Medical practices

Evaluation of the ThinPrep Pap test as an adjunct to the conventional Pap smear Jennifer M Roberts, A Marion Gurley, Julia K Thurloe, Ronald Bowditch and Colin R A Laverty For editorial comment, see Wain Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au/>". Abstract - Introduction - Methods - Results - Discussion - Acknowledgements - References - Authors' details - - ©MJA1997 Abstract Objective: To evaluate the ThinPrep Pap test as an adjunct to the conventional Pap smear. Design and setting: Prospectively collected cervical samples were split for independent screening at a large specialised private gynaecological pathology practice in Sydney. Main outcome measures: Detection of additional significant abnormalities (cervical intraepithelial neoplasia 1, or more severe); changed management recommendations from "repeat smear in 12 months" or "...six months" to "colposcopy"; a reduction in unsatisfactory reports. Results: 35 560 paired (split-sample) conventional and ThinPrep slides were prepared. Significant abnormalities were detected in 724 conventional smears (2%). Additional significant abnormalities were found in 85 ThinPrep slides whose corresponding conventional smear was negative or unsatisfactory even after review, representing a 12% increase in the detection of significant abnormalities. As a result of the addition of ThinPrep, management recommendations were changed from "repeat smear in 12 months" or "...six months" to "colposcopy" for 89 of 1669 women whose conventional Pap smears showed minor non-specific changes or papillomavirus. There were 1258 conventional smears (3.5%) that were unsatisfactory compared with 235 ThinPrep slides (0.7%); for only 74 samples (0.2%) were both slides unsatisfactory. Conclusions: The addition of the ThinPrep Pap test improves detection and clinical management of cervical abnormalities, and reduces the number of unsatisfactory samples which would otherwise require repeat tests. MJA 1997; 167: 466-469 Introduction The conventional Pap smear has been the mainstay of screening for cervical precancer for approximately 50 years, without major changes in techniques relating to preparation and interpretation. In recent years there has been increasing pressure to improve Pap smear standards to reduce the occurrence of false negative results, with their associated morbidity and mortality for the women involved, and legal consequences for health professionals.1,2 Many false negative cases are the result of problems of sample transfer and smear quality.3,4 The conventional Pap smear represents only a small subsample of the cellular material collected on the sampling implement,5,6 with the remainder traditionally being discarded. Importantly, this small subsample is a non-random fraction of the material collected and is not necessarily representative of the whole. The traditional smearing technique also produces slides which vary greatly in quality, making it sometimes difficult to find and interpret abnormal cells. Improvements in both subsample selection and slide quality should therefore reduce the number of false negative cases. The ThinPrep Pap test is an automated technique developed with the aim of producing better quality slides.7 Each cervical sample is rinsed into a vial of fixative fluid. In the laboratory the sample is homogenised, and a representative subsample is collected on a single-use filter, transferred to a glass slide, and stained in the routine fashion. This technique results in a better quality preparation, as the problems of poor fixation, uneven thickness of the cellular spread, and obscuring of cells by blood and inflammatory exudate are overcome. We aimed to evaluate the impact of this technology on detection rates of significant abnormalities, accuracy in reporting, and numbers of unsatis - fact ory preparations when it was used in addition to the conventional Pap smear. Methods In a pilot study in the first half of 1996, we evaluated 1051 samples to test the performance of the ThinPrep 2000 machine (Cytyc Corporation, Boxborough, Mass., USA), as previous experience with an earlier model had highlighted some technical problems.8 The pilot study showed slides of uniformly high quality and so, in July 1996, the test was offered to all referring practitioners at a charge to patients of $20.30. For this study, we analysed paired samples collected between 1 July 1996 and 30 May 1997 by practitioners who chose to offer ThinPrep (TP) to their patients as an adjunct test. We used a split-sample protocol in which routine samples of the transformation zone (squamo-columnar junction) -- the area of the cervix where abnormalities are most likely to occur -- were taken with a broom-like device, the Cervex brush (Rovers BV, Netherlands). We insisted on the use of this implement as our earlier study8 had shown that sampling with a spatula was more likely to result in a TP slide with insufficient cells. For postmenopausal women and those who had had previous cervical surgery or ablation, an endocervical brush sample was added. After a conventional Pap smear was made, the sampling implement was vigorously rinsed in a vial of PreservCyt fixative fluid (Cytyc Corporation, Boxborough, Mass., USA). TP slides were prepared from the vial by the ThinPrep 2000 machine and both slides were stained routinely. A duplicate of each request form was made, and the conventional smear and the TP slide were examined independently by different cytotechnologists. Nineteen of our 28 cytotechnologists were involved in reading the TP slides; all 28 read conventional smears. Any slide showing minor abnormalities was reviewed by a second experienced cytotechnologist. If cervical intraepithelial neoplasia (CIN) of any grade was detected, or if a high grade lesion was suspected, the case was referred to a cytopathologist for review. Slide pairs (conventional Pap smear and TP) displaying a significant discrepancy were also reviewed. The final report issued to the referring doctor was structured according to the National Health and Medical Research Council (NHMRC) reporting terminology9 (see Box 1) and contained any relevant information obtained from both slides, together with the relevant management recommendations. For the purposes of analysis, separate final results of the two slides were also recorded. A significant abnormality was defined as one for which the management recom mendation is "colposcopy" (see Box 1). This includes any grade of CIN, adenocarcinoma in situ (AIS), invasive carcinoma or the suspicion of a high grade abnormality ("inconclusive" in NHMRC terminology). Results Approximately 500 referring practitioners chose to offer TP to their patients as an adjunct test. Some of these doctors offered the TP test to all women, and others only to selected women. Thirty-five per cent of women having non-screening cytological tests had TP, whereas only 24% of women having routine cytological tests had TP. Currently, about 30% of all smears received in this laboratory are accompanied by a TP sample. We received 35 560 paired split-sample conventional Pap smear slides and TP samples for analysis in this study. Box 1 presents a comparison of final (i.e., reviewed) results for TP and conventional Pap smear slides. Of the 34 141 paired results that were satisfactory for both slides, 32 195 (94.3%) showed total agreement and 1946 (5.7%) did not. Of the latter group, the TP slide showed a more severe abnormality than the conventional slide in 1194 (61.4%), and the opposite was true in 752 (38.6%; chi-squared = 100.4; df = 1; P < 0.001). For the 1946 paired slides which did not show total agreement, 271 had substantial discrepancies in which colposcopy would have been recommended based on the result of one test, but not the other. Of these, colposcopy would have been recommended on the basis of the TP result alone for 167 (61.6%), and on the basis of the conventional Pap smear result alone for 104 (38.4%; chi-squared = 14.6; df = 1; P < 0.001). Significant abnormalities were detected in 724 conventional smears. An additional 85 significant abnormalities were detected on TP slides for which the corresponding conventional slides, even after review, were either negative (78) or unsatisfactory (7). This represents a 12% increase in the detection of significant abnormalities. There were 14 TP slides that predicted high-grade epithelial abnormality (HGEA), while the corresponding conventional smears were either negative or unsatisfactory. Histological follow-up of 11 of these confirmed HGEA in eight, and low-grade epithelial abnormality (LGEA) in three. There were also 27 TP slides reported as inconclusive, while the corresponding conventional smears were either negative or unsatisfactory. Histological follow-up of 17 of these showed HGEA in seven, LGEA in eight, and inflammation in only two. Further, 15 conventional smears originally reported as negative, and reviewed because their TP counterparts showed an abnormality, were subsequently reported as HGEA or inconclusive. These "screening" false negatives are not represented in Box 1, which shows only final results (i.e., after review). ThinPrep slides not only increased detection of abnormalities, but also influenced management recommendations. On the basis of conventional smear tests, the recommendations for 1669 women (4.7%) with minor non-specific changes or features of human papillomavirus (HPV) effect were "repeat smear in 12 months" or "...six months", respectively. For 89 of these women (5.3%), this recommendation was changed to "colposcopy" as a result of a higher grade abnormality being detected on the TP slide. Histological follow-up concentrated on the predicted HGEA and inconclusive categories. Findings are shown in Box 2 (below). Of conventional smears, 1258 (3.5%) were found to be unsatisfactory, but only 74 (0.2%) were unsatisfactory by both methods, representing a 94% reduction in unsatisfactory reports. TP predicted HGEA in four smears reported as unsatisfactory on the conventional smear; all four have been histologically confirmed to be HGEA. Minor abnormalities (minor non- specific changes/HPV) were detected in 1669 (4.7%) of the conventional smears and 2027 (5.7%) of the TP slides. An endocervical component was absent in 8.3% of the conventional smears and in 20.0% of the TP slides. In 6.4% of cases, neither slide displayed an endocervical component. With respect to glandular abnormalities, 13 cases of adenocarcinoma in situ (AIS) or adenocarcinoma were detected by both methods. There was one histologically confirmed case of adenocarcinoma in which the conventional smear was reported as inconclusive (suspicious of AIS) and the corresponding TP slide was negative. Further TP slides were made from the remaining sample in the vial but none of these contained abnormal cells. Discussion Overall, we found that diagnostic agreement between the conventional Pap smear and the TP Pap test was high. Where there was disagreement, a higher grade abnormality was predicted significantly more often by TP than by the conventional smear. The addition of the TP test resulted in detection of additional histologically confirmed HGEA, leading to a reduction in the number of false negatives and a corresponding increase in sensitivity. The fact that the same proportion of HGEA (83%) was histologically confirmed for both conventional Pap smear and TP shows that TP does not result in undue overreporting, and that, although sensitivity has been increased, the positive predictive value of the test has been maintained. The extra significant abnormalities detected on TP and the increased accur acy of reporting are mainly attributable to improved subsampling of the cervical specimens, resulting in fewer "subsampling" false negatives. Computer-assisted rescreening of conventional smears would have no impact on this area. "Screening" false negatives (although few in this study) will also be reduced by the addition of a slide of much higher quality, and by the independent screening of additional material. That there were cases in which an abnormality detected on conventional smear was not detected on TP is not surprising, as the split-sample protocol favours the conventional smear. In these cases it appears that all abnormal cells in the sample were transferred to the conventional smear, leaving none in the TP vial. A "direct-to-vial" protocol in which 100% of the material collected is rinsed into the fluid fixative should eliminate this problem. However, if only a very small number of abnormal cells are present in the vial, then they may not be represented on any single TP slide. The addition of the TP test resulted in a large reduction in reporting of unsatisfactory preparations. This is a direct result of removal of the many variables associated with the conventional Pap smear technique (poor fixation, uneven thickness of the cellular spread, and obscuring of cells by blood and inflammatory exudate), and represents significant time and cost savings, as the need for a repeat smear is avoided. Significantly, there were four women with histologically confirmed HGEA which was predicted by TP, but whose conventional smears were reported as unsatisfactory. These abnormalities may have been detected by the recommended repeat smear three months later, but there is always a risk that women will not return for follow-up. A very small percentage of TP slides were unsatisfactory and this was always the result of there being insufficient cells in the specimen. In these cases the machine reported a dilute specimen. We believe that a number of these cases were the result of the use of incorrect sampling implements, contrary to the recommended protocol. Of some surprise to us was the increase in reporting of minor non-specific changes. While some of this no doubt represents a genuine increase in detection of minor abnormalities, we feel that other factors must be contributing. In particular, we believe that the improved nuclear morphology obtained in the TP process requires cytotechnologists and cytopathologists to "relearn" subtle criteria used at this end of the diagnostic spectrum. We noted this trend early in the study period and have been addressing it with numerous educational sessions. We continue to monitor reporting in this category. The addition of TP tests resulted in a decreased proportion of combined reports lacking an endocervical compon ent. However, TP alone had a higher rate of absent endocervical compon ent. The reason for this is unclear, but may relate to TP slides being prepared from left-over cellular material. We can only speculate that the "endocervical component present" rate may improve in a direct-to-vial situation. Even though there were more TP slides lacking an endocervical component, there was only one case in which there was a significant glandular abnormality present in the conventional smear and not in the TP slide. As further slides prepared from the TP vial also failed to show abnormal cells, we assume that no abnormal cells were present in the vial. An important advantage of the TP process is the presence of further cellular material in the fluid fixative which can be used to prepare more slides, or for HPV typing which may have clinical relevance in the triage of patients with low grade abnormalities.10 In the United States, the Food and Drug Administration has approved the TP Pap test as a replacement for the conventional Pap smear.11 This decision was based on analysis of data on 7360 paired samples derived from six different centres (Data on file, Cytyc Corporation, Boxborough, Mass., USA). Our data, on a larger sample from a single practice, support the assertion that the TP Pap test performs substantially better than the conventional smear on analysis of significant parameters. While we continue to offer the TP Pap test as an additional proced ure, we anticipate that the machine-made slide may replace the conventional smear in the future. Acknowledgements We thank Tabatha Lovelace, Lisa Wong, Rozanne Van Gramberg and Samira Bounassif for technical assistance. Statement of potential conflict of interest: The Cytyc Corporation (Boxborough, Mass., USA) lent us one of the Thinprep 2000 machines used in this study. The company was not involved in the design of the study, collection of data, analysis of results or preparation of the manuscript. References DeMay RM. To err is human -- to sue American. Diagn Cytopathol 1996; 15: iii-vi. Skoumal SM, Florell SR, Bydalek MK, Hunter WJ. Malpractice protection: communication of diagnostic uncertainty. Diagn Cytopathol 1996; 14: 385-389. Gay JD, Donaldson LD, Goellner JR. False-negative results in cervical cytologic studies. Acta Cytol 1985; 29: 1043-1046. Joseph MG, Cragg F, Wright VC, et al. Cyto-histological correlates in a colposcopic clinic: a 1 year prospective study. Diagn Cytopathol 1991; 7: 477-481. Hutchinson ML, Isenstein LM, Goodman A, et al. Homogeneous sampling accounts for the increased diagnostic accuracy using the ThinPrep Processor. Am J Clin Pathol 1994; 101: 215-219. Goodman A, Hutchinson ML. Cell surplus on sampling devices after routine cervical cytologic smears. A study of residual cell populations. J Reprod Med 1996; 41: 239-241. Zahniser DJ, Hurley AA. Automated slide preparation system for the clinical laboratory. Cytometry 1996; 26: 60-64. Laverty CRA, Thurloe JK, Redman NL, Farnsworth A. An Australian trial of ThinPrep: a new cytopreparatory technique. Cytopathology 1995; 6: 140-148. National Pathology Accreditation Advisory Council, Commonwealth Department of Health and Family Services. Requirements for gynaecological (cervical) cytology. Canberra: AGPS, 1997. Sherman ME, Schiffman MH, Lorincz AT, et al. Cervical specimens collected in liquid buffer are suitable for both cytological screening and ancillary Human Papilloma virus testing. Cancer 1997; 81; 89-97. Communication, 20 May 1996. Rockville, Md.: Center for Devices and Radiological Health, Food and Drug Administration, 1996. (Received 3 Feb, accepted 14 July 1997) Authors' details Dr Colin Laverty & Associates, Pathologists, Eastwood, NSW. Jennifer M Roberts, MB BS, FRCPA, Pathologist; A Marion Gurley, MB ChB, FIAC, Pathologist; Julia K Thurloe, BSc, MEc, Statistician; Ronald Bowditch, BScAg CT(ASC), Senior Cytotechnologist; Colin R A Laverty, MB BS, FRCPA, Principal Pathologist. Reprints: Dr C R A Laverty, Dr Colin Laverty & Associates, Pathologists, 18 Glen Street, Eastwood, NSW 2122. ©MJA 1997 <URL: http://www.mja.com.au/> © 1997 Medical Journal of Australia.

What are the risks of diagnostic medical radiation? Richard C Smart It is both ethically and economically desirable to restrict the use of diagnostic medical radiation to only those who will benefit from it. However, patients should not refuse diagnostic tests based on an exaggerated estimation of the risks because most of these tests involve low doses of radiation. It is probable that the risks derived from studies of the atomic bomb survivors, who were exposed to high doses of radiation, overestimate the risks at low doses. No evidence of thyroid cancer, leukaemia or non-Hodgkin's lymphoma has been found in patients exposed to diagnostic levels of ionising radiation. For most diagnostic tests, the risks arising from the radiation exposure are too small to be observed and the benefits will almost always outweigh the risk. (MJA 1997; 166: 589-591) Readers may print a single copy for personal use. No further reproduction or distribution of the articles should proceed without the permission of the publisher. For permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au/>". Introduction - The effect of dose - The effects of dose rate and fractionation - The effects of age at radiation exposure - Risks of medical diagnostic tests - Risks from natural background radiation - Risks to children from parental radiation exposure - Conclusion - References - Authors' details Register to be notified of new articles by e-mail - Current contents list - (c)MJA1997 Introduction Recently in this Journal, Roebuck suggested that, for many diagnostic radiology and nuclear medicine investigations, the risk from the ionising radiation may outweigh the benefits of these tests.1He outlined steps which referring doctors and radiologists may consider to reduce patient irradiation, and strongly advocated better education of the medical profession as to the risks of diagnostic radiation. However, Roebuck's estimates of the number of radiation-linked fatal malignancies in Australia each year must be questioned because these data were based on the risk of radiation-associated deaths determined from studies of the survivors of the 1945 atomic bombing of Hiroshima and Nagasaki. Many were exposed to radiation doses hundreds of times greater than those encountered in medical diagnostic procedures.2Most medical diagnostic studies result in effective doses in the range 1-20 mSv, although chest x-rays give only 0.05 mSv. There is increasing evidence that the risk associated with medical diagnostic radiation exposure is substantially less than that predicted from studies of high-dose radiation. The effect of dose The effects of radiation at high doses are dependent on the administered dose.3Solid tumours show a linear incidence with dose, while leukaemia (chronic and acute myelogenous, and acute lymphocytic) shows a linear-quadratic dose-response function (Figure, below).4In both cases, the dose-response curve flattens off at high doses (> 3 Sv), probably a direct effect of cell death. This does not necessarily imply that such a relationship continues to zero dose -- the basis of the "linear no-threshold hypothesis", which purports that any dose of radiation carries a risk. An alternative hypothesis is that there is no risk up to a certain dose but that the risk increases above this threshold; another hypothesis (commonly known as "radiation hormesis") proposes that there is a reduced risk for low levels of radiation and a higher risk at higher doses. The Hiroshima and Nagasaki survivor data showed no increase in the risk of leukaemia at low radiation doses (range, 0-200 mSv).4For solid tumours, the lowest radiation dose at which there was a statistically significant excess risk was 50 mSv,4a level well above that of most diagnostic radiological testing. The effects of dose rate and fractionation The atomic bomb survivors received all their radiation over a short period of time, and thus at high dose rates, while medical diagnostic radiation is given at low dose rates. How does dose rate affect the consequences of radiation exposure? Animal studies have shown a reduction in effect of between two and five for life-shortening and between one and 10 for tumour induction when the same total dose was given at a significantly lower dose rate.3What happens if the total dose is delivered in a number of smaller doses spread over a prolonged time period (fractionated), such as would be received by a patient having a series of diagnostic x-rays? In the past, patients with tuberculosis were monitored with fluoroscopy every two weeks over a five-year period and hence received a substantial amount of radiation (average lung dose, 840 mGy). In one United States study, there was no evidence of an increased risk of lung cancer in a cohort of such patients (standardised mortality ratio, 0.8).5However, other studies have shown an increase in breast cancer similar to that found in Hiroshima and Nagasaki survivors.6,7The risk varied with the total dose received. For those receiving a dose of 1-2 Sv, the relative risk (RR) of breast cancer was 1.38 (95% confidence interval [95% CI], 1.07-1.77), whereas for those receiving a dose of 0.5-0.9 Sv the RR was 1.11 (95% CI, 0.86-1.43).7 The United Nations Scientific Committee on the Effects of Atomic Radiation suggested a dose and dose rate effectiveness factor (DDREF) of 2.0 for leukaemia and 1.4 for other cancers.3The DDREF is a factor applied to the risk from the radiation observed in the high-dose atomic bomb studies to predict the risk from radiation observed in low-dose and low-dose-rate studies. The Committee stated that the DDREF should be applied if the total dose is less than 200 mGy or the dose rate is below 0.1 mGy/min, as is typically found in diagnostic x-ray studies.3 The effects of age at radiation exposure The most recent report on the mortality of the atomic bomb survivors (which includes an extended period of follow-up until 1990 and which includes an additional 10 500 survivors) provides estimates of excess risk specific to sex and age at exposure.4 Those exposed at age 50 had one-third of the lifetime risk per Sv for solid cancers as those exposed at age 30. Those exposed in childhood had 1-1.8 times the estimates for those aged 30.4The excess lifetime risk of leukaemia for those exposed at age 50 was about two-thirds of the risk if exposed at an earlier age.4Similarly, the Canadian studies on breast cancer following extended monitoring of treatment with fluoroscopy for tuberculosis found that the excess relative risk (ERR) decreased with increasing age at exposure.7Women exposed at age 50 or more showed no increase in breast cancer, while an ERR of 1.25 per Sv was found for those irradiated between birth and nine years. Risks of medical diagnostic tests Several recent studies have investigated the cancer risk in patients who had received diagnostic x-ray procedures. Inskip et al. identified all patients with papillary and follicular thyroid carcinoma diagnosed between 1980 and 1992 in Uppsala, Sweden.8 An equal number of control patients were matched for age, sex and country of residence. The case patients' hospital medical records were examined and the number and type of x-rays were recorded. Radiation received three to five years before diagnosis was excluded because of the known latent period for thyroid cancers. They found that each of the 484 case patients had received an average of 6.1 x-ray procedures, while the control patients received an average of 6.6 procedures; a similar number in each group received no x-rays (115 and 117, respectively). No association was found between higher doses to the thyroid and cancer, and there was no change in the relative risk with thyroid doses up to the highest dose of 80 mGy. Iodine 131 (I 131) has long been used for the diagnosis and treatment of thyroid disorders. Hall et al. recently reported on the risk of thyroid cancer in 34 104 patients given diagnostic doses of I 131 of between 0.04-37 MBq.9An excess of thyroid cancers was observed only among patients referred for a suspected thyroid cancer. No increase was seen among the 23 319 patients referred for other reasons. In a 1991 United States study, all x-ray procedures were reviewed for 565 patients with all types of leukaemia, 318 patients with non-Hodgkin's lymphoma and 208 patients with multiple myeloma.10The case patients were selected from a prepaid health plan in Oregon and California from 1956 to 1982 and were matched to 1390 control patients. The probable bone marrow dose was assigned for each x-ray and a cumulative bone marrow dose was estimated for each case patient. A latent period of four years was assumed for both leukaemia and solid tumours. A similar number of case and control patients had not had any x-rays during that time. The average number of x-rays was identical in both groups (11.6). The incidence of leukaemia in the irradiated patients was not significantly increased compared with those who had not been irradiated and there was no evidence of an increased risk with increasing dose (RR, 1.13; 95% CI, 0.7-1.8). Similarly, there were no significant increases in RRs for non-Hodgkin's lymphoma (RR, 1.24; 95% CI, 0.80-2.0) or for multiple myeloma (RR, 1.07; 95% CI, 0.6-2.0). The only group showing any possible effects of the radiation was a small group who had received the highest number of x-rays (average of 35) and who had an RR of 4.5 for multiple myeloma. Although these studies yielded negative results, they do not imply zero risk from diagnostic medical radiation but suggest that the risk is very small. Further studies involving large numbers of patients are required to detect the low levels of risk at these low doses. Risks from natural background radiation People are continuously exposed to natural background radiation (e.g., cosmic radiation, terrestrial radiation sources such as soils and building materials, and radon gas). The level of background radiation varies substantially around the world and has provided an alternative means of assessing the risks of low doses of radiation. One study involved 80 000 individuals living in two adjacent regions in China where the levels of background radiation differed by more than a factor of two.11The leukaemia mortality data indicated that there was no increasing risk with dose; if anything, there was a decreasing risk with dose. The average background radiation at sea level in Australia has been estimated to be 2.1 mSv per annum.12Therefore, a patient having a chest x-ray receives the same effective dose as he or she would receive naturally in only six days. Risks to children from parental radiation exposure The Oxford Survey of Childhood Cancers has estimated that the absolute risk of mortality from cancer following radiation exposure in utero is 1 in 20 000 per mSv.13 Data from the atomic bomb survivors indicate an increased risk of mental retardation to the fetus if the mother is exposed to radiation between eight to 25 weeks' gestation.2During the most sensitive period (eight to 15 weeks' gestation), there may be a reduction in IQ of 0.03 units per mSv.2There is no apparent increased risk of congenital malformation below a dose of 100 mSv.14The possible association of childhood cancer with paternal irradiation has also recently been investigated by the Oxford Survey of Childhood Cancers.15Using data from 14 869 children dying from cancer in the United Kingdom in the period 1953-1981 matched to an equal number of control patients, paternal irradiation before conception was found not to be a risk factor for childhood leukaemia. Gardner et al. suggested that there was a risk of childhood cancer following irradiation of the father before conception.15 They investigated a cluster of cases of childhood leukaemia and lymphoma observed in the vicinity of the Seascale nuclear processing plant in the United Kingdom and suggested an association between these cases and external irradiation of the father, particularly in the six months before conception. There have been many attempts to reproduce this finding in France, Germany, Canada and the United States, but none of these studies found a similar association.17Furthermore, clusters of leukaemia were observed at six potential nuclear sites in the United Kingdom and two nuclear installations that had been built but not operated.18As no increased levels of radiation existed at these sites, it was apparent that radiation exposure was not the cause of these clusters. Conclusion It is both ethically and economically desirable to restrict the use of diagnostic radiation to only those who will benefit from it. Wherever possible, diagnostic procedures which do not use ionising radiation should be used if these alternative techniques can give the same information. However, when a radiological study is clinically indicated it is equally important that patients do not refuse such tests based on an exaggerated estimation of the risks. For example, about 2400 women die of breast cancer in Australia each year.19Periodic mammographic screening of women over the age of 50 has been shown to reduce breast cancer mortality by 30%.20In the most recent report of the Canadian Breast Cancer Study, Howe and McLaughlin concluded that "even a very small benefit to women from routine mammographic screening would outweigh any possible risks of radiation-induced breast cancer".7If diagnostic radiation studies are used appropriately, with all routine steps taken to minimise patient radiation exposure,1the benefits will almost always outweigh the risk. References Roebuck DJ. Ionising radiation in diagnosis: do the risks outweigh the benefits? Med J Aust 1996; 164: 743-747. International Commission on Radiological Protection. 1990 Recommendations of the International Commission on Radiological Protection. (ICRP Publication 60.) Oxford: Pergamon Press, 1991. United Nations Scientific Committee on the Effects of Atomic Radiation (UNSCEAR). Sources and effects of ionizing radiation, Annex A. Epidemiological studies of radiation carcinogenesis. 1994 Report. New York: UNSCEAR, 1994. Pierce DA, Shimizu Y, Preston DL, et al. Studies of the mortality of atomic bomb survivors. Report 12, Part 1. Cancer: 1950-1990. Radiat Res 1996; 146: 1-27. Davis FG, Boice JD Jr, Hrubec Z, et al. Cancer mortality in a radiation-exposed cohort of Massachusetts tuberculosis patients. Cancer Res 1989; 49: 6130-6136. Boice JD Jr, Preston DL, Davis FG, et al. Frequent chest x-ray fluoroscopy and breast cancer incidence among tuberculosis patients in Massachusetts. Radiat Res 1990; 125: 214-222. Howe GR, McLaughlin J. Breast cancer mortality between 1950 and 1987 after exposure to fractionated moderate-dose-rate ionizing radiation in the Canadian fluoroscopy cohort study and a comparison with breast cancer mortality in the atomic bomb survivors study. Radiat Res 1996; 145: 694-707. Inskip PD, Ekbom A, Galanti MR, et al. Medical diagnostic x-rays and thyroid cancer. J Natl Cancer Inst 1995; 87: 1613-1621. Hall P, Mattsson A, Boice JD Jr. Thyroid cancer after diagnostic administration of iodine-131. Radiat Res 1996; 145: 86-92. Boice JD Jr, Morin MM, Glass AG, et al. Diagnostic x-ray procedures and risk of leukaemia, lymphoma and multiple myeloma. JAMA 1991; 265: 1290-1294. Wei L, Zha Y, Tao Z, et al. Epidemiological investigation of radiological effects in high background radiation areas of Yangjiang, China . J Radiat Res 1990; 31: 19-136. Costello JM. Radioactivity in the environment. Radiat Prot Aust 1983; 1: 21-27. Mole RH. Childhood cancer after prenatal exposure to diagnostic x-ray examinations in Britain. Br J Cancer 1990; 62: 152-168. Mettler FA, Moseley RD. Medical effects of ionizing radiation. Grune & Stratton, 1985. Sorahan T, Lancashire RJ, Temperton DH, Heighway WP. Childhood cancer and paternal exposure to ionizing radiation: A second report from the Oxford Survey of Childhood Cancers. Am J Ind Med 1995; 28: 71-78. Gardner MJ, Snee MT, Hall AJ, et al. Results of a case-control study of leukaemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. BMJ 1990; 300: 423-429. McLaughlin JR, Clarke EA, Nishri D, et al. Childhood leukaemia in the vicinity of Canadian nuclear facilities. Cancer Causes Control 1993; 4: 51-58. Cook-Mozaffari P, Darby SC, Doll R. Cancer near potential sites of nuclear installations. Lancet 1989; 2: 1145-1147. Australian Institute of Health and Welfare. Cancer in Australia, 1989-90. Canberra: AIHW, 1996. Fletcher SW, Black W, Harris R, et al. Report of the International Workshop on Screening for Breast Cancer. J Natl Cancer Inst 1993; 85: 1644-1656. Authors' details Department of Nuclear Medicine, St George Hospital, Kogarah, NSW. Richard C Smart, MSc, PhD, Principal Medical Physicist. Reprints: Dr R C Smart, Department of Nuclear Medicine, St George Hospital, Kogarah, NSW 2217. E-mail: r.smart @ unsw.edu.au ©MJA 1997 <URL: http://www.mja.com.au/> (c) 1997 Medical Journal of Australia.

Richard C Smart

Cancer and TV towers: association but not causation A more complete knowledge of the causes of childhood leukaemia is essential before progressing from association to causation MJA 1996; 165: 599 Readers may print a single copy for personal use. No further reproduction or distribution of the articles in whole or in part should proceed without the permission of the publisher. For copyright permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au/>". - ©MJA1996 Nothing concentrates public concern more than links -- real or supposed -- between pollution from man-made objects and malignant diseases. This is probably made worse if we do not particularly enjoy the aesthetic appeal of the objects in question, and worse still if we do not understand what the biological effects might be of the pollutants involved. This situation is exemplified by the study by Hocking et al. in this issue of the Journal. They report an excess of leukaemia in people living near the TV towers just north of Sydney, the excesses being mainly in children and in certain time periods. Technically epidemiologists will regard such an observation as an "association" -- that is to say, pretty difficult (if not impossible) to understand in isolation and one which has to be put into context. Is this association in Sydney the product of one of these quirks of distribution? How does one interpret the study of Hocking et al.? Such a question is effectively asking: how can an epidemiological study lead to a causal link being established between radiofrequency irradiation and leukaemia in children? Several informative steps are required. Firstly, is this a new observation? The answer is, nearly but not quite. The authors themselves refer to a rather unsatisfactory study from Honolulu, which might show a non-statistically significant leukaemia excess near radio transmitters. 1 In addition, and potentially more interesting, an extensive study in Britain has shown a nearly twofold excess in leukaemia in residents close to one particular TV transmitter complex -- but only in mixed types of leukaemia in adults and not in children. 2 However, when all the other TV and related radio transmitters in Britain were studied this excess was not confirmed, although a decline in leukaemia incidence with distance from the towers was observed. It should be noted that the adult leukaemias are very different diseases, in terms of causation, from childhood acute leukaemia, 3 as indeed the different types of adult leukaemias are from each other. 4 As these other studies do not unequivocally support the findings of Hocking et al., we are left with two further lines of enquiry. One is to ask whether there are other explanations for the leukaemia excesses in northern Sydney, and the second is to question what is known of the physical properties of this type of non-ionising irradiation and whether there are any abnormal biological effects consequent on exposure of human tissue. There is not a great deal known of the possible harmful biological effects of radiofrequency non-ionising irradiations and nothing which might suggest that they cause malignant diseases. 5 A great deal more is known of the lower energy electromagnetic fields (50-60 Hz), but here, broadly speaking, no studies have yet suggested a likely biological link with a leukaemogenic process. This unsatisfactory state of affairs fuels public and clinical concerns and has led to a small epidemic of studies on childhood leukaemias aimed at investigating (using case-control methods) all known and hypothesised causes of the condition. These studies, in New Zealand, Canada, the United States, the United Kingdom and Germany, are all coming to fruition over the next few years. Not only do they represent the concern of the public, but also the general lack of knowledge of common risk factors for childhood leukaemias. Thus, other possible explanations for the observation of Hocking et al. are limited at the moment. Nevertheless, a good deal of descriptive epidemiology of childhood leukaemias is available, mainly from the United Kingdom. 6 This throws up some remarkable features of the diseases. Leukaemia is not evenly distributed among the childhood population -- it is roughly twice as common in areas that are either geographically isolated from major conurbations or of greater affluence. 7 Furthermore, childhood leukaemias do occasionally form close case aggregations or clusters. 8 Rare diseases will form striking clusters by chance, but in the case of childhood acute lymphoblastic leukaemia more occur than simply by chance. Where and why is not known, but is under investigation. 9 Is this association in Sydney the product of one of these quirks of distribution? Is there a local case excess in an affluent area, or even an unrecognised cluster of leukaemia? The way forward would certainly include a closer look at the Sydney data for evidence of these phenomena. Area cluster statistics should be used, 10 as well as a more rigorous application of point-source statistics, which, for example, look for decline in rates by distance for a putative point hazard. 11 These possible local studies, together with epidemiological studies from elsewhere, might show similar, and thus supporting, results. There is also the possibility of new biological studies supporting a causal link. However, a more complete knowledge of the causes of childhood leukaemia is essential in order to go down the road from association to causation. In that regard we have taken only the first few steps of a very long journey. Ray A Cartwright Professor of Cancer Epidemiology Leukaemia Research Fund, University of Leeds, Leeds, UK Maskarinec G, Cooper J, Swygert L. Investigation of increased incidence in childhood leukaemia near radio towers in Hawaii: preliminary observations. J Environ Pathol Toxicol Oncol 1994; 13: 33-37. Dolk H, Elliott P, Shaddick G, et al. Leukaemia incidence near high power radiotransmitters [abstract]. Epidemiology 1996; 7(Suppl 4): S95. Doll R. The epidemiology of childhood leukaemia. J R Statist Soc Ser A 1989; 152: 341-351. McKinney PA, Alexander FE, Roberts BE, et al. Yorkshire case-control study of leukaemias and lymphomas parallel multivariate analyses of seven disease categories. Leuk Lymphoma 1990; 2: 67-80. National Radiological Protection Board. Electromagnetic fields and the risk of cancer. Report of an Advisory Group on Non-Ionising Radiation. Vol 3, No. l. Chilton, Didcot, Oxfordshire: National Radiological Protection Board, 1992. Draper G, editor. The geographical epidemiology of childhood leukaemia and non-Hodgkin's lymphoma in Great Britain 1966-1983. London: HMSO, 1990. Alexander FE, Ricketts TJ, McKinney PA, Cartwright RA. Community lifestyle characteristics and risk of acute lymphoblastic leukaemia in children. Lancet 1990; 336: 1461-1465. Cartwright RA, Alexander FF, McKinney PA, Ricketts TJ. Leukaemia and lymphoma: an atlas of distribution within areas of England and Wales 1984-1988. London: Leukaemia Research Fund, 1990. Alexander FE, Wray N, Boyle P, et al. Clustering of childhood leukemia: a European study in progress. J Epidemiol Biostat 1996; 1: 13-24. Chen R, Mantel N, Kingberg M. A study of three techniquest of time-space clusters in Hodgkin's disease. Stat Med 1984; 3: 173-184. Bithell JF, Stone RA. On statistical methods for analysing the geographical distribution of cancer cases near nuclear installations. J Epidemiol Commun Health 1989; 43: 79-85. - To top of article - ©MJA 1996 <URL: http://www.mja.com.au/> © 1996 Medical Journal of Australia.

Ray A Cartwright

Cancer incidence and mortality and proximity to TV towers Bruce Hocking, Ian R Gordon, Heather L Grain and Gifford E Hatfield MJA 1996; 165: 601 For editorial comment, see Cartwright Readers may print a single copy for personal use. No further reproduction or distribution of the articles in whole or in part should proceed without the permission of the publisher. For copyright permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au/>". Abstract - Introduction - Methods - Results - Discussion - Conclusion - References - Authors' details - ©MJA1996 Abstract Objective: To determine whether there is an increased cancer incidence and mortality in populations exposed to radiofrequency radiations from TV towers. Design: An ecological study comparing cancer incidence and mortality, 1972-1990, in nine municipalities, three of which surround the TV towers and six of which are further away from the towers. (TV radiofrequency radiation decreases with the square of the distance from the source.) Cancer incidence and mortality data were obtained from the then Commonwealth Department of Human Services and Health. Data on frequency, power, and period of broadcasting for the three TV towers were obtained from the Commonwealth Department of Communications and the Arts. The calculated power density of the radiofrequency radiation in the exposed area ranged from 8.0 µW/cm 2 near the towers to 0.2 µW/cm 2 at a radius of 4 km and 0.02 µW/cm 2 at 12 km. Setting: Northern Sydney, where three TV towers have been broadcasting since 1956. Outcome measures: Rate ratios for leukaemia and brain tumour incidence and mortality, comparing the inner with the outer areas. Results: For all ages, the rate ratio for total leukaemia incidence was 1.24 (95% confidence interval [CI], 1.09-1.40). Among children, the rate ratio for leukaemia incidence was 1.58 (95% CI, 1.07-2.34) and for mortality it was 2.32 (95% CI, 1.35-4.01). The rate ratio for childhood lymphatic leukaemia (the most common type) was 1.55 (95% CI, 1.00-2.41) for incidence and 2.74 (95% CI, 1.42-5.27) for mortality. Brain cancer incidence and mortality were not increased. Conclusion: We found an association between increased childhood leukaemia incidence and mortality and proximity to TV towers. MJA 1996; 165: 601-605 Introduction The biological effects of low level electromagnetic fields and any relation to cancer causation are controversial. There have been several epidemiological studies of possible effects of extremely low frequency (50 Hz) fields, 1 but few have looked at radiofrequency radiations (RFR) (i.e., frequencies of 300 kHz to 300 GHz). Goldsmith, 2 in a recent review, concluded that there may be an association between RFRs and cancer; however, a World Health Organization review concluded that there is no clear evidence of detrimental health effects in humans exposed to RFR. 3 An opportunity for studying the effect of RFR presents itself in northern Sydney, New South Wales, where three TV towers are sited in a triangle close to each other (Figure 1). The towers have been used to broadcast three TV services since 1956 and four since 1965. The channel frequencies range from 63 to 215 MHz; the wavelengths (ranging from 5 m to 1 m) are close to body reson ances and hence are maximally absorbed. 3 We compared cancer incidence and cancer mortality for the three municipalities (Lane Cove, Willoughby and North Sydney -- population, 135 000) which immediately surround the TV towers (inner area) with data for six adjacent municipalities (Ryde, Ku-ring-gai, Warringah, Manly, Mosman and Hunters Hill -- population, 450 000) (outer area) (Figure 1), on the basis that the RFR becomes progressively weaker with the square of the distance from the towers across these municipalities. The control municipalities were selected because of the similar distance from the towers to their nearest borders, their resi dents having a similar upper-middleclass socioeconomic status, 4 and their areas being large enough for there to be a decrease in power density. Cancers of interest were leukaemia and brain tumour, especially in childhood, given findings from community studies of extremely low frequency (50 Hz) electro magnetic fields. 1 We had no prior knowledge of, nor had concerns been raised about, clusters of leukaemia cases in the areas close to the towers. Methods Radiofrequency radiation Data for frequency and power of the RFR sources on the towers for the period 1956-1990 were obtained from the Commonwealth Department of Communications and the Arts 5 and are shown in Box 1.1. The TV signals are composed of 100 kW video amplitude modulated (AM) and 10 kW audio frequency modulated (FM) signals, on carrier frequencies which range from 63 to 215 MHz. The combined field strengths at increasing distances were calculated by the method of the United States National Council on Radiation Protection and Measurement 6 (Box 1). There were no TV repeater stations in the inner or outer areas during the survey period. Cancer data The NSW Cancer Registry maintains a comprehensive database allowing distinction between incidence and mortality, and giving residence at the time of report. 4 Data from the registry for 1972 to 1990 are available from HealthWiz 7 and were extracted by municipality, and for sex and age bands 0-14 years, 15-69 years and 70 years and over. The data are available only for the three-digit code categories identified by the International classification of diseases, injuries and causes of death , ninth revision (ICD-9). More refined data are not available for reasons of privacy. Cancer data from before 1972 are not available. Statistical analysis The data were analysed using a Poisson regression model, 8 in which the number of cases or deaths were regarded as Poisson random variables, whose mean is a product of the person-years (i.e., the sum of appropriate mid-year populations) pertaining to the observation and the functions of the explanatory variables. These models give rate ratio estimates for comparisons of interest, adjusted for the other variables. Interactions were examined, and the model tested for goodness-of-fit. We made adjustment for extra-Poisson variation, when necessary, using the "quasi-likelihood" method of McCullagh and Nelder. 9 The explanatory variables fitted in these models were: age in years (0-14, 15-69, 70 and over), sex, calendar period (1972-1978, 1979-1984 and 1985-1990), and area ("inner" [close to the TV towers] and "outer" [more distant]) (Figure 1). For comparisons between the areas of interest and the whole of New South Wales, standardised incidence ratios (SIRs) and standardised mortality ratios (SMRs) were calculated. For these analyses the stratification was by calendar-year (19 separate years), age and sex. Confidence intervals were calculated by the "exact" method. 10 Results Box 2 shows the data structure used in the analysis; the leukaemia cases and person-years in each cell were obtained by summing across the years for that age-group and sex combination. The rate ratios comparing the inner with the outer areas are shown in Boxes 3 and 4. No increase in brain cancer incidence or mortality was found, but there was an increased leukaemia incidence and mortality in the municipalities close to the towers. The rate ratio for childhood leukaemia incidence (Box 4) was 1.58 (95% CI, 1.07-2.34) and for mortality was 2.32 (95% CI, 1.35-4.01). These rates were broadly consistent across the types of leukaemia; for lymphatic leukaemia, the rate ratio was 1.55 for incidence and 2.74 for mortality. Our analysis pooled data from the inner and outer municipalities. To see whether results within each municipality were similar, we performed tests of homogeneity for childhood leukaemia incidence and mortality. No significant heterogeneity was found ( P = 0.10 for incidence and P = 0.13 for mortality). We found no significant overall trends across time for brain cancer or leukaemia incidence, for all ages combined or for children alone. For children, there was a significant overall reduction in leukaemia mortality over time ( P = 0.008), but no significant evidence of a change over time in the differences between the outer and inner areas in brain cancer or leukaemia incidence or mortality, for all ages combined or for children alone. Because a small part of Hunters Hill projects close to the TV towers (Figure 1) and there is a potential confounder there (a factory which used radium until the 1970s in Hunters Hill), the data were analysed excluding Hunters Hill. The incidence rate ratio for childhood leukaemia was 1.56 (95% CI, 1.09-2.22), and for all ages was 1.23 (95% CI, 1.06-1.43). Childhood cancer incidence and mortality (brain cancer and leukaemia) for the inner and outer areas were compared with cancer incidence and mortality data for the whole of New South Wales (Box 5). There was no difference for cancer of the brain. Leukaemia incidence and mortality were significantly increased in the inner area, but incidence and mortality data for the outer area were similar to data for the State as a whole. Discussion This ecological study found an association between residential proximity to TV towers and increased incidence of childhood leukaemia. Study biases Studies of this type are prone to biases. 1. Comparison of the inner and outer areas: Socioeconomic class has been associated with leukaemia, with a positive association with higher socio economic status. However, all muni cipalities considered in the inner and outer areas are ranked in the top two socioeconomic quintiles; further, two out of three of the inner municipalities are in the top quintile, and four out of six of the outer municipalities are in the top quintile. 4 Moreover, for New South Wales as a whole there is no evidence of a socioeconomic gradient for leukaemia. 4 There are small pockets of light industry in the surveyed municipalities, but they are mainly residential. The area closer to the TV towers is subject to much higher traffic density than the outer area, and exhaust fumes contain small traces of benzene, a proven leukaemogen. 11 However, a causal relationship between exhaust fumes and childhood leukaemia has not been established; 11 in occupational studies benzene exposure is related predominantly to acute myeloid leukaemia, 12 but we found an increased incidence/mortality of lymph atic leukaemia in the inner areas. 2. Confounding variables affecting individuals can not be adjusted for. The few recognised causes of leukaemia include ionising radiation, cytotoxic drugs and some uncommon genetic conditions. 13 The only known potential community exposure to ionising radiation in the study area is a factory in Hunters Hill that used radium until the 1970s. There are no high voltage power lines traversing the inner area, but one traverses the outer area and runs along the border between Lane Cove and Ryde in a national park. Individual (household) exposure cannot be determined, and therefore local enhancements and attenuations of RFR , which might influence dose-response calculations, cannot be allowed for. Usually, exposures in flats and houses will be lower than those for free space, such as gardens, parks and schoolyards. 3. Population movement cannot be adjusted for. Thus, miscalculations arise if people move out of, or into, particular areas for selective reasons (e.g., treatment of cancer is offered at Royal North Shore Hospital, which is in the inner area). This would not influence incidence, but could influence mortality data if patients with cancer came to live closer to the hospital for ease of access. However, it appears most childhood leukaemia cases attend children's hospitals not in the study area. A linkage study of cases could resolve this. On the other hand, social mobility would tend to obscure effects that have long latency periods. Duration of residence would need to be determined in a more detailed study. Migration to new towns has been suggested as a confounding factor in childhood leukaemia clusters, 14 with viral spread to susceptible persons, but the areas surveyed in this study are long established. Greaves, 15 using a similar argument, postulated that fewer infectious stimuli in early postnatal life, with later infection at a critical period, may play a major role in precipitating acute lymphoblastic leukaemia. According to this theory, less dense populations mean less exposure to infections early in life and higher rates of leukaemia. However, of the areas surveyed the inner area is the more densely populated (2818 per km 2 , compared with 1378 per km 2 ). 16 Effects of radiofrequency radiation The calculated exposure levels of 8.0 to 0.2 µW/cm 2 in the inner area are very low compared with the Australian Standard 17 public exposure level of 0.2 mW/cm 2 . The mechanism whereby such low energies could cause biological effects is a matter of intense research. A recent report by the Commonwealth Scientific and Industrial Research Organisation (CSIRO) concluded that reliance on thresholds for heat build-up in setting the Australian safety standard may be insufficient. 18 The TV frequencies considered, because of their wavelengths in relation to body heights, are close to body resonance, 3 leading to maximum absorption by both adults (including pregnant women 19 ) and children. However, in considering any biological effects, regard must be given to the modulations (50 Hz to 5 MHz) as much as to the carrier wave. The key video modulation frequencies are pulsed at 50 Hz and 15.6 kHz. Many and conflicting reports have been published about possible biological effects at low energy levels with low frequency amplitude modulations. 3,18,20,21 Stuchly et al. found that 60 Hz low-level fields may act as promoters of cancer. 22 It has been suggested that the biological effects may be on the cell membrane rather than the genetic material, 23 and that low energy signals are detected through non-linear mechanisms, such as stochastic resonance. 24 The disparity between our calculations and the measured power densities could result from various mechanisms, including absorption of the signal and cancellation due to reflections, especially as the minimum of one signal is unlikely to coincide with the minimum of another, even being different for the audio and vision signals of one channel. An extensive measurement program is needed to develop detailed contour maps to better define dose-response relationships. Techniques such as isotonic regression could be used; this enables effects of a point source on a surrounding community to be analysed. 25 The new services since 1980 will have increased the power density due to Tower 1 by four times, most of this being due to ultra high frequency (UHF) TV (526-533 MHz) and FM radio. A number of other services, such as mobile phone and paging services, may have also been established in the surveyed areas. However, these are of much lower power and/or use different carrier frequencies and/or modulations to the TV broadcast services. Radiofrequency radiation and cancer? An association between RFR and childhood leukaemia has not been reported previously. None of the previous studies of RFR has looked at exposure of such a large population (including children) for so long a time to frequencies of maximal body absorption. A study of people working on TV towers did not find evidence of chromosome damage, 26 and among 32 cases of neoplasms of the blood in Telecom Australia employees (retiring for medical reasons or dying) there was no excess in radiocommunication occupations (Hocking, unpublished data). A small study from Honolulu (Hawaii), where broadcast towers are also situated in populated areas, compared census tracts with towers with those without towers and found a non-significant standarised incidence ratio of 1.5 for all types of leukaemia. 27 A preliminary report of a small area study of leukaemia near 20 TV/FM transmission sites in the United Kingdom found a decline in incidence of adult leukaemia with distance, but concluded that "the results give, at most, no more than weak support for an association between residence near transmitters and leukaemia risk". 28 However, that study was restricted to adult leukaemia and incidence, whereas our most significant results were for childhood leukaemia incidence and mortality. The time trend for childhood leukaemia incidence has remained fairly stable, consistent with a constant exposure, and the reduction noted in the childhood leukaemia mortality rate most likely reflects improvements in treatment. If RFR exposure is a relevant (causal) factor, classification of population into inner and outer areas is a proxy for the appropriate exposure variable, which would tend to bias the rate ratios towards the null (i.e., towards no effect) due to non-differential misclassification. This is relevant to the irregular natural boundaries of the municipalities (Figure 1). Analysis by postcode or census collector units would yield more refined data in relation to distance from the towers. Finally, our observation of a more marked association between proximity to TV towers and leukaemia mortality than incidence (Box 4) could be of biological interest if a putative exposure not merely caused the disease but influenced its progression. Conclusion The calculated levels of RFR in the areas with increased childhood leukaemia incidence and mortality are substantially below the current Australian public safety standard. More detailed studies (e.g., relating cases to power density contours) are required to replicate any association and to look for dose-response relationships before any conclusions can be drawn. References Savitz D, Ahlbom A. Epidemiologic evidence of cancer in relation to residential and occupational exposures. In: Carpenter D, Ayrapetyan S, editors. Biological effects of electric and magnetic fields. Chapter II. Vol 2. Sydney: Academic Press, 1994: 233-261. Goldsmith JR. Epidemiologic evidence of radio-frequency (microwave) effects on health in military broadcasting and occupational studies. Int J Occup Med Environ Health 1995; 1: 47-57. World Health Organization. Environmental Health Criteria 137 Electromagnetic fields (300 Hz to 300 GHz). WHO: Geneva, 1993: 164-168, 74-75. Smith D, Taylor R, Coates M. Socioeconomic differentials in cancer incidence and mortality in urban New South Wales 1987-1991. Aust N Z J Public Health 1996; 20: 129-137. Department of Communications and the Arts. Radio and Television Broadcasting Stations. Canberra: AGPS, 1994. National Council on Radiation Protection and Measurement. A practical guide to the determination of human exposure to radiofrequency fields. Bethesda, Md: NCRP, 1993. (NCRP Report No. 119.) HealthWiz. National health database. Commonwealth Department of Human Services and Health. 1991-1996. Canberra: Prometheus Pty Ltd, 1996. Frome EL. The analysis of rates using Poisson regression models. Biometrics 1983; 39: 665-674. McCullagh P, Nelder JA. Generalized linear models. London: Chapman and Hall, 1983: 80-81. Liddell J. Simple exact analysis of the standardised mortality ratio. J Epidemiol Community Health 1984; 38: 85-88. UK Department of the Environment. Expert Panel on Air Quality Standards. Benzene. London: HMSO, 1994. Akasoy M. Benzene carcinogenicity. Boca Raton, Fl: CRC Press, 1988: 119-125. Doll R, Darby S. Childhood leukaemia in the United Kingdom. Radiat Protect Aust 1990; 8(3): 55-61. Childhood leukaemia: an infectious disease? [editorial]. Lancet 1990; 336: 1477-1479. Greaves MF. Etiology of childhood acute lymphoblastic leukemia: a soluble problem? In: Gale RP, Hoelzer D, editors. Acute lymphoblastic leukemia. UCLA Symposium on Molecular and Cellular Biology. New Series Vol 108. New York: Academic Press, 1989. Digital Cadastral Database. Bathurst, NSW: Land Information Centre. Australian Standard AS2772.1. Radiofrequency Radiation Part 1: Maximum Exposure Levels -- 100 kHz to 300 GHz. Sydney: Standards Australia, 1990. Barnett S. Status of research on biological effects and safety of electromagnetic radiation: telecommunications frequencies. Chatswood: Division of Radiophysics CSIRO, 1994. Fleming AHJ, Joyner KH. Estimates of the absorption of radiofrequency radiation by the embryo and foetus during pregnancy. Health Phys 1992; 63: 149-159. Adey R. Effects of weak amplitude modulated microwave fields on calcium efflux from awake cat cerebral cortex. Bioelectromagnetics 1982; 3: 295-307. Adey R. Frequency and power windowing. Proc IEEE (Proceedings of the Institution of Electrical and Electronic Engineers) 1980; 68 (1): 119-125. Stuchly MA, McLean J, Burnett R, et al. Modification of promotion in the mouse skin by exposure to an alternating magnetic field. Cancer Lett 1992; 65: 1-7. Weaver J, Astumian RD. The thermal noise limit for threshold effects of electric and magnetic fields in biological systems. In: Carpenter D, Ayrapetyan S, editors. Biological effects of electric and magnetic fields. Chapter 3, Vol 1. Sydney: Academic Press, 1994: 83-104. Moss F, Wiesenfield K. The benefits of background noise. Sci Am 1995; August: 50-54. Stone RA. Investigations of excess environmental risk around a putative source. Statistical problems and a proposed test. Stat Med 1988; 7: 649-660. Garson MO, McRobert TL, Campbell LJ, et al. A chromosomal study of workers with long-term exposure to RFR. Med J Aust 1991; 155: 289-292. Maskarinec G, Cooper J, Swygert L, et al. Investigation of increased incidence in childhood leukaemia near radio towers in Hawaii: preliminary observations. J Environ Pathol Toxicol Oncol 1994; 13: 33-37. Dolk H, Elliott P, Shaddick P, et al. Leukaemia incidence near high power radio transmitters [abstract]. Epidemiology 1996; 7(Suppl 4): S95.(Received 3 Oct 1995, accepted 25 Sep 1996) Authors' details Bruce Hocking and Associates, 9 Tyrone Street, Melbourne, VIC. Bruce Hocking , FAFOM, FAFPHM, Consultant in Occupational Medicine. Statistical Consulting Centre, University of Melbourne, Melbourne, VIC. Ian R Gordon , PhD, Director. System Innovations in Health, PO Box 125, Melbourne, VIC. Heather L Grain , ADipMRA, GDipDP, Consultant. Broadcast & Radiation Safety Consulting, Melbourne, VIC. Gifford E Hatfield , MSEE, Managing Director. No reprints will be available. Correspondence: Dr B Hocking, 9 Tyrone Street, Camberwell, VIC 3124. To top of article - ©MJA 1996 <URL: http://www.mja.com.au/> © 1996 Medical Journal of Australia.

Bruce Hocking · Ian R Gordon · Heather L Grain · Gifford E Hatfield

Ensuring quality in all phases of the pathology cycle Laboratories need to monitor and improve their procedures to ensure that they always deliver the correct test results for the correct patient to the correct doctor MJA1996; 165: 125-126. There is an unfortunate perception that pathology laboratories are only interested in performing tests, whereas, in reality, they are responsible for a cycle of activities which begin before and conclude after the actual testing. These activities include consultation between the clinician and pathologist, accurate identification of the patient and referring doctor, correct collection and storage of the specimen, timely and accurate delivery and readability of the result, and, if necessary, more consultation between the pathologist and the clinician. As a consequence of more than 30 years of the Royal Col lege of Pathologists of Australasia (RCPA) leading the provision of well-organised proficiency testing programs, and the advent of medical (pathology) laboratory accreditation with mandatory National Association of Testing Authorities-RCPA registration in 1986, 1 the quality of analysis in Australian laboratories is as good as, if not better than, that in most developed countries. However, it has been recognised that there are still significant problems with the activities that surround analysis in many laboratories. In this issue of the Journal, Khoury et al. ( page 128 ) present an insight into the extent of the problem. Although there were an equal number of laboratories which compared less than favourably with their peers in each of the analytical and transcription-error aspects of the study, the error rate would appear to be much higher in the latter. While some of the 14 laboratories they looked at had near-perfect records in transcribing patient details, median error rates for various elements of the data were in the order of 3%. One laboratory documented an overall transcription-error rate of nearly 40%. If these findings are extrapolated to all laboratory activity in Australia, errors must abound in alarming numbers. In a similar study of 417 laboratories (eight from Australia), Nakhleh and Zarbo 2 found deficiencies of identification and recording in 6% of surgical pathology requests. While neither of these studies addressed the question of whether such errors affect patient outcomes, there is indirect evidence that a minor, preventable systems error can have serious consequences. 3 The Quality in Australian Health Care Study 4 showed that a significant number of patient adverse events were related to system errors. These were often errors of omission related to investigations, such as failure to act upon results of findings, or failure to carry out indicated tests. In such cases it is likely that one health care worker's error of omission results from another's error of commission, such as the incorrect delivery of a report. It is likely that most errors occur because of system problems. Typical examples are poor staff training and non-existent work manuals. Inadequate computer systems with insufficient internal audit checks and poor or no access to the hospital Master Patient Index and ward information system are the bane of many laboratories' existence. Despite a belief that those in private practice are more skilled at the activities surrounding analysis, the article by Khoury et al. indicates that the problems are widespread in both public and private sectors. Where then does the remedy lie? The first step in rectifying this situation is to recognise and document that the problems exist. Pathology laboratories should recognise that their professional responsibility encompasses a whole cycle of events, starting with the requesting doctor initiating an investigation necessary for the management of a patient, right through to recognising the significance of the results for the management of the patient. Then, laboratories should be given, and accept, the authority and responsibility necessary to fulfil their professional obligations to both patient and clinician. Australian pathology laboratories have been slow to adopt total quality improvement processes. Unfortunately, there are even those who believe that too much money is already being spent on mandatory quality activities. This near-sighted stance ignores the significant costs to both patients and the community of preventable adverse events, not to mention potentially expensive medicolegal exposure. Without some mechanism for documenting the problems they experience or cause, laboratories are in no position to begin improving the quality of their services. An example of such documentation stipulating the requirements for laboratory quality systems has been provided by the International Standards Organisation. 5 Because so few laboratories monitor their pre- and post-analytical activities, there is little information from which to develop the necessary benchmarks. The College of American Pathologists has introduced "quality probes" programs to gather data from which realistic benchmarks may be re commended. These programs also document the features of laboratory operations and organisation that correlate with better quality outcomes. A good example of this is a study of wristband identification errors which showed that hospitals with the lowest error rate had systems where the pathology phlebotomists played a crucial role in monitoring wristband conformity to the required standard. 6 Such peer review techniques, which have delivered improvement in the quality of laboratory analysis, may bring about similar improvement in these other areas of laboratory operations. Another technique that has been productive in other procedural areas is incident monitoring. 7 It may very well yield fruitful results if applied to laboratory activities. There is obviously much scope for documenting, understanding and minimising the types of laboratory errors highlighted by Khoury et al. However, a systematic and co ordinated approach will be required from those involved in laboratory medicine to achieve the all-round quality of performance that clinicians and their patients should come to expect. Stewart J Bryant General Manager, SouthPath, St George Hospital Campus, Sydney, NSW Maynard JH. The regulation of medical laboratories in Australia. Clin Lab Med 1991; 11: 777-791. Nakhleh RE, Zarbo RJ. Surgical pathology specimen identification and accessioning. Arch Pathol Lab Med 1996; 120: 227-233. Craig JC, Knight JF, Smith GH. Communication breakdown: a preventable cause of acute renal failure in a newborn infant. Med J Aust 1996; 164: 663-664. Wilson RM, Runciman WB, Gibberd RW, et al. The Quality in Australian Health Care Study. Med J Aust 1995; 163: 458-471. International Standards Organisation guide 25. General requirements for the competence of calibration and testing laboratories. 3rd Ed. Geneva: International Standards Organisation, 1990. Renner SW, Howanitz PJ, Bachner P. Wristband identification error reporting in 712 hospitals. A College of American Pathologists' Q-Probes Study of Quality Issues in Transfusion Practice. Arch Pathol Lab Med 1993; 117: 573-577. Runciman WB, Sellen A, Webb RK, et al. The Australian Incident Monitoring Study. Errors, incidents and accidents in anaesthetic practice. Anaesth Intensive Care 1993; 21: 506-519. To top of article ©MJA; 1996 < URL: http://www.mja.com.au/> © 1996 Medical Journal of Australia.

Stewart J Bryant

Error rates in Australian chemical pathology laboratories Mounira Khoury, Leslie Burnett and Mark A Mackay MJA 1996; 165: 128-130 For editorial comment, see Bryant Readers may print a single copy for personal use. No further reproduction or distribution of the articles in whole or in part should proceed without the permission of the publisher. For copyright permission, contact the Australasian Medical Publishing Company Journalists are welcome to write news stories based on what they read here, but should acknowledge their source as "an article published on the Internet by The Medical Journal of Australia <http://www.mja.com.au/>". Abstract - Introduction - Methods - Measurement criteria - Participants - Survey - Statistical analysis - Results - Participation - Transcription errors - Analytical performance - Combined error rates - Discussion - Acknowledgements - References - Authors' details Register to be notified of new articles by email - - ©MJA1996 Abstract Objective: To measure transcription and analytical errors made by Australian chemical pathology laboratories. Design: Retrospective data collection covering the period 1 November 1993 to 1 April 1994. Setting and participants: Fourteen pathology laboratories in five Australian States (seven in the public sector, and seven in the private sector). Main outcome measures: Error rates in transcribing information from request forms to computer record systems, and laboratory performance on chemical analysis. Results: Pathology laboratories had a transcription-error rate of up to 39% and an error rate of up to 26% for analytical results. The worst-performing laboratory had errors (of patient identification or results of analysis) in 46% of requests. The three best-performing laboratories achieved 85% error-free reporting, with one achieving 95%. Conclusions: Error rates in Australian pathology laboratories vary widely, but may be as high as 46% for all specimens in some laboratories. The types of errors reported were under the control of the laboratory, and would affect the accuracy of reported pathology test results, with potential adverse outcomes for patient care and inefficient use of health-care resources. There is a need to establish broader quality assurance programs and performance requirements to reduce these types of error. MJA 1996; 165: 128-130 Introduction T he Quality in Australian Health Care Study 1 described an unacceptable rate of preventable adverse events involving patients which occur in Australian hospitals. Errors introduced into the process of requesting pathology tests can adversely affect patient management and will detract from quality of care and services. 2 At best, a sample may need to be collected again to repeat the analysis, inconveniencing the patient, delaying treatment and increasing the cost. At worst, a doctor may act on incorrect results and the patient may be treated inappropriately. In its most common diagnostic setting, the process of requesting pathology tests seems simple: the doctor writes a request, the pathology laboratory interprets the request, transcribes information from the request form to its computer system, performs the analysis, and the pathologist produces a printed report for return to the doctor. This process assumes that the patient has been correctly identified, an appropriate specimen has been obtained, analysis has been correctly performed, and the results of analysis have been communicated to the original requesting practitioner. We recently examined the accuracy of one part of the process of obtaining pathology test results in a hospital environment and found that the address of the referring doctor had been incorrectly recorded on 3% of all printed reports issued, delaying or preventing delivery of the final pathology report. 3 If this rate of error were typical for the many other steps of the requesting process for pathology tests, then the total error rate in obtaining results would be extremely high. To address this question, we surveyed clinical pathology laboratories from both public and private sectors of five Australian States to determine whether error rates in pathology are as widespread and as clinically significant as our earlier study suggested. Methods Measurement criteria We identified potential indicators using national Quality Award criteria for benchmarking indicators, 4 focusing on factors affecting medical outcomes which arose from the pathology laboratory. In this paper, we report on transcription errors and analytical performance in chemical pathology quality assurance programs. Details of these criteria and how they were measured are shown in Box 1. We did not examine how laboratories functioned to deliver these outcomes. Participants We approached 18 large National Association of Testing Authorities-registered laboratories, well recognised in the industry, representing both public and private sectors, in five Australian States (Queensland, New South Wales, Victoria, Tasmania and South Australia). Survey Survey forms, which included detailed instructions and operational defini tions of the information required, were completed by the laboratories between 1 May and 31 July 1994. One hundred hand-written pathology request forms dated between 1 November 1993 and 1 April 1994 were randomly selected by the participating laboratories. We excluded request forms which also contained requests for blood transfusion or tissue pathology because some laboratories adopt more stringent identification criteria for these types of specimens. Laboratories scored the number of transcription errors (defined as any instances where the data on individual request forms were not identical to the data entered into the laboratory's computer system 5 ). Chemical pathology quality assurance (QA) data were from one cycle of analysis within the QA program (there are two or three cycles annually, depending on the program) during 1993 and 1994. Laboratories scored the total number of QA samples analysed, and the proportion of results of these analyses that lay outside the allowable limits of error of the QA program. All data, as well as a summary of the relative performance of all other participants (each assigned a number so that they were not identifiable), were returned to each laboratory for validation before final analysis. Statistical analysis Data analysis was based on the binomial distribution. 5 Calculation of combined error rates was based on the product of constituent error-free components, with the standard deviation calculated from the sum of variances of individual component errors. Results Participation Fourteen laboratories (seven public and seven private, numbered 1-14 to preserve anonymity) completed the entire benchmarking study; all five States were represented. Transcription errors Box 2 shows our findings for particular types of transcription errors. Most laboratories lay within a tight distribution clustered around median error rates of 1% to 3%. For each transcription error listed, a single outlier -- a different laboratory in each case -- performed particularly badly. The Figure shows that, while two laboratories were able to achieve overall transcription-error rates of less than 3%, four had error rates more than two standard deviations above the mean error rate (13%) of participating laboratories ( P < 0.05). The laboratory with the poorest performance reported a transcription-error rate of 39%. Analytical performance Laboratories recorded errors outside the allowable limits of error in up to 26% of analytical results. As shown in the Figure, four laboratories reported unacceptable performance more than two standard deviations above the mean rate (11.4%) of participating laboratories ( P < 0.05). Combined error rates By matching the transcription-error rate with analytical performance, we were able to measure the proportion of all pathology requests that would have been free of either type of error (Figure). Three laboratories showed superior performance compared with peers, achieving 85% error-free reporting, with one achieving 95% error-free reporting (a total error rate of less than 5%). These three laboratories represented both the public and private sectors, were all from different States, and each used a different computer system and clinical chemistry analyser. The two laboratories which showed the poorest performance reported transcription errors or unacceptable ana lytical performance in more than a third of all requests. They were both hospital laboratories from the same State. Their high total error rates were particularly influenced by transcription errors. Discussion This pilot survey of laboratory performance is the first of its kind to have been done in Australia. Previous studies have reported particular types of laboratory errors (e.g., patient ward location or address) in 0.08% 6 to 3% 3,6,7 of requests. We have confirmed that median error rates for individual types of errors in routine pathology requests are of the order of 3% 3 (see Box 2), and overall error rates can be as high as 46%. While some of the laboratories operated with few or no errors, eight of the fourteen reported significantly inferior performance (Figure). Further, our findings indicate that some laboratories may be making transcription errors in up to 39% of all request forms, and can have error rates of 9% in patient identification data and 17% in data such as sex and date-of-birth. Such error rates should be a cause of clinical concern. Risk management in pathology laboratories recognises the relationship between proper identification of specimens and the prevention of adverse patient-care-related incidents. 2 Our study was not designed to trace particular errors to final clinical outcomes, or to distinguish between errors on the basis of severity. However, it would be surprising if transcription-error rates of the magnitude we encountered did not contribute to adverse clinical outcomes. Are the errors caused by the requesting doctor, or by the laboratory? Several laboratories used direct electronic capture of patient data from a database, which should minimise errors in patient identification data arising from illegible handwriting. These laboratories did not have lower rates of transcription error for patient identification data than those relying solely on handwritten pathology requests. Indeed, the two laboratories with the highest overall rates of transcription error both obtained their patient identification data from a central database. This leads us to conclude that poor handwriting by the requesting doctor was not the major factor in high rates of transcription errors. Errors in pathological tests come at a considerable cost to the community. Expenditure on pathology investigations in Australia amounts to some $1.4 billion -- approximately 4% of total health services expenditure (Australian Association of Pathology Practices, personal communication). Incorrect tests being performed in up to 15% of cases wastes medical resources. Requesting doctors being incorrectly identified on up to 17% of requests will result in laboratories incurring unnecessary delivery costs and doctors incurring additional costs, as they will need to obtain duplicate copies of reports not delivered, and will request repeat testing for results unable to be located. We believe that most of these errors could have been prevented by the laboratories. The three best-performing laboratories in our study used different equipment and computer systems, and operated in both public and private sector environments. We therefore suggest that superior overall performance can be achieved by controlling error rates at each step in the process, and that this can be done in diverse ways. All recognised medical testing laboratories within Australia are required to be accredited, and to participate in QA programs for analytical quality. However, there are no minimum standards of performance which laboratories are required to maintain within these QA programs. There are also no Australian QA programs which monitor non-analytical aspects of pathology laboratories, such as transcription errors. We recommend that QA programs should be introduced for a range of clinically important areas of laboratory performance, such as request transcription accuracy, transport and analysis times for key analytes (e.g., potassium), rapid reporting of critical patient data, and client satisfaction, in addition to existing programs for analytical quality. We also recommend that summaries of actual and achievable performance in these QA programs be published. Without laboratory participation and satisfactory performance in such programs, doctors cannot be confident that test results are correct. Indeed, they cannot even be confident that the result was obtained for the correct patient. Lower error rates in pathology tests should lead to improved patient outcomes and less wastage of health resources. Acknowledgements We thank the staff of the participating laboratories for their assistance with this study. We are grateful to Elizabeth Benson, Jon Currie, Jeremy Chapman, Doug Chesher and Ruth Pojer for their critical comments on the manuscript. References Wilson RM, Runciman WB, Gibberd RW, et al. The Quality in Australian Health Care Study. Med J Aust 1995; 163: 458-471. Lord JT. Risk management in pathology and laboratory medicine. Arch Pathol Lab Med 1990; 114: 1164-1167. Banning J, Brown J, Hooper L, et al. Reduction of errors in laboratory test reports using continuous quality improvement (CQI) techniques. Clin Lab Management Rev 1993; 7: 424-437. Australian Quality Awards Foundation. Assessment criteria and application guidelines. Sydney: Australian Quality Council, 1994. Grant EL, Leavenworth RS. Statistical quality control. 6th ed. Singapore: McGraw-Hill. 1988. Burnett L, Banning J. Reduction of errors in laboratory test reports: comparison of continuous quality improvement techniques with laboratory information system techniques. In: Parkany M, editor. Quality assurance and TQM for analytical laboratories. Cambridge, UK: The Royal Society of Chemistry, 1995: 97-101. Chambers AM, Elder J, O'Reilly DStJ. Blunder-rate in a clinical biochemistry service. Ann Clin Biochem 1986; 23: 470-473. (Received 20 Nov 1995, accepted 22 May 1996) Authors' details Faculty of Clinical Chemistry, School of Life Sciences, University of Technology (Gore Hill Campus), Sydney, NSW. Mounira Khoury, MSc(Clin Biochem), Graduate Student. Institute of Clinical Pathology and Medical Research, Westmead Hospital, Sydney, NSW. Leslie Burnett, PhD, FRCPA, Director of Clinical Chemistry. Mark A Mackay, MSc, Hospital Scientist and Quality Facilitator. Reprints: Associate Professor Leslie Burnett, Assistant Director, Institute of Clinical Pathology and Medical Research, Westmead Hospital, Westmead, NSW 2145. < URL: http://www.mja.com.au/> © 1996 Medical Journal of Australia.

Mounira Khoury · Leslie Burnett · Mark A Mackay