- Email: [email protected]
Analyse transplant outcomes in distinct epochs of follow-up
Correspondence
discover whether the availability of level 1 data from the trial has led to any convergence of views among the previously studied cohort of surgeons and radiologists. I declare that I have no conflict of interest.
Andrew W Bradbury [email protected]. uk
Science Photo Library
Department of Vascular Surgery, University of Birmingham, Heart of England NHS Foundation Trust, Birmingham B9 5SS, UK
Rights were not granted to include this image in electronic media. Please refer to the printed journal.
1816
Analyse transplant outcomes in distinct epochs of follow-up Andrew Burroughs and colleagues (Jan 21, p 225)1 report country-unadjusted predictive models for 3-month and 12-month mortality after the first liver transplant in 1988–2003 for adults in Europe. No-names UK data were reported to the European Liver Transplant Registry (ELTR) by UK Transplant which, from the late 1980s, had recognised, and promoted, the importance of analysing transplant outcomes—kidney,2 liver,3 or heart4— in distinct epochs of follow-up. It is not merely disappointing but, more importantly, prejudicial to transplantation sciences if new statistical analyses overlook old intelligence. And more so when the same issue—tumour recurrence 3–12 months after liver transplantation for hepatocellular carcinoma—persists as in earlier analysis.3 Change in sign of prognostic effect plus non-overlapping confidence intervals1 should alert readers (and analysts) that the prognostic influence associated with hepatocellular carcinoma had switched from being favourable for surviving the first 3 months after liver transplantation to being a risk factor for (metastasisrelated) mortality 3–12 months after the procedure. This occurrence was recognised in the first major statistical analysis of the UK’s Cambridge-Kings Liver Transplant
series3—with two consequences. First, Addenbrooke’s Hospital introduced staging laparotomy to obviate liver transplantation in the presence of pre-existing metastatic disease. Second, the finding passed into (and now apparently out of) the folklore of transplantation and medical statistics as a classic example of the need for analysts to look not just for frailty of prognostic influences, which is a common departure from constant proportional hazards or odds, but also to check for qualitatively different, but potentially biologically relevant, effects. Burroughs and colleagues are to be congratulated that the above major change in prognostic effect, although obscured by the chosen modelling framework, was clearly recognisable by us from their excellent table 1 of descriptive statistics. 12% of around 20 000 patients had died within 3 months, and 18% within 12 months, of first liver transplantation. But, for the subgroup of around 2000 patients transplanted for hepatocellular carcinoma, corresponding figures were 9% (95% CI 7·8–10·2) and 20%. The same table points to a second familiar reason for transplant analyses in distinct epochs. Whereas 3-month mortality, before and after adjustment for covariates, had decreased importantly in 1996–2003 compared with 1988–95, the same does not apply for the unadjusted death rate for the distinct epoch of 3–12 months after transplantation, which hovered at around 9% of 3-month survivors in each 3-year calendar period of liver transplants. Thus, we hope that ELTR will focus on how patients’ characteristics at operation (if at all, other than hepatocellular carcinoma) influence mortality rate by 1 year after liver transplantation for patients who had survived to 3 months. In view of current or projected UK epidemics of alcohol-related, hepatitis-Crelated,5 or obesity-related late liver disease, particular attention to those
who received liver transplants for alcoholic cirrhosis, hepatitis-C-related cirrhosis, obesity-related cirrhosis, or combinations, is key. We declare that we have no conflict of interest.
*Sheila M Bird, Roy Y Calne, Linda D Sharples [email protected] Medical Research Council Biostatistics Unit, Cambridge CB2 2SR, UK (SMB, LDS); and Office of Sir Roy Calne, Cambridge CB2 2AS, UK (RYC) 1
2
3
4
5
Burroughs AK, Sabin CA, Rolles K, et al. 3-month and 12-month mortality after first liver transplant in adults in Europe: predictive models for outcome. Lancet 2006; 367: 225–32. Gilks WR, Gore SM, Bradley BA. Analysing transplant survival data. Transplantation 1986; 42: 46–49. Gore SM, Barroso E, White DJG. Risk factors in orthotopic first liver transplantation. In: Calne R, ed. Liver transplantation, 2nd edn. London: Grune and Stratton, 1987: 513–30. Sharples LD, Caine N, Mullins P, et al. Risk factor analysis for the major hazards following heart transplantation—rejection, infection and coronary occlusive disease. Transplantation 1991; 52: 244–52. Hutchinson SJ, Bird SM, Goldberg DJ. Modelling the current and future disease burden of hepatitis C among injection drug users in Scotland. Hepatology 2005; 42: 711–23.
Authors’ reply Sheila Bird and colleagues correctly note that, although a diagnosis of hepatocellular carcinoma is a significant favourable prognostic factor for survival over the first 3 months after transplantation, it does not remain significantly associated with survival over the first year, suggesting that any possible beneficial effect in the first 3 months might be reversed with longer follow-up. We had not overlooked this fact, but due to lack of space, comment on it was omitted. The explanation is likely to lie in the fact that selection criteria for hepatocellular carcinoma in those with cirrhosis have varied in terms of tumour load. Many centres confine selection to either a single nodule of no more than 5 cm in diameter, or to three nodules or less, each no more than 3 cm in diameter on the basis of preoperative imaging.1 Moreover, these www.thelancet.com Vol 367 June 3, 2006
discover whether the availability of level 1 data from the trial has led to any convergence of views among the previously studied cohort of surgeons and radiologists. I declare that I have no conflict of interest.
Andrew W Bradbury [email protected]. uk
Science Photo Library
Department of Vascular Surgery, University of Birmingham, Heart of England NHS Foundation Trust, Birmingham B9 5SS, UK
Rights were not granted to include this image in electronic media. Please refer to the printed journal.
1816
Analyse transplant outcomes in distinct epochs of follow-up Andrew Burroughs and colleagues (Jan 21, p 225)1 report country-unadjusted predictive models for 3-month and 12-month mortality after the first liver transplant in 1988–2003 for adults in Europe. No-names UK data were reported to the European Liver Transplant Registry (ELTR) by UK Transplant which, from the late 1980s, had recognised, and promoted, the importance of analysing transplant outcomes—kidney,2 liver,3 or heart4— in distinct epochs of follow-up. It is not merely disappointing but, more importantly, prejudicial to transplantation sciences if new statistical analyses overlook old intelligence. And more so when the same issue—tumour recurrence 3–12 months after liver transplantation for hepatocellular carcinoma—persists as in earlier analysis.3 Change in sign of prognostic effect plus non-overlapping confidence intervals1 should alert readers (and analysts) that the prognostic influence associated with hepatocellular carcinoma had switched from being favourable for surviving the first 3 months after liver transplantation to being a risk factor for (metastasisrelated) mortality 3–12 months after the procedure. This occurrence was recognised in the first major statistical analysis of the UK’s Cambridge-Kings Liver Transplant
series3—with two consequences. First, Addenbrooke’s Hospital introduced staging laparotomy to obviate liver transplantation in the presence of pre-existing metastatic disease. Second, the finding passed into (and now apparently out of) the folklore of transplantation and medical statistics as a classic example of the need for analysts to look not just for frailty of prognostic influences, which is a common departure from constant proportional hazards or odds, but also to check for qualitatively different, but potentially biologically relevant, effects. Burroughs and colleagues are to be congratulated that the above major change in prognostic effect, although obscured by the chosen modelling framework, was clearly recognisable by us from their excellent table 1 of descriptive statistics. 12% of around 20 000 patients had died within 3 months, and 18% within 12 months, of first liver transplantation. But, for the subgroup of around 2000 patients transplanted for hepatocellular carcinoma, corresponding figures were 9% (95% CI 7·8–10·2) and 20%. The same table points to a second familiar reason for transplant analyses in distinct epochs. Whereas 3-month mortality, before and after adjustment for covariates, had decreased importantly in 1996–2003 compared with 1988–95, the same does not apply for the unadjusted death rate for the distinct epoch of 3–12 months after transplantation, which hovered at around 9% of 3-month survivors in each 3-year calendar period of liver transplants. Thus, we hope that ELTR will focus on how patients’ characteristics at operation (if at all, other than hepatocellular carcinoma) influence mortality rate by 1 year after liver transplantation for patients who had survived to 3 months. In view of current or projected UK epidemics of alcohol-related, hepatitis-Crelated,5 or obesity-related late liver disease, particular attention to those
who received liver transplants for alcoholic cirrhosis, hepatitis-C-related cirrhosis, obesity-related cirrhosis, or combinations, is key. We declare that we have no conflict of interest.
*Sheila M Bird, Roy Y Calne, Linda D Sharples [email protected] Medical Research Council Biostatistics Unit, Cambridge CB2 2SR, UK (SMB, LDS); and Office of Sir Roy Calne, Cambridge CB2 2AS, UK (RYC) 1
2
3
4
5
Burroughs AK, Sabin CA, Rolles K, et al. 3-month and 12-month mortality after first liver transplant in adults in Europe: predictive models for outcome. Lancet 2006; 367: 225–32. Gilks WR, Gore SM, Bradley BA. Analysing transplant survival data. Transplantation 1986; 42: 46–49. Gore SM, Barroso E, White DJG. Risk factors in orthotopic first liver transplantation. In: Calne R, ed. Liver transplantation, 2nd edn. London: Grune and Stratton, 1987: 513–30. Sharples LD, Caine N, Mullins P, et al. Risk factor analysis for the major hazards following heart transplantation—rejection, infection and coronary occlusive disease. Transplantation 1991; 52: 244–52. Hutchinson SJ, Bird SM, Goldberg DJ. Modelling the current and future disease burden of hepatitis C among injection drug users in Scotland. Hepatology 2005; 42: 711–23.
Authors’ reply Sheila Bird and colleagues correctly note that, although a diagnosis of hepatocellular carcinoma is a significant favourable prognostic factor for survival over the first 3 months after transplantation, it does not remain significantly associated with survival over the first year, suggesting that any possible beneficial effect in the first 3 months might be reversed with longer follow-up. We had not overlooked this fact, but due to lack of space, comment on it was omitted. The explanation is likely to lie in the fact that selection criteria for hepatocellular carcinoma in those with cirrhosis have varied in terms of tumour load. Many centres confine selection to either a single nodule of no more than 5 cm in diameter, or to three nodules or less, each no more than 3 cm in diameter on the basis of preoperative imaging.1 Moreover, these www.thelancet.com Vol 367 June 3, 2006