1416

detectable in a national cohort of 350 000 women of relevant agewho were taking OCs in the 1960s.The above Office of Population Censuses and Surveys data

simply demonstrate that there has been a rise in the notificationrates of breast carcinoma and the higher the age the greater the rise.The age-group allegedly most at risk had a smaller rise than did olderwomen supposedly not at risk. These figures do not support Pike’shypothesis.Whilst data which demonstrate a negative correlation between the

incidence of carcinoma and OC sales can never disprove the claimedcausal association, they can, and in this case do, cast substantialdoubts on the validity of Pike and colleagues’ generalconclusions-that a significant sized population subgroup has a 4 or5 fold increased risk of breast cancer.

Schering Chemicals Ltd,Burgess Hill, West Sussex RH15 9NE R. A. WISEMAN

PROGESTAGEN "POTENCY" AND BREAST CANCER

SIR,-It is now clear that endocrinologists reject Professor Pikeand his colleagues’ table of oral contraceptive formulations rankedaccording to "progestagen potency" (Oct 22, p 926) and the tablederived from it and distributed by the Committee on Safety ofMedicines and published in the CSM’s Current Problems no 2. Thedelay-of-menses test relates only to endometrial haemostatic

efficacy and provides no evidence for ranking syntheticprogestagens by any other property. Even by this limited yardstickonly those formulations containing 50 JJg ethinyloestradiol and theprogestagens norethisterone, norethisterone acetate, ethynodioldiacetate, norgestrel, and levonorgestrel can be compared: for therest of the formulations in the CSM table, acceptable data do notexist.

PROGESTAGEN POTENCY RATIO (DELAY-OF-MENSES TEST)

*All contain the oestrogen ethinyloestradiol 50 g.tNot currently marketed in UK.

For only twelve oral contraceptive formulations can potencyratios (such as they are) be deduced from the delay-of-menses datafor mixtures containing 50 ug ethinyloestradiol (see table).2 Prince Arthur Road,London NW3 6AU G. I. M. SWYER

TWO-STAGE VACCINATION PROGRAMME AGAINSTRUBELLA

SIR,-Rubella vaccines have been licensed for over a decade in theUK and USA, but there is still uncertainty about the relative meritsof the vaccination policies of the two countries. The strategy in theUS is to immunise large numbers of children at a preschool age, toreduce substantially the transmission rate of the virus within thecommunity and hence the incidence of congenital rubella syndrome(CRS). UK policy aims to encourage the acquisition of immunity bynatural infection during early childhood by limiting vaccination toschoolgirls aged 10-15. Concern over the failure of UK policy tosuppress the incidence of rubella significantly below that pertainingbefore widespread immunisation (the interepidemic period hasremained at about 4 years, major outbreaks being recorded in1974-75, 1978-79, and 1982-83) has generated some discussion

1. Swyer GIM. Potency of progestogens in oral contraceptives: further delay of mensesdata Contraception 1982; 26: 23-27.

about the advantages of a two-stage vaccination programme. Onesuggestion has been to vaccinate boys and girls at around 2 years ofage (perhaps as a combined measles-rubella vaccine) in addition tothe vaccination of girls between the ages of 10 and 15.Mathematical models of the epidemiology of rubella provide a

framework for examining the relative merits of current UK policyversus a two-stage programme. 1-4 A recent study suggests thatcurrent UK policy will have a substantial impact on the incidence ofCRS only after more than 20 years of immunising cohorts of10-15-year-old girls (fig 173). In accord with observation, thepredicted reduction in CRS in the short term (10-15 years) is low.3 3Our research also suggests that current policy will have little impacton the frequency and magnitude of rubella epidemics, since theaverage age at vaccination (about 12 years) exceeds the average age atinfection (9 years).2,3 We suggest that the situation in the UK todayis not necessarily indicative of an inadequate policy since the fullbenefits of the current strategy will only become apparent in 5-10years’ time.

3As a supplement to this research3 we have examined the long-termimpact of a two-stage programme (fig 1) in terms of a ratio w(al,a2)denoting the average number of CRS cases arising during one yearfrom pregnancies in the age range 16 (al) to 40 (a2) years after thevaccination programme is established, divided by the

corresponding number of cases in the same age range beforevaccination. The ratio is recorded for various levels of vaccination

coverage of 12-year-old girls (the proportion P) and 2-year-old boysand girls (the proportion P’). Our principal conclusion is that thereis little benefit to be gained from a two-stage programme (over thatarising from the current policy) unless the acceptance rate amongyoung children is very high. For example, our current acceptance1. Knox EG. Strategy for rubella vaccination. Int J Epidemiol 1980; 9: 13-23.2. Anderson RM, May RM. Directly transmitted infectious diseases: Control by

vaccination. Science 1982, 215: 1053-60.3. Anderson RM, May RM. Vaccination against rubella and measles: Quantitative

investigations of different policies. J Hyg Camb 1983; 90: 259-325.4. Hethcote HW. Measles and rubella in the United States. Am J Epidemiol (in press).

Fig I-The ratio w(al,a2) of the predicted number of CRS casesarising from pregnancies in the age range 16-40 years aftervaccination, divided by the number in the same age range beforecontrol, plotted as a function of the proportion (P’) of boys and girlsvaccinated at 2 years of age and the proportion (P) of girlsvaccinated at 12 years of age.

The average age at infection before immunisation was assumed to be 9 yearsand the methods used in the calculation ofw(al,a2) are detailed elsewhere 3 Inthe shaded regions a two-stage policy is predicted to increase the number ofCRS cases over the number of cases arising under the current policy.

1417

Age (yr)Fig 2-Age-related changes in force of infection, (a).

A(a) is the per caput rate at which susceptibles acquire rubella infectionestimated3 from UK serological data.

rate of 84% amongst 12-year-old girls and assuming an acceptancerate of 50% for 2-year-old boys and girls (the current level of measlesvaccine coverage) the predicted values of the ratio w(al,a2), in thelong-term, are 0-16 for a two-stage programme and 0 17 for currentpolicy.Our calculations are based on an average force of infection

(averaged over all age classes) calculated from age-specific values (fig2). Ifwe were to take into account differences in transmission withinand between age classes and within family groups of differing sizesthe quantitative detail of our calculations would change, but not ourconclusion that there is little benefit to be gained from changing thecurrent.programme of rubella vaccination in the UK. A substantialreduction in the incidence of CRS and even eradication of rubella

might be achieved if acceptance rates among young children wereraised to a very high level.

Department of Pure and Applied Biology,Imperial College,London SW7 2BB ROY M. ANDERSON

Biology Department,Princeton University,Princeton, New Jersey, USA ROBERT M. MAY

BILE REFLUX AND PANCREATIC CANCER

SiR.—Dr Braganza (Oct 29, p 100) suggests bile reflux as a

pathogenetic factor in pancreatic disease, including carcinoma.Referring to the experiments of Pour et al,l she claims that themodel for preventing bile reaching the pancreas by the procedure ofcholecystoduodenostomy combined with suprapancreatic bileductresection is irrelevant because the animal used, the Syrian goldenhamster, lacks a duodenal sphincter. We have extended andconfirmed the work of Pour et al in similar experiments in 143hamsters.2 Furthermore, no tumours were found in the pancreatichead in the non-bile-deviated hamsters (controls); indeed, in ourexperiments there has been a preponderance of tumours in the mostteleduodenal part of the splenic lobe of the pancreas, probably dueto the greater mass of the gland there.Braganza also claims that the hepatic lesions in acute

haemorrhagic pancreatitis may be due to aberrations in the liver. Inother experiments at our hospital acute pancreatitis was induced by

5. Miller E, Cradock-Watson JE, Pollock TM. Consequences of confirmed maternalrubella in successive stages of pregnancy. Lancet 1982 ii: 781-84.

1. Pour P, Donnelly T. Effect of cholecystoduodenostomy and choledochostomy inpancreatic carcinogenesis. Cancer Res 1978; 38: 2048-51.

2. Andrén-Sandberg A, Dawiskiba S, Ihse I. Effect of bili-digestive anastomosis onexperimental pancreatic carcinogenesis. Acta Chir Scand 1982; 148: 511-15.

retrograde injection of sodium deoxycholate and chenodeoxy-cholate into rat pancreas. A shunt was established between the

proximal common duct and the duodenum in these animals. Focalnecroses also appeared in the liver. 3,4 Thus, in experimental acutepancreatitis in rats, concomitant lesions can arise in the liver.We think that Braganza’s conclusions are wrong and that, at least

in the model for experimentally induced ductal carinoma in theSyrian golden hamster, the evidence speaks against bile reflux as anaetiological agent for pancreatic cancer.

Departments of Surgeryand Pathology,

University of Lund,S-221 85 Lund, Sweden

ÅKE ANDRÉN-SANDBERGSIGMUND DAWISKIBAIHGEMAR IHSE

METRONIDAZOLE OR VANCOMYCIN FORCLOSTRIDIUM DIFFICILE ASSOCIATED

DIARRHOEA

SIR,-Dr Teasley and colleagues (Nov 5, p 1043), in a randomisedtrial of metronidazole and vancomycin in colitis due to Clostridiumdifficile, found a failure rate of 13% in one group and 12% in theother but, properly, did not reject the null hypothesis of nodifference. However, they went further and concluded that"metronidazole is as effective as vancomycin" and they infer that itshould be the preferred treatment because it is much cheaper.A clinical trial reported as negative will often be merely

inadequate.5 If too few patients are studied, a clinically importantdifference between two treatments may not be demonstrated at the

required level of statistical significance, even though it exists.

Teasley et al stu’died groups of 42 and 52 cases. The sample sizetable in a standard statistical text indicates that, if you are

interested in a three-fold difference in failure rates (eg, 30% vs 10%)and require a p value of 0 - 05 to reject the null hypothesis, you willneed 80 patients in each treatment group to achieve an 80%probability of reaching a true conclusion, and over 100 for 90%probability. For a two-fold difference in failure rates even morepatients are needed.Blackwelder7 has proposed a different approach to the statistical

analysis of trials such as this, in which the object is to show that acheaper (or less risky or less painful) treatment is not worse thananother to a clinically important extent. Instead of seeking to reject anull hypothesis of no difference, he suggests formulating a

hypothesis that the cheaper treatment is worse to a degreepreselected on the basis of clinical judgment, and using the data toattempt to reject that hypothesis, thus justifying the use of thecheaper therapy. If Teasley and colleagues were to analyse their datafollowing Blackwelder’s suggestion they might find that theevidence is not yet convincing and be able to extend their

investigation until firm conclusions can be drawn.

National Institutes of Health,Bethesda, Maryland 20205, USA ROBERT S. GORDON

URTICARIA INDUCED BY D-PSICOSE

SIR,-A food reaction is often blamed for urticaria although firmevidence of such a role is usually lacking.8 We describe a case ofurticaria in which lesions were provoked by ingestion of a byproductof high-fructose syrup, D-psicose.A 30-year-old man had had urticarial attacks for 6 months when he

ate foods such as hamburgers, spaghetti, or cakes or took certaindrinks. He had never had attacks after bathing in warm water orafter exercise. His general condition was normal and he did notshow dermographism. Skin tests with allergen extracts of pork,

3. Hansson K, Lundh G, Stenram U. Morphological and secretory studies in experi-mentally induced pancreatitis in rats. Acta Chir Scand 1961, 121: 274-83.

4. Hansson K, Lundh G, Stenram U, Wallerström A. Pancreatitis and free bile acids. ActaChir Scand 1963; 126: 338-45.

5. Freiman JA, Chalmers TC, Smith H Jr, Kuebler RR Importance ofbeta, type II errorand sample size in the randomized control trial N Engl J Med 1978; 299: 690-94.

6. Fleiss JL. Statistical methods for rates and proportions. New York: John WIley, 1973:table A3.

7. Blackwelder WC "Proving the null hypothesis" in clinical trials. Controlled Clin Trials1982; 3: 345-53.

8. Warin RP, Champion RH. Urticaria. In: Major problems in dermatology: Vol I,London: WB Saunders, 1974, 50-56.