Bloembergen WE, Port FK, Mauger EA, Wolfe RA
A comparison of mortality between patients treated with
hemodialysis and peritoneal dialysis and a comparison of
causes of death (two back-to-back papers)
J Am Soc Nephrol
(Aug) 6:177-191 1995
These papers use the enormous USRDS database to compare
mortality rates between HD and PD. Previous studies from different
countries on this topic have shown conflicting results. Bloembergen et
al, however, show a significantly higher mortality
on PD with a relative risk of 1.19 compared to HD. This is after correction
for age, race, sex and diabetes. The excess risk is only significant at
age over 55 and is most marked in diabetics and females. It is most
marked for deaths from infection and cardiovascular disease. Overall,
the excess mortality represents 4 deaths per 100 patient years.
These studies have some notable features. Firstly, they have
enormous statistical power with 170,000 patient years and 42,000
deaths from 1987 to 1989. Secondly, they deal with the difficult issue of
modality switches by treating each patient as a member of a new cohort
at the start of each year. This approach, which has obvious
advantages, contrasts with that in other studies which ignore or censor
modality switches.
These studies may have a detrimental effect on PD, particularly in
the US. A number of points should be noted, however, before they are
overinterpreted. Firstly, Bloembergen et al adjust for age, sex, race and
diabetes, but not for other comorbidities. It is possible that the
latter explain the
excess mortality on PD. Registry data are not controlled trials, modality
selection is not random and unmeasured characteristics of the population
directed to PD in the US may explain the higher mortality. Such a view is
supported by data from the recently published 1993 Canadian Registry
showing a significantly lower mortality on PD than on HD with relative
risks of .89 and .76, under and over age 65, after correction for sex,
age, time and diabetes. Modality selection in Canada is unlike that in the
US and the striking differences in relative outcomes suggest case mix
may be confounding all these analyses.
A second point is that the practice of both PD and HD is
constantly changing. PD practices were not optimal in the US during the
period covered by the Bloembergen studies. 'Flush before fill' giving sets
were less widely used than in Canada or Europe and a high proportion
of patients received less than 8 liters of dialysate daily. It may be
relevant that data from the multicenter, CANUSA, PD adequacy study to
be presented at ASN 1995 show over 10% higher mortality at 2 years in
he US as compared to Canada, even after correction for demographics,
comorbidity, dialytic dose and nutritional indices. This again suggests
either a case mix difference not picked up by the parameters for which
the data were corrected or, alternatively, a persisting problem with PD
practice in the US. Examples of such a problem might be greater
noncompliance with exchanges or less widespread use of newer, bettergiving
sets.
In summary, there are reasons to justify some scepticism about
how applicable the findings of Bloembergen et al are in the mid 1990's
inside and, particularly, outside the US. It would be unjustifed to
conclude that PD is an inferior therapy to HD. Nevertheless, these
studies are an incentive to give attention in PD to issues such as dialytic
dose, newer giving sets, compliance, volume overload, malnutrition,
dyslipidemia and appropriate use of automated PD. A randomized trial
comparing PD and HD may one day be done although many doubt it's
feasibility. For now, both modalities should be viewed as complementary
and in need of constant improvement.
(Blake)
To go back use the BACK button on your browser.
Otherwise click on the desired link to this article below:
CRF: Problem Areas :
Outcomes (Morbidity, Mortality)