Europe and North America, where labour costs are high and skilled technicians are scarce. Such a system would also benefit developing countries, where prevention of cervical cancer has low priority in


national health care schemes. Automation would be especially valuable in India and South America, where cervical cancer is the commonest malignant disease in women and death rates from this condition are among the highest in the world. Automated analysis of cervical smears depends on the interaction of three pieces of equipment: (a) a microscope photometer for acquisition of optical data; (b) a motor-driven stage that will sweep the slide through the field of view of a line camera in a predetermined search pattern; and (c) a computer programmed to analyse the optical data generated as the smear is scanned. Data acquisition and computer analysis were formerly slow and the scanning stages were clumsy and unreliable, but the latest generation of machines permit extremely rapid computed image analysis, and slide presentation is very reliable. Most of the systems being developed in the USA are designed to "interact" with the cytotechnologist and some are already being marketed as prescreening devices.1 The machines select problem smears for the attention of the cytotechnician; these are smears that may contain one or two suspicious cells only. The publicity associated with development of these automated screening systems has tended to overshadow other approaches to automation. This is unfortunate because there is a clear need for methods of detecting precancerous lesions in the cervix that are more accurate than Papanicolaou smears. Thus scant attention has been paid to research by Titley et all2 who investigated the use of flow cytometry for detection of neoplastic cells in cervical scrapes. Discrimination between neoplastic and nonneoplastic cells was based on forward and wide-angle light scatter, double-stranded nucleic acid content, and transferrin receptor status. Although further work is needed to sort out technical difficulties before this system is ready for clinical testing, preliminary results are encouraging. The speed, objectivity, and reproducibility of flow cytometry make this a potentially useful method for automated analysis of cervical smears. Hopes for a better test were raised by a report from Partington and co-workers,3who claim that it is possible to identify women at risk of cervical cancer by microdensitometric measurement of acid-labile nuclear DNA in exfoliated cervical cells. Confidence in the usefulness of this test as a potential screening tool is limited by small sample sizes and overlap of the relative optical densities between the normal and abnormal groups. Refinement of the test system and confirmation of the findings by an independent research group in larger numbers of patients would be welcome in view of the variable results with the modified Feulgen hydrolysis technique described by Partington et al.

The ultimate test of any automated cervical cancer screening system is a clinical trial. Several American systems, well advanced in their development, are already being evaluated in a laboratory setting, although it is not clear whether the conditions under which the systems are being tested are sufficiently stringent to satisfy the Food and Drug Administration. If they are not, a good case could be made for comparative trials in other countries in view of the enormous impact these automated systems may have on organisation of the cervical cancer screening programmes. on automated cytology systems as submitted by their developers. Analyt Quant Cytol Histol 1991; 13: 300-06. Titley I, Tree DEH, Driver M, Davies DC, Adams RL. Can flow cytometry reduce the workload for cervical screening? The results of a series of 622 specimens. Cytopathology 1991; 2: 193-203. Partington CK, Sincock AM, Steele SJ. Quantitative determination of acid-labile DNA in cervical intraepithelial neoplasia. Cancer 1991; 67:

1. Data 2.






"I would rather be kept alive in the efficient if cold altruism of a large hospital than expire in a gush of wann sympathy in a small one".1 Aneurin Bevan expressed it best, but ever since the National Health Service was founded in the UK there have been those who claim that the best way to survive childbirth, cancer, a nasty smash on the road, or just about any surgical procedure is to go to a large centre. Thus the small friendly maternity unit is "a liability";2 children with various cancers treated at paediatric oncology centres have "significantly higher" survival rates than those treated elsewhere;3 for "optimum results" the management of severely injured patients should be carried out in trauma centres;4 and across fifteen categories of surgical patients "strong and consistent evidence" is found that high volume is associated with better outcomes5 (ie, larger centres do better). Nor are these differences trivial: for vascular surgery and transurethral resection of the prostate, death rates in the large centres may be 25-40% lower than in the small hospitalswhile the 3-year survival rate for children with acute non-lymphoblastic leukaemia in one study varied from 28% at a specialist centre to 9% at non-teaching hospitals.3Such data provide good copy for the Sunday newspapers, with claims of a "life and death lottery" for those who may be consigned to a

large centre or a peripheral hospital.7

Much of the earlier debate about occasional operators and poorly equipped units centred on the safety and standards of care in cottage hospitals and general practitioner units, especially GP maternity units, but questions are now raised about a much wider range of activity. If "very few patients outside the large cancer treatment centres receive the full benefit of modem expertise and technology"7 is it right to treat any cancer outside a specialist centre? Should any patient be offered vascular surgery or a transurethral prostatectomy in a general hospital?


What about hip replacements? Or cataract surgery? Will purchasing staff in the new NHS seek, in the name of quality, to place all contracts at specialist centres? Small units can defend their record in several ways. Most obviously, they can meet the challenge directly with audit results,8-11 but audit is so time-consuming that it could never be achieved systematically for all activities in a hospital. Furthermore, audit results for major events such as mortality from a single small unit are often statistically unconvincing. Small units will therefore need to deploy more general lines of argument to counter the claims of the specialist centres.

First, the evidence

Many of the childhood conditions published such as retinoblastoma,12 Wilm’s tumour, and the childhood leukaemias. There are few published data to show that specialist centres do any better at treating the common adult cancers of breast, bowel, and bladder that comprise the routine workload of peripheral hospitals, and there is, for example, no difference between centres in survival rates for Hodgkin’s disease in children.3Furthermore, many clinicians are aware of their limitations and increasingly refer patients with the rarer tumours to specialist centres: the proportion of children with malignant disease in Britain referred to the specialist centres increased from 44% to 71 % between 1977 and 1984.3 This trend is welcome and should be encouraged. For the somewhat commoner cancers (eg, acute lymphoblastic leukaemia and multiple myeloma), non-specialist centres can get good results provided they use standard protocols for treatment and do not rely on "clinical judgment" .13-15 The implications of the evidence about surgical procedures are less straightforward. It was shown over 30 years ago16,17 that teaching hospitals had lower studies relate



to rare

mortality rates (or more precisely case-fatality ratios) than non-teaching hospitals for several common diagnoses such as appendicitis, hernia, hyperplasia of the prostate, and head injury. However, as was pointed out at the time, there are two main lines of explanation for this finding: either the teaching hospitals offer better treatment, or the non-teaching hospitals are at an initial disadvantage. The question of initial disadvantage is still not settled conclusively, 18 but many studies have shown that, even after controlling for severity of illness at admission, the survival advantage of large hospitals persists.5,6,19 It is not clear whether the patients whom the small hospital fail are the most ill’9 or the least ill;5 nor do we really know whether the key to a good result is a good hospitalz° or an experienced clinician,21 although the weight of evidence points to the former, as would be expected if the whole team of theatre, ward, and laboratory staff is important. The implication is that a clinician who achieves excellence in his or her main base may not do so in a small unit such as a private


The other feature of the evidence on surgical procedures is that there seems to be a threshold volume at which clinical excellence is achieved. For example, hospitals doing 50-100 total hip replacements a year can obtain mortality rates as low as those of hospitals doing 200 or more such operations a year.Many district general hospitals will be achieving adequate volumes for the types of operation they do, especially since the threshold of excellence seems to be lower for simple operations such as cholecystectomy than for more complex procedures such as open heart surgery. Conversely, the threshold for some apparently standard district general hospital operations such as transurethral prostatectomy is as high as for coronary

bypass grafting. Patients with rare conditions are best treated at centres with considerable experience. At the other extreme, a volume of 50 or so cases a year is, for most conditions, enough to ensure that the threshold of excellence is reached. At lower volumes and when new

techniques (eg, laparoscopic cholecystectomy) are adopted, there are doubts to dispel and it might be reasonable to conduct an audit. Finally we should remember that not everyone would agree with Bevan: for some patients warm sympathy may matter more than cold efficiency.

1. Foot M. Aneurin Bevan: a biography. Vol two: 1945-1960. London: Davis-Poynter, 1973: 132. 2. Reynolds F. Obstetric anaesthetic services. Br Med J 1986; 293: 403-04. 3. Stiller CA. Centralisation of treatment and survival rates for cancer. Arch Dis Child 1988; 63: 23-30. 4. Anderson ID, Woodford M, de Dombal FT, Irving M. Retrospective study of 1000 deaths from injury in England and Wales. Br Med J 1988; 296: 1305-08. 5. Flood AB, Scott WR, Ewy W. Does practice make perfect? Med Care

1984; 22: 98-114. HS, Bunker JP, Enthoven AC. Should operations be regionalised? N Engl J Med 1979; 301: 1364-69. 7. Hunt L. The life and death lottery. Independent on Sunday Jan 5, 1992: 3. 8. Waddell TK, Kalman PK, Goodman SJ, Girotti MJ. Is outcome worse in a small volume Canadian trauma centre? J Trauma 1991; 31: 958-61.

6. Luft

9. Schiowitz MF. Patient throughput and mortality rate in a trauma service.

Br J Surg 1990; 77: 497-98. RA, Reinken J, Shoemack. Is obstetrics safe in small hospitals? Lancet 1985; ii: 429-32. Sanger R, Clyne CA. The surgical value of community hospitals: a closer look. Ann R Coil Surg Engl 1991; 73: 77-80. Sanders BM, Draper GJ, Kingston JE. Retinoblastoma in Great Britain 1969-80: incidence, treatment and survival. Br J Ophthalmol 1988; 72:

10. Rosenblatt 11. 12.


Karjalainen S, Palve I. Do treatment protocols improve end results? A study of survival of patients with multiple myeloma in Finland. Br Med J 1989; 299: 1069-72. 14. Meadows AT, et al. Survival in childhood acute lymphocytic leukaemia: effect of protocol and place of treatment. Cancer Invest 1983; 1: 49-55. 15. Stiller CA, Draper GJ. Treatment centre size, entry to trials, and survival in acute lymphoblastic leukaemia. Arch Dis Child 1989; 64: 657-61. 16. Lee JAH, Morrison SL, Morris JN. Fatality from three common surgical conditions in teaching and non-teaching hospitals. Lancet 1957; ii:



Lipworth L, Lee JAH, Morris JN. Case fatality in teaching and non-teaching hospitals 1956-59. Med Care 1963; 1: 71-76. 18. Knaus WA, Wagner DP. Interpretation of hospital mortality rates: the current state of the art. Mayo Clin Proc 1990; 65: 1627-29. 19. Dubois RW, Rogers WH, Moxley JH, Draper D, Brook RH. Hospital inpatient mortality. N Engl J Med 1987; 317: 1674-80. 20. Kelly JV, Hellinger FJ. Physician and hospital factors associated with mortality of surgical patients. Med Care 1986; 24: 785-800. 21. Roos LL, Cageorge SM, Roos PR, Danziger R. Centralisation, certification and monitoring. Med Care 1986; 24: 1044-66. 17.

Dead or alive.

964 Europe and North America, where labour costs are high and skilled technicians are scarce. Such a system would also benefit developing countries,...
315KB Sizes 0 Downloads 0 Views