M et al. Duration of anti-microbial prophylaxis in vascular surgery. Am. J. Surg. 1998;175: 87–90.