computerized advice on drug dosage to improve prescribing practice

Computerized advice on drug dosage to improve prescribing

practice (Review)

Durieux P, Trinquart L, Colombet I, Niès J, Walton RT, Rajeswaran A, Rège-Walther M,

Harvey E, Burnand B

This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library2010, Issue 10

http://www.thecochranelibrary.com

Computerized advice on drug dosage to improve prescribing practice (Review)

Copyright © 2010 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd.

http://www.thecochranelibrary.com

T A B L E O F C O N T E N T S

1HEADER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2PLAIN LANGUAGE SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3BACKGROUND . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3OBJECTIVES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3METHODS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

5RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

9DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10AUTHORS’ CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

11REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14CHARACTERISTICS OF STUDIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

42DATA AND ANALYSES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Analysis 1.1. Comparison 1 Dose of drug used, Outcome 1 Dose administered to the patient. . . . . . . . . 44

Analysis 1.2. Comparison 1 Dose of drug used, Outcome 2 Number of doses adjustments. . . . . . . . . . 45

Analysis 2.1. Comparison 2 Serum concentrations and therapeutic range, Outcome 1 Serum concentrations. . . . 45

Analysis 2.2. Comparison 2 Serum concentrations and therapeutic range, Outcome 2 Percentage of patients within

therapeutic range. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

Analysis 2.3. Comparison 2 Serum concentrations and therapeutic range, Outcome 3 Toxic Drug Levels. . . . . 47

Analysis 3.1. Comparison 3 Physiological parameters, Outcome 1 Mean proportion of time spent within target. . . 47

Analysis 4.1. Comparison 4 Time to achieve therapeutic control, Outcome 1 Time to achieve therapeutic range. . . 48

Analysis 4.2. Comparison 4 Time to achieve therapeutic control, Outcome 2 Time to stabilization. . . . . . . 48

Analysis 5.1. Comparison 5 Clinical events, Outcome 1 Death. . . . . . . . . . . . . . . . . . . . 49

Analysis 5.2. Comparison 5 Clinical events, Outcome 2 Adverse reactions. . . . . . . . . . . . . . . . 50

Analysis 5.3. Comparison 5 Clinical events, Outcome 3 Improvement. . . . . . . . . . . . . . . . . 50

Analysis 6.1. Comparison 6 Health care costs, Outcome 1 Length of stay. . . . . . . . . . . . . . . . 51

51APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52FEEDBACK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

55WHAT’S NEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

55HISTORY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56CONTRIBUTIONS OF AUTHORS . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56DECLARATIONS OF INTEREST . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

56INDEX TERMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

iComputerized advice on drug dosage to improve prescribing practice (Review)


[Intervention Review]

Computerized advice on drug dosage to improve prescribingpractice

Pierre Durieux1 , Ludovic Trinquart2 , Isabelle Colombet3 , Julie Niès3, RT Walton4 , Anand Rajeswaran5 , Myriam Rège-Walther6,

Emma Harvey7, Bernard Burnand6

1Epidemiology and Clinical Research Unit, Georges Pompidou European Hospital, Paris Descartes University, INSERM CIE4, Paris,

France. 2Clinical Research Unit and INSERM CIE 4, Georges Pompidou European Hospital, Paris, France. 3Medical Informatics

Department, Georges Pompidou European Hospital, Paris Descartes University, INSERM U872 eq20, Paris, France. 4Centre for

Health Sciences, Barts and the London Medical School, London, UK. 5Clinical Epidemiology Centre, University Institute of Social

and Preventive Medicine, Lausanne, Switzerland. 6Health Care Evaluation Unit & Clinical Epidemiology Centre, Institute of Social

and Preventive Medicine, Centre Hospitalier Vaudois and University of Lausanne, Lausanne, Switzerland. 7SaltaSustainable, Leeds,

UK

Contact address: Pierre Durieux, Epidemiology and Clinical Research Unit, Georges Pompidou European Hospital, Paris Descartes

University, INSERM CIE4, 20 rue Leblanc, Paris, 75015, France. [email protected].

Editorial group: Cochrane Effective Practice and Organisation of Care Group.

Publication status and date: Edited (no change to conclusions), comment added to review, published in Issue 10, 2010.

Review content assessed as up-to-date: 13 May 2008.

Citation: Durieux P, Trinquart L, Colombet I, Niès J, Walton RT, Rajeswaran A, Rège-Walther M, Harvey E, Burnand B. Computerized

advice on drug dosage to improve prescribing practice. Cochrane Database of Systematic Reviews 2008, Issue 3. Art. No.: CD002894.

DOI: 10.1002/14651858.CD002894.pub2.


A B S T R A C T

Background

Maintaining therapeutic concentrations of drugs with a narrow therapeutic window is a complex task. Several computer systems have

been designed to help doctors determine optimum drug dosage. Significant improvements in health care could be achieved if computer

advice improved health outcomes and could be implemented in routine practice in a cost effective fashion. This is an updated version

of an earlier Cochrane systematic review, by Walton et al, published in 2001.

Objectives

To assess whether computerised advice on drug dosage has beneficial effects on the process or outcome of health care.

Search strategy

We searched the Cochrane Effective Practice and Organisation of Care Group specialized register (June 1996 to December 2006),

MEDLINE (1966 to December 2006), EMBASE (1980 to December 2006), hand searched the journal Therapeutic Drug Monitoring

(1979 to March 2007) and the Journal of the American Medical Informatics Association (1996 to March 2007) as well as reference

lists from primary articles.

Selection criteria

Randomized controlled trials, controlled trials, controlled before and after studies and interrupted time series analyses of computerized

advice on drug dosage were included. The participants were health professionals responsible for patient care. The outcomes were: any

objectively measured change in the behaviour of the health care provider (such as changes in the dose of drug used); any change in the

health of patients resulting from computerized advice (such as adverse reactions to drugs).

1Computerized advice on drug dosage to improve prescribing practice (Review)


mailto:[email protected]

Data collection and analysis

Two reviewers independently extracted data and assessed study quality.

Main results

Twenty-six comparisons (23 articles) were included (as compared to fifteen comparisons in the original review) including a wide range

of drugs in inpatient and outpatient settings. Interventions usually targeted doctors although some studies attempted to influence

prescriptions by pharmacists and nurses. Although all studies used reliable outcome measures, their quality was generally low.

Computerized advice for drug dosage gave significant benefits by:

1.increasing the initial dose (standardised mean difference 1.12, 95% CI 0.33 to 1.92)

2.increasing serum concentrations (standradised mean difference 1.12, 95% CI 0.43 to 1.82)

3.reducing the time to therapeutic stabilisation (standardised mean difference -0.55, 95%CI -1.03 to -0.08)

4.reducing the risk of toxic drug level (rate ratio 0.45, 95% CI 0.30 to 0.70)

5.reducing the length of hospital stay (standardised mean difference -0.35, 95% CI -0.52 to -0.17).

Authors’ conclusions

This review suggests that computerized advice for drug dosage has some benefits: it increased the initial dose of drug, increased serum

drug concentrations and led to a more rapid therapeutic control. It also reduced the risk of toxic drug levels and the length of time

spent in the hospital. However, it had no effect on adverse reactions. In addition, there was no evidence to suggest that some decision

support technical features (such as its integration into a computer physician order entry system) or aspects of organization of care (such

as the setting) could optimise the effect of computerised advice.

P L A I N L A N G U A G E S U M M A R Y

Computerized advice on drug dosage to improve prescribing practice

Physicians and other health care professionals often prescribe drugs that will only work at certain levels. These drugs are said to have a

narrow therapeutic window. This means that if the level of the drug is too high or too low, they may cause serious side effects or not

provide the benefits they should. For example, blood thinners are prescribed to thin the blood to prevent clots. If the level is too high,

people can bleed to death. On the other hand, if the level is too low, a clot could form and cause a stroke. For these types of drugs, it

is important that the right amount of the drug is prescribed.

Calculating and prescribing the right amount can be complicated and time-consuming for health care professionals. Sometimes

determining the right amount can take a long time since health professionals may not want to prescribe high doses of the drugs

right away or sometimes they make mistakes. Several computer systems have been designed to do these calculations and assist health

professionals to prescribe these types of drugs.

A review of studies that evaluated these computer systems showed that computerized advice for drug dosage can benefit both health

professionals and patients. When using the computer system, health professionals prescribed higher doses of the drugs right away and

the right amount of the drug was reached quicker. Using the computer systems also reduced the length of time patients spent in the

hospital while the right amount of the drug was reached. However, the computer systems did not increase or decrease how often serious

side effects, such as strokes or death, occurred.



B A C K G R O U N D

Medication errors still represent 20% of medical errors although

many efforts have focused in recent years on reducing them (

Kaushal 2006). Maintaining therapeutic concentrations of drugs

is a complex task requiring knowledge of evidence based clinical

guidelines, clinical pharmacology and skills in dose calculation.

The potential for error is great since many of the drugs commonly

used have a narrow ’window’ within which therapeutic benefits

can be obtained with a low risk of unwanted effects.

Monitoring drug therapy to optimise effects and minimise dan-

gers can be very time-consuming. Practitioners may need access

to a large amount of information to make an appropriate pre-

scription in situations such as prevention of deep vein thrombo-

sis or management of patients with renal insufficiency (Durieux

2005). Under these conditions, health professionals make errors

of judgement because their ability to process information is fi-

nite (McDonald 1976). Moreover, physicians’ computational skills

are often inadequate to perform calculations about drug dosage

(Baldwin 1995). For example, 82 out of 150 hospital doctors were

unable to calculate how many milligrams of lidocaine were in a

10 ml ampoule of 1% solution (Rolfe 1995).

Decision support systems, either computerized or not, have been

proposed to improve clinical practice (Kawamoto 2005). Com-

puters are very good at collecting information and performing

repetitive calculations. Moreover, the drugs which cause most of

the problems have often been in use for many years, the phar-

macology of the drugs is therefore well understood and computer

models may be used to generate advice on dosage. Several com-

puter systems have been designed to help doctors in the task of

determining the optimum dosage of drugs. Significant improve-

ments in health could be achieved if computer advice was shown

to be beneficial and was provided by the computers that clinicians

now use for their everyday work.

In addition, the logistics by which the advice on drug dosage is

delivered to the health professional is critical to its effectiveness

and to the transferability of this effectiveness in other settings.

Computer physician order entry (CPOE) systems, which allow

physicians to enter orders directly into a computer rather than

handwriting them, have the potential to incorporate clinical de-

cision support into daily practice (Kuperman 2003). According

to two recent systematic reviews (Garg 2005; Kawamoto 2005),

clinical decision support systems are more often associated with

improvement of practice when the decision aid is automatically

prompted, integrated in clinicians workflow and provided at time

and location of decision making.

This is an updated version of an earlier Cochrane systematic review

(Walton 2001).This earlier review provided evidence to support

the use of computer assistance in determining drug dosage but

concluded that further clinical trials were necessary to confirm

those results.

O B J E C T I V E S

To determine:

1. whether there is evidence that computerized advice on drug

dosage is beneficial

2. whether any technical features of computerized systems or

organizational aspects concerning their implementation are

critical to obtain this benefit.

Hypotheses Tested

Effect on Process of care (health professional related)

1. Computer advice leads to a change in drug dosage.

Effect on Outcome of care (patient related)

1. Decisions on drug dosage based on computer advice lead

more often to drug levels within the therapeutic range.

2. Decisions on drug dosage based on computer advice lead

more often to a physiological parameter being maintained within

the desired range (for example, blood pressure or prothrombin

time).

3. Decisions on drug dosage based on computer advice lead to

more rapid therapeutic control, assessed by a physiological

parameter.

4. Decisions on drug dosage based on computer advice lead to

fewer unwanted effects than conventional dose adjustment.

5. Computer advice reduces the cost of health care or the use

of resources (length of stay).

Effect of decision support logistics and organization of

care

1. Computer advice given in real time is more effective than

that given by delayed feedback.

2. Computer advice integrated in CPOE system is more

effective than other systems.

3. System-initiated computer advice is more effective than

user-initiated computer advice.

4. Direct intervention (system delivers advice directly to the

provider) is more effective than indirect intervention (advice is

made available to the provider by the intermediate of a third

party actor, i.e. system is not directly used by the provider).

5. The impact of computer advice depends on the setting

where it is implemented (inpatient versus outpatient care).

M E T H O D S



Criteria for considering studies for this review

Types of studies

We included the following types of studies:

• Randomized controlled trials

• Controlled clinical trials

• Controlled before and after studies

• Interrupted time series analyses

(see the EPOC checklist for definition of designs)

Types of participants

Any health professional (for example doctors, nurses or pharma-

cists) with responsibility for patient care.

Patients receiving drug therapy based on:

1. Advice from a computer

2. Advice from any other source

3. Unassisted clinical judgement

Types of interventions

We sought to identify all comparative studies of computer advice

on drug dosage. We defined computer advice on drug dosage as

follows: after a health professional types in data, for example about

the patient’s age, weight and previous drug levels, the program

calculates the most appropriate drug dose, often using individu-

alised mathematical models of the distribution of the drug in the

patient’s body. In most interventions, the drug was administered

by a nurse in tablet form. However, we included studies where

the computer directly administered the drug to the patient, for

example as an infusion. Studies where the computer-controlled

infusion was not under the control of a clinician were excluded.

Some recent studies evaluated systems allowing patient self-man-

agement of oral anticoagulation, but whether a computerized sys-

tem was used or not was rarely accurately reported. A Cochrane

protocol (Garcia 2002) (Garcia 2002) addresses the evaluation of

anticoagulant self-management. Interventions based on comput-

erized advice delivered directly to the patient through self-dosing

device were excluded from this review.

Types of outcome measures

In order to test our hypotheses, we defined the following outcomes.

Process of care (health professional related):

1. Difference in therapeutic regimen across study groups:

initial, maintenance and total doses.

2. Proportion of patients where the therapeutic regimen is

changed due to computer advice: number or proportion of dose

changes, proportion of appropriate orders.

3. Toxic drug levels.

Outcome of care (patient related):

1. Changes in therapeutic drug levels, proportion of patients,

or patient-time with plasma drug concentrations within

therapeutic range.

2. Proportion of patients, or patient-time with physiological

parameters within therapeutic range.

3. Time to achieve therapeutic control.

4. Proportion of patients with unwanted effects of drug

therapy.

5. Proportions of deaths

6. Proportion of patients with clinical improvement.

7. Resources used: Length of stay.

Search methods for identification of studies

For the initial review, the authors searched the Cochrane Effec-

tive Practice and Organisation of Care Group (EPOC) special-

ized register (June 1996 to December 2006), MEDLINE (1966

to June 1996), EMBASE (1980 to June 1996), hand searched

the Therapeutic Drug Monitoring journal (1979 to June 1996),

reference lists from primary articles, and made contact with ex-

perts. The search strategy had no language restrictions. Search

terms were (“Computer Systems”[MESH] OR “Artificial intelli-

gence”[MESH]) AND (prescr* OR “drug therapy”[MESH] AND

(“Comparative Study”[MESH] OR “Clinical Trials”[MESH]).

From the initial review, all included studies were reviewed again

and all studies awaiting assessment were searched and examined.

We further searched the Cochrane Effective Practice and Organ-

isation of Care Group (EPOC) specialized register (June 1996 to

December 2006), MEDLINE (June 1996 to March 2007), EM-

BASE (June 1996 to March 2007), reference lists from primary ar-

ticles. We hand searched the Therapeutic Drug Monitoring jour-

nal (June 1996 to March 2007) and the Journal of the American

Medical Informatics Association (January 1996 to March 2007).

We searched without language restrictions. The search strategy is

included in Appendix 1.

Five additional studies were identified through a personal com-

munication. These were added to the studies awaiting assessment.

The search will be modified in the next update so that these five

studies are retrieved.

Data collection and analysis

Two reviewers (PD, JN) independently screened the search results

for relevance. Each selected study was then randomly allocated to

two reviewers (among IC, PD, MR, AR) who reviewed it and ex-

tracted data independently. Disagreements were resolved by group

discussion with the four reviewers and a statistician (LT).

The quality of the studies was assessed using the criteria de-

scribed by the EPOC group and data were extracted using the



EPOC checklist (see Editorial Information under Group Details

for Methods used in Reviews).

The checklist was adapted to the specific subject. In order to study

the decision support technical features by which the advice on

drug dosage was delivered to the health professional, four key

items drawn from a previous systematic review (Nies 2006) were

extracted from each paper :

• Was the computerized advice delivered in real time (at the

moment of the practitioners decision making) or by delayed

feedback?

• Was the computerized advice integrated in a CPOE?

• Was the computerized advice user-initiated or system-

initiated?

• Was the intervention direct or indirect (a third party brings

advice from computer and transfers it to user)?

For dichotomous variables, we used the relative risk. When the

outcomes were continuous variables, we calculated standardised

mean differences (SMD) with 95% confidence intervals. The stan-

dardised mean difference is a statistical measure of the impact of

the intervention, which is independent of the units used to mea-

sure study outcomes. This measure allows studies of the same in-

tervention using different outcomes to be compared.

For example, measurement of drug concentrations in blood in

different studies may use different assays in several laboratories

and results may be reported in different units. The standardised

mean difference compares differences between experimental and

control groups to the standard deviation of the outcome for each

study. Hence, a quantitative approximation can be made of the

overall effect of decision support on plasma levels.

The effect sizes were combined to give an overall effect for each sub-

group of studies, using a random effects model with correction for

small sample size. The random effects model was chosen because

it does not assume that all interventions have the same underlying

effect.

Each study was deemed favourable to computerized advice when a

statistically significant improvement, in favour of the intervention,

was observed in at least 50% of all its abstracted outcomes (Garg

2005). Then, we assessed whether the decision support logistics by

which the advice was given were associated with positive studies

or not (Garg 2005).

R E S U L T S

Description of studies

See: Characteristics of included studies; Characteristics of excluded

studies.

When several relevant comparisons were reported in one study,

each comparison was considered for our review. Of the 57 com-

parisons that were reviewed for potential inclusion, 15 compar-

isons were included in the initial review and 42 were identified by

literature search. We identified 26 comparisons (23 articles) that

met the eligibility criteria and were included in the final review.

Thirteen comparisons were added to 13 of the comparisons anal-

ysed in the initial review. Four included comparisons had a po-

tential unit of analysis error (Ageno 1998; Chertow 2001; Poller

1998; Poller 1998a). We didn’t include those comparisons in the

meta-analysis but we reported their results in parallel with the

comparisons without unit of analysis error, in the appropriate re-

sults section . We excluded 31 comparisons, among them seven

for an inappropriate design, five for absence of relevant data for

primary outcome. Two comparisons included in the initial review

were excluded from this review because the design did not fit the

inclusion criteria (Alvis 1985) or because the intervention was not

addressed to health professionals (Willcourt 1994).

In one publication concerning warfarin dosage adjustment (Carter

1987), three groups were studied: we reviewed only the compar-

ison between the group using an analog-computer method and

the group using empiric dosing (control). The third group, using

a linear regression model, was excluded because it didn’t involve

any computer assistance.

In one publication (Manotti 2001), two different groups of pa-

tients were studied: a group starting oral anticoagulants (induc-

tion) and a group on long term treatment (maintenance). The

maintenance study was not reviewed because of the absence of

relevant data for primary outcome.

In one publication, two different studies were reported: the first

study (Vadher 1997 pop 1) considered patients starting warfarin

with a targeted INR between 2 and 3; the second study (Vadher

1997 pop2) considered patients on long-term treatment with a

targeted INR between 3 and 4.5.

In all comparisons except one, the objective was to increase the

dose of drug administered (health professionals are reluctant to

expose the patients to adverse effects of drug therapy). In con-

trast, in two comparisons dealing with anaesthesiology (Theil 1993

fentanyl;Theil 1993 midazolam), the objective was to obtain a

lower administered drug dose in order to provide a reduction in

time for extubation. Both Midazolam and Fentanyl infusions were

analysed in the same population. Therefore, these two drugs were

reviewed separately and this study was not included in the corre-

sponding meta-analyses.

In one publication (Fitzmaurice 2000), there were two levels of

randomisation. Practices were randomised to intervention or con-

trol. The study used two control populations: patients individ-

ually randomly allocated to control in the intervention practices

(intrapractice controls) and all patients in the control practices

(interpractice controls). We didn’t analyze interpractice controls

to avoid a possible unit of analysis error.

Characteristics of the providers



The providers were primarily doctors, although ten studies tar-

geted several categories of health professionals including pharma-

cists (Destache 1990; Mungall 1994) and nurses (Destache 1990).

Two studies addressed only nurses’ behaviour (Ruiz 1993; White

1991). Thirteen studies were conducted in North America (12

in the USA, one in Canada). One study took place in Australia

(Hurley 1986), two in New Zealand (Begg 1989; Hickling 1989),

one in Israel (Verner 1992). Nine studies were conducted in Eu-

rope (UK, France, Spain, Italy).

Target behaviour

The target behaviour of the health care provider was the prescrip-

tion and the dosing of drugs.

Characteristics of the interventions

Most of the studies provided advice about appropriate drug

dosages to health care professionals who then decided whether to

follow this or not. Among them, twelve studies evaluated anti-

coagulants (eleven oral anticoagulants, one heparin), four stud-

ies evaluated the administration of aminoglycoside (Begg 1989;

Burton 1991; Destache 1990; Hickling 1989), three studies eval-

uated theophylline (Casner 1993; Gonzalez 1989; Verner 1992),

two comparisons (Theil 1993 fentanyl; Theil 1993 midazolam)

evaluated computer-controlled infusions of anaesthetic agents.

Most of the computer support systems used a mathematical model

of the pharmacokinetics of the drug to predict the required dose.

These models represent the compartments in the body in which

the drug is distributed, with rate constants determining the move-

ment of the drug between different compartments. These systems

allowed the operator to specify a target serum drug level, which the

computer attempted to achieve using Bayesian forecasting meth-

ods. Where the effect of the drug was more important than the

serum level, pharmacodynamic parameters based on population

data could be added to the model (White 1987).

The advice was given in real time to the health professional in

all studies except four, where it was unclear (Poller 1998; Poller

1998a; Vadher 1997; Verner 1992). The computer support sys-

tem was integrated into a CPOE in four studies (Casner 1993;

Chertow 2001; Theil 1993 fentanyl; Theil 1993 midazolam). The

computerized advice was user-initiated in ten studies, system-ini-

tiated in seven studies and it was unclear in nine studies. The in-

tervention was direct in 12 studies, indirect in three studies and

it was unclear in 11 studies. The setting was outpatient care for

nine studies and inpatient care for 16 studies; it was mixed in one

study.

Risk of bias in included studies

Concealment of allocation

All studies were randomized controlled trials, except three (Burton

1991; Chertow 2001; Manotti 2001) which were classified as con-

trolled clinical trials.

Seven studies reported adequate concealment of allocation (the

unit of allocation was by patient and there was some form of

centralised randomisation scheme, for example random numbers

in opaque envelopes).

Protection against contamination

When studies were randomized by patient, the same health pro-

fessional may have given treatment both to intervention and con-

trol groups: it is possible that computerized advice influenced the

treatment of the control groups. Protection against contamination

was considered to be done only in one study (Fitzmaurice 2000).

Power calculation

Only four studies reported a power calculation (Destache 1990;

Fitzmaurice 2000; Manotti 2001; Mungall 1994).

Follow-up of patients and professionals

All studies except three (Casner 1993; Destache 1990; Vadher

1997) reported an adequate follow-up of patients (more than 80%

of patients). But only nine studies reported an adequate follow-up

of professionals.

Assessment of primary outcome

In most studies, the assessment of primary outcome was blinded

and the measure of the primary outcome was reliable. In one

study on warfarin adjustment (Carter 1987), the therapeutic range

chosen for prothrombin time ratio was lower (1.3-2.5) than the

conventional one (1.5-2.5).

Baseline measurement

Baseline measurement was done in only four studies. In all other

studies, it was impossible to score the existence of baseline mea-

surement, since most computerized advices worked at the moment

of the prescription of a new specific drug. In one study the review-

ers thought that there was little room for improvement because

the performance of the health professional was adequate without

the intervention (White 1991).

Effects of interventions

Hypothesis 1. Giving the health professional

computer advice led to a change in drug dosage.

For this comparison, we analysed the following outcomes: the dose

administered to the patient and the number of dosage changes.



Comparisons concerning administered doses were separated into

three groups: initial dose, maintenance dose, and total amount of

drug used. Five comparisons provided outcomes for the analysis on

initial dose (Burton 1991;Gonzalez 1989; Hurley 1986; Rodman

1984; Theil 1993 fentanyl; Verner 1992). Initial doses tended

to be higher with computerized advice. (SMD 1.12, 95%CI

0.33 to 1.92). Eight comparisons provided data on maintenance

dose (Burton 1991; Carter 1987; Gonzalez 1989; Hurley 1986;

Rodman 1984; Ruiz 1993; Vadher 1997 pop 1; Vadher 1997

pop2). Overall, the pooled effect showed no difference between

both groups (SMD 0.19, 95%CI -0.10 to 0.48). Four trials re-

ported data on total administered dose (Begg 1989; Lesourd 2002;

Mungall 1994; Rodman 1984). Overall, the pooled effect showed

no difference between both groups (SMD 0.43, 95%CI -0.29 to

1.16). For the three outcomes, statistical heterogeneity was impor-

tant.

Theil 1993 fentanyl and Theil 1993 midazolam provided out-

comes for the initial, maintenance and total doses for both Fen-

tanyl and Midazolam infusions. Computerized advice had no ef-

fect on Fentanyl drug dosage but reduced significantly Midazolam

initial, maintenance and total drug doses (SMD -2.0, 95%CI -2.9

to -0.96 for initial dose; -1.21, 95%CI -2.04 to -0.3 for mainte-

nace dose; -1.10, 95%CI -1.92 to -0.21 for total dose).

Two comparisons analysed the number of drug adjustments

(Destache 1990; Theil 1993 fentanyl). The pooled effect showed

no difference between both groups (SMD 0.26, 95%CI -0.48 to

0.98). There were three eligible studies with potential unit of anal-

ysis error which analysed the proportion of dose changes (Ageno

1998; Poller 1998; Poller 1998a). For those three studies, the pro-

portion of dose changes was less important in the computer group

than in the control group (median RR 0.68). One eligible study

with potential unit of analysis error (Chertow 2001) showed that

a computerized decision support system for prescribing drugs in

patients with renal insufficiency improved the proportion of ap-

propriate orders (RR 1.71, 1.64-1,78).

In summary, in spite of high heterogeneity, the computerized ad-

vice seems to lead to a change in the initial dose of drug but to

have no effect on the maintenance dose and the total amount of

drugs used. In addition, computerized advice doesn’t reduce the

number of dosage adjustments.

Hypothesis 2. Decisions on drug dosage based on

computerized advice led more often to drug levels

within the therapeutic range.

For this comparison, the outcomes analysed were serum concen-

trations, proportion of patients within therapeutic range and pro-

portion of time spent within therapeutic range.

Eight comparisons reported serum concentrations ( Begg 1989;

Burton 1991; Casner 1993; Gonzalez 1989; Hickling 1989;

Hurley 1986; Rodman 1984; Verner 1992). Although these com-

parisons concerned three different drugs (Theophylline, Lido-

caine, Aminoglycoside), we pooled the results because we used

standardised mean differences. Overall, drug levels are higher in

the computer group than in the control group (SMD 1.12, 95%CI

0.43 to 1.82).

Theil 1993 fentanyl and Theil 1993 midazolam reported serum

concentrations for both Fentanyl and Midazolam infusions: com-

puterized advice had no effect on Fentanyl serum concentrations

but reduced significantly Midazolam serum concentrations (SMD

-1.48, 95%CI -2.33 to -0.53).

Three comparisons (Begg 1989; Destache 1990; Hickling 1989)

analysed the proportion of patients within therapeutic range. Gen-

erally, the proportion of patients with drug levels in the therapeu-

tic range were higher in the computer groups, but this failed to

reach significance (RR 1.50, 95%CI 0.79 to 2.86).

One study (Verner 1992) analysed the percentage of time spent

within therapeutic range and was in favour of the computer group

compared to the control group, although it included a small num-

ber of patients (SMD 2.85, 95%CI 1.67 to 4.02).Three compar-

isons (Vadher 1997; Vadher 1997 pop 1; Vadher 1997 pop2) re-

ported the number of days per 100 patient-days of treatment spent

in the INR therapeutic range: patients in the intervention group

tended to spend more time in the therpaeutic range but this was

not significant (combined incidence rate ratio 1.21, 95%CI, 0.98

to 1.49, inconsistency I²=0%).

In four comparisons (Burton 1991; Casner 1993; Hurley 1986;

White 1987), the risk of toxic drug level was significantly lower

in the computer group than in the control group (RR 0.45, 95%

CI 0.30 to 0.70).

In summary, the results tend to suggest that computerized advice

leads to higher serum concentrations increases the percentage of

patients within therapeutic range - although a high heterogeneity

was observed - and reduces the risk of toxic drug level.


computerized advice led more often to a physiological

parameter being maintained within the desired range

(for example, blood pressure or prothrombin time).

Four studies without unit of analysis error yielded outcomes for

this hypothesis (Manotti 2001; Ruiz 1993, White 1987, White

1991). In two comparisons (Ruiz 1993; White 1987), computer-

ized advice seems to increase the mean proportion of time spent

within therapeutic target, but this was not significant (pooled

SMD 1.62, 95%CI -0.35 to 3.59). Another comparison (White

1991) showed no evidence of difference in anticoagulant control

in patients whose dose was determined by computer, compared

to those who were treated by a nurse specialist (RR 0.87, 95%CI

0.47 to 1.61). Finally, one comparison (Manotti 2001) analysed

the proportion of patients reaching a stable state of anticoagula-

tion (three INR measurements within therapeutic range) and was

in favour of the computer group compared to the control group

(RR 1.46, 95%CI 1.07 to 2.00).

There were two eligible comparisons (Poller 1998; Poller 1998a)

assessing the proportion of time spent within therapeutic target



with potential unit of analysis error (the randomisation was by

patient while the unit of analysis was the INR measurement) which

were not included in the analysis; the median SMD for these two

studies was 0.4. Besides, the proportion of INR measurements

within therapeutic range was reported in two eligible studies (

Ageno 1998; Fitzmaurice 2000) with unit of analysis error; the

median RR for these two studies was 1.1.

Considering the small number of studies and their heterogeneity,

it is impossible to draw any conclusion concerning this hypothesis.


computer advice led to more rapid therapeutic

control, assessed by a physiological parameter

The five included comparisons concerned anticoagulant therapy.

Outcomes analysed for this comparison were: time to achieve ther-

apeutic range (first INR measurement within target), time to stable

dose (INR measurement or prothrombin time ratio maintained

within a given range during at least three days).

In two comparisons (Vadher 1997; White 1987), there was no

evidence of difference between the intervention and control groups

for time to achieve therapeutic range (SMD -0.22, 95%CI -0.69

to 0.26) but, in three comparisons (Carter 1987; Vadher 1997;

White 1987), the computer advice reduced time to stabilization

(SMD -0.55, 95%CI -1.03 to -0.08).

In addition, one comparison performed among patients receiv-

ing aminoglycosides (Destache 1990) analysed clinical parameters

for resolution of infection. The time for elevated temperature to

decrease was shorter in the computer group than in the control

group (50.0 +/- 79.4 hours versus 92.23 +/- 122.50 hours).

Although we observed that computer advice reduced significantly

time to stabilisation, it is difficult to draw any conclusion con-

cerning this comparison because of heterogeneity of these results.


computer advice led to fewer unwanted effects

For this comparison we considered two outcomes: death and ad-

verse reactions.

Six comparisons analysed death rates (Begg 1989; Burton 1991;

Destache 1990; Fitzmaurice 2000; Hurley 1986; Vadher 1997).

No difference was observed between the computer and control

groups (RR 0.81, 95%CI 0.37 to 1.81).

Ten comparisons assessed the effect of computer support on ad-

verse reactions. The assessment of the quality of evidence on ad-

verse reactions is reported in Table 1. There was a great diversity

of drugs and of type of adverse reactions. Consequently, we did

not pool the results.

Table 1. Logistics and organisational aspects of care

Characteristics Negative studies Positive studies

Setting

• Outpatient 5 (63) 3 (37)

• Inpatient 9 (53) 8 (47)

• Mixed 1 (100) 0 (0)

Advice integrated in a CPOE

• No 8 (50) 8 (50)

• Yes 3 (75) 1 (25)

• Not clear 4 (67) 2 (33)

Starter

• User-initiated 8 (80) 2 (20)

• System-initiated 2 (29) 5 (61)



Table 1. Logistics and organisational aspects of care (Continued)

• Not clear 5 (71) 4 (29)

Type of intervention

• Direct 8 (67) 4 (33)

• Indirect 2 (67) 1 (33)

• Not clear 5 (45) 6 (55)

Abbreviations

CPOE - Computer physician order entry

In addition, three comparisons analysed the impact of computer

advice on positive events: the absence of change in creatinine clear-

ance (Begg 1989, RR 1.34, 95%CI 0.61 to 2.98), the cure of

the disease after aminoglycoside therapy (Burton 1991, RR 0.95,

95%CI 0.55 to 1.64), the pregnancy rate after ovarian stimula-

tion (Lesourd 2002, RR 1.15, 95%CI 0.59 to 2.27). Individual

comparisons showed no evidence of difference between computer

and control groups. We did not pool the results because of the

diversity of outcomes.

Hypothesis 6. Computer advice reduced the use of

resources (length of stay).

Six comparisons reported the length of time spent in hospital

(Burton 1991; Casner 1993; Destache 1990; Hurley 1986; Verner

1992; White 1987). Overall they showed a significant reduction in

hospital stay duration in the computer group (SMD -0.35, 95%CI

0.52 to 0.17). There was one eligible study (Chertow 2001) with

potential unit of analysis error which showed a reduction in length

of stay during the intervention periods (SMD -0.04, 95%CI -0.07

to -0.01).

Hypothesis 7. Computerized advice given in real time

was more effective than that given by delayed

feedback.

We didn’t find any studies where the computerized advice was

delivered by delayed feedback.

Hypotheses 8 to 11.

The results addressing decision support technical features (such as

its integration into a computer physician order entry system) or

aspects of organization of care (such as the setting) are presented

in Table 1. Because the numbers of comparisons are small, these

results are only descriptive and cannot answer the tested hypothe-

sis. System-initiated decision support seems to be associated with

comparisons favourable to computerized advice.

D I S C U S S I O N

This review suggests that the computerized advice for drug dosage

has some benefits: it increases the initial dose of drug and tends to

increase serum concentrations; it leads to a more rapid therapeutic

control and it reduces the length of time spent in the hospital.

Those results were similar in the initial review. It also suggests that

computer advice significantly decreases toxic drug levels but has

no effect on adverse reactions. In the initial review, both outcomes

showed significant benefits for computerized advice. In addition,

we did not find strong evidence of any logistic and organization

of care aspects favouring the effect of computer support. Finally,

the review does not provide enough evidence to identify specific

therapeutic areas where the intervention is particularly beneficial.

In the comparisons that we identified, unaided health profession-

als tended to be cautious in estimating the amount of drug to use.

This caution presumably results from an unwillingness to expose

the patient to adverse effects of drug therapy. Unaided clinicians

tended to use lower loading, maintenance and total doses, than

when computer support was available. Lower doses lead to lower

blood levels and often to sub-optimal therapeutic effects. Although

doses with computer support tended to be higher than those used

by unaided doctors, toxic drug levels were significantly reduced.

This suggests that the computers helped doctors to tailor the dose



of the drug more accurately to the individual patient. Higher doses

with computer support may lead to more rapid therapeutic con-

trol, bringing benefits for patients and reducing the time spent in

hospital.

However, these findings need to be read with caution.

First, from 1996 to 2006, we identified only nine comparisons

since the first publication of this review, concerning a small num-

ber of drugs. Half of the comparisons concerned anticoagulants

and three comparisons concerned theophylline, a drug which is

not considered as the first choice treatment of asthma at present.

However, monitoring serum concentration levels of theophylline

is essential to ensure that non-toxic doses are achieved (National

Asthma 2002). We didn’t find any study concerning some new

drugs for which it is considered important to monitor drug levels

such as glycopeptides, antifungal (fluconazole) and antiretroviral

drugs.

Second, the quality of studies was generally low. The unit of allo-

cation and the unit of analysis of a well-designed study would be

the community, institution or practice. Here, the unit of allocation

was the patient in all comparisons; we excluded four comparisons

with practice as unit of allocation because of unit of analysis er-

ror . Only seven comparisons reported an adequate concealment

of allocation. Protection against contamination was considered to

be done in only one study. In most comparisons sample size was

small. The decision support logistics of interventions were fully

described in only seven comparisons.

Third, the heterogeneity accross individual comparisons was high

for most outcomes. We had to deal with different outcomes, a

great diversity of clinical contexts and organization of care. Pool-

ing those outcomes was sometimes impossible. The number of

extracted outcomes varied greatly from one study to another and

consequently the set of included comparisons varied greatly ac-

cross outcomes. Furthermore, for a given hypothesis, different out-

comes could be correlated. This makes the interpretation of re-

sults sometimes difficult: for instance, 1) computer advice did not

increase the total dose while it had an effect on the initial dose:

this might be explained by the fact that only two comparisons

contribute to the two outcomes 2) concerning hypothesis 02, two

comparisons concerning aminoglycosides appeared in two out-

comes, serum concentrations and percentage of patients within

therapeutic range.

Fourth, for some indicators (length of stay, mortality), crude results

can be affected by unknown confounding factors.

This review suggests that computer support leads to safer and more

accurate determination of drug dosage. Since physicians more and

more use computers for prescribing (including CPOE in hospitals)

the opportunity exists to make comprehensive support for drug

dosage widely available.

Finally, we had to deal with different outcomes, due to the great di-

versity of clinical contexts, and pooling those outcomes was some-

times impossible. In five studies it was even impossible to extract

outcomes, for which full data were not available.

A U T H O R S ’ C O N C L U S I O N SImplications for practice

1. Analysis of trials suggests that computerized advice for drug

dosage has some benefits. It increased the initial dose of drug and

led to a more rapid therapeutic control. It also reduced the

length of time spent in the hospital. However, it did not reduce

adverse reactions.

2. Those results are based on studies mainly of low quality,

concerning various clinical specialties and a small number of

drugs. Drugs studied so far have a narrow therapeutic range;

benefits may be confined to this type of drug.

3. No definitive conclusion could be drawn concerning the

logistics of the computerized support and organization of care

aspects. It is not certain that these benefits could be achieved

with different computer systems in different clinical situations.

Implications for research1. More studies are needed to demonstrate that the use of

computers improves the quality of care. Well-designed trials

randomized by clusters are mandatory for assessment of the

effect of computerized support systems on drug dosage.

2. These studies should address the identification of the

factors that predict a successful and acceptable system: the

decision support logistics, the organization of care and the

healthcare professionals’ characteristics.

3. Studies evaluating other drugs with a narrow therapeutic

window or complicated pharmacokinetics (e.g. antibiotics) are

needed.

These studies should address the identification of the factors that

predict a successful and acceptable system: the decision support

logistics, the organization of care and the healthcare professionals’

characteristics.

Studies evaluating other drugs with a narrow therapeutic window

or complicated pharmacokinetics (e.g. antibiotics) are needed.

A C K N O W L E D G E M E N T S

The authors thank Sophie Guiquerro for literature seach support,

Elske Ammenwerth for her review of an article in German, and

Lucienne Boujon who provided English corrections. We would

also like to thank Jessie McGowan and Doug Salzwedel for their

assistance with searching and Alan Forster and Jim Wright for their

helpful comments on earlier versions of this review.



R E F E R E N C E S

References to studies included in this review

Ageno 1998 {published data only}

Ageno W, Turpie AG. A randomized comparison of a computer-

based dosing program with a manual system to monitor oral

anticoagulant therapy. Thrombosis Research 1998;91(5):237–40.

Begg 1989 {published data only}

Begg EJ, Atkinson HC, Jeffery GM, Taylor NW. Individualised

aminoglycoside dosage based on pharmacokinetic analysis is

superior to dosage based on physician intuition at achieving target

plasma drug concentrations. British Journal of Clinical

Pharmacology 1989;28(2):137–41.

Burton 1991 {published data only}

Burton ME, Ash CL, Hill DP Jr, Handy T, Shepherd MD, Vasko

MR. A controlled trial of the cost benefit of computerized bayesian

aminoglycoside administration. Clinical Pharmacology and

Therapeutics 1991;49(6):685–94.

Carter 1987 {published data only}

Carter BL, Taylor JW, Becker A. Evaluation of three dosage-

prediction methods for initial in-hospital stabilization of warfarin

therapy. Clinical Pharmacy 1987;6(1):37–45.

Casner 1993 {published data only}

Casner PR, Reilly R, Ho H. A randomized controlled trial of

computerized pharmacokinetic theophylline dosing versus empiric

physician dosing. Clinical Pharmacology and Therapeutics 1993;53

(6):684–90.

Chertow 2001 {published data only}

Chertow GM, Lee J, Kuperman GJ, Burdick E, Horsky J, Seger

DL, et al.Guided medication dosing for inpatients with renal

insufficiency. JAMA 2001;286(22):2839–44.

Destache 1990 {published data only}

Destache CJ, Meyer SK, Bittner MJ, Hermann KG. Impact of a

clinical pharmacokinetic service on patients treated with

aminoglycosides: a cost-benefit analysis. Therapetic Drug

Monitoring 1990;12(5):419–26.

Fitzmaurice 2000 {published data only}

Fitzmaurice DA, Hobbs FD, Murray ET, Holder RL, Allan TF,

Rose PE. Oral anticoagulation management in primary care with

the use of computerized decision support and near-patient testing:

a randomized, controlled trial. Archives of Internal Medicine 2000;

160(15):2343–8.

Gonzalez 1989 {published data only}

Gonzalez ER, Vanderheyden BA, Ornato JP, Comstock TG.

Computer-assisted optimization of aminophylline therapy in the

emergency department. American Journal of Emergency Medicine

1989;7(4):395–401.

Hickling 1989 {published data only}

Hickling K, Begg E, Moore ML. A prospective randomised trial

comparing individualised pharmacokinetic dosage prediction for

aminoglycosides with prediction based on estimated creatinine

clearance in critically ill patients. Intensive Care Medicine 1989;15

(4):233–7.

Hurley 1986 {published data only}

Hurley SF, Dziukas LJ, McNeil JJ, Brignell MJ. A randomized

controlled clinical trial of pharmacokinetic theophylline dosing.

American Review of Respiratory Disease 1986;134(6):1219–24.

Lesourd 2002 {published data only}

Lesourd F, Avril C, Boujennah A, Parinaud J. A computerized

decision support system for ovarian stimulation by gonadotropins.

Fertility and Sterility 2002;77(3):456–60.

Manotti 2001 {published data only}

Manotti C, Moia M, Palareti G, Pengo V, Ria L, Dettori AG. Effect

of computer-aided management on the quality of treatment in

anticoagulated patients : a prospective, randomized, multicenter

trial of APROAT (Automated Program for Oral Anticoagulant

Treatment). Haematologica 2001;86(10):1060–70.

Mungall 1994 {published data only}

Mungall DR, Anbe D, Forrester PL, Luoma T, Genovese R, Mahan

J, et al.A prospective randomized comparison of the accuracy of

computer-assisted versus GUSTO nomogram--directed heparin

therapy. Clinical Pharmacology and Therapeutics 1994;55(5):591–6.

Poller 1998 {published data only}

Poller L, Shiach CR, MacCallum PK, Johansen AM, Munster AM,

Magalhaes A, et al.Multicentre randomised study of computerised

anticoagulant dosage. European Concerted Action on

Anticoagulation. Lancet 1998;352(9139):1505–9.

Poller 1998a {published data only}

Poller L, Shiach CR, MacCallum PK, Johansen AM, Munster AM,

Magalhaes A, et al.Multicentre randomised study of computerised

anticoagulant dosage. European Concerted Action on

Anticoagulation. Lancet 1998;352(9139):1505–9.

Rodman 1984 {published data only}

Rodman JH, Jelliffe RW, Kolb E, Tuey DB, de Guzman MF, Wagers

PW, et al.Clinical studies with computer-assisted initial lidocaine

therapy. Archives of Internal Medicine 1984;144(4):703–9.

Ruiz 1993 {published data only}

Ruiz R, Borches D, Gonzalez A, Corral J. A new sodium-

nitroprusside-infusion controller for the regulation of arterial blood

pressure. Biomedical Instrumentation & Technology 1993;27(3):

244–51.

Theil 1993 fentanyl {published data only}

Theil DR, Stanley TE 3rd, White WD, Goodman DK, Glass PS,

Bai SA, et al.Midazolam and fentanyl continuous infusion

anesthesia for cardiac surgery: a comparison of computer-assisted

versus manual infusion systems. Journal of Cardiothoracic and

Vascular Anesthesia 1993;7(3):300–6.

Theil 1993 midazolam {published data only}

Theil DR, Stanley TE 3rd, White WD, Goodman DK, Glass PS,

Bai SA, et al.Midazolam and fentanyl continuous infusion



anesthesia for cardiac surgery: a comparison of computer-assisted

versus manual infusion systems. Journal of Cardiothoracic and

Vascular Anesthesia 1993;7(3):300–6.

Vadher 1997 {published data only}

Vadher B, Patterson DL, Leaning M. Evaluation of a decision

support system for initiation and control of oral anticoagulation in

a randomised trial. BMJ 1997;314(7089):1252–6.

Vadher 1997 pop 1 {published data only}

Vadher BD, Patterson DL, Leaning M. Comparison of oral

anticoagulant control by a nurse-practitioner using a computer

decision-support system with that by clinicians. Clinical and

Laboratory Haematology 1997;19(3):203–7.

Vadher 1997 pop2 {published data only}

Vadher BD, Patterson DL, Leaning M. Comparison of oral

anticoagulant control by a nurse-practitioner using a computer

decision-support system with that by clinicians. Clinical and

Laboratory Haematology 1997;19(3):203–7.

Verner 1992 {published data only}

Verner D, Seligmann H, Platt S, Dany S, Almog S, Zulty L, et

al.Computer assisted design of a theophylline dosing regimen in

acute bronchospasm: serum concentrations and clinical outcome.

European Journal of Clinical Pharmacology 1992;43(1):29–33.

White 1987 {published data only}

White RH, Hong R, Venook AP, Daschbach MM, Murray W,

Mungall DR, et al.Initiation of warfarin therapy: comparison of

physician dosing with computer-assisted dosing. Journal of General

Internal Medicine 1987;2(3):141–8.


White RH, Mungall D. Outpatient management of warfarin

therapy: comparison of computer-predicted dosage adjustment to

skilled professional care. Therapeutic Drug Monitoring 1991;13(1):

46–50.

References to studies excluded from this review

Abbrecht 1982 {published data only}

Abbrecht PH, O’Leary TJ, Behrendt DM. Evaluation of a

computer-assisted method for individualized anticoagulation:

retrospective and prospective studies with a pharmacodynamic

model. Clinical Pharmacology and Therapeutics 1982;32(1):129–36.

Alvis 1985 {published data only}

Alvis JM, Reves JG, Govier AV, Menkhaus PG, Henling CE, Spain

JA, et al.Computer-assisted continuous infusions of fentanyl during

cardiac anesthesia: comparison with a manual method.

Anesthesiology 1985;63(1):41–9.

Bury 2005 {published data only}

Bury J, Hurt C, Roy A, Cheesman L, Bradburn M, Cross S, et

al.LISA: a web-based decision-support system for trial management

of childhood acute lymphoblastic leukaemia. British Journal of

Haematology 2005;129(6):746–54.

Carter 1987 linear {published data only}

Carter BL, Taylor JW, Becker A. Evaluation of three dosage-

prediction methods for initial in-hospital stabilization of warfarin

therapy. Clinical Pharmacy 1987;6(1):37–45.

Chiarelli 1990 {published data only}

Chiarelli F, Tumini S, Morgese G, Albisser AM. Controlled study in

diabetic children comparing insulin-dosage adjustment by manual

and computer algorithms. Diabetes Care 1990;13(10):1080–4.

Collins 2004 {published data only}

Collins CD, Pedersen CA, Schneider PJ, Miller AS, Sierawski SJ,

Roux RK. Effect on amphotericin B lipid complex use of a clinical

decision support system for computerized prescriber order entry.

American Journal of Health-System Pharmacy 2004;61(13):1395–9.

Fihn 1994 {published data only}

Fihn SD, McDonell MB, Vermes D, Henikoff JG, Martin DC,

Callahan CM, et al.A computerized intervention to improve timing

of outpatient follow-up: a multicenter randomized trial in patients

treated with warfarin. National Consortium of Anticoagulation

Clinics. Journal of General Internal Medicine 1994;9(3):131–9.


Fitzmaurice DA, Hobbs FD, Murray ET, Bradley CP, Holder R.

Evaluation of computerized decision support for oral

anticoagulation management based in primary care. British Journal

of General Practice 1996;46(410):533–5.


Fitzmaurice DA, Hobbs FD, Murray ET. Primary care

anticoagulant clinic management using computerized decision

support and near patient International Normalized Ratio (INR)

testing: routine data from a practice nurse-led clinic. Family

Practice 1998;15(2):144–6.

Hobbs 1996 {published data only}

Hobbs FD, Delaney BC, Carson A, Kenkre JE. A prospective

controlled trial of computerized decision support for lipid

management in primary care. Family Practice 1996;13(2):133–7.

Horn 2002 {published data only}

Horn W, Popow C, Miksch S, Kirchner L, Seyfang A. Development

and evaluation of VIE-PNN, a knowledge-based system for

calculating the parenteral nutrition of newborn infants. Artificial

Intelligence in Medicine 2002;24(3):217–28.

Hwang 2004 {published data only}

Hwang HG, Chang IC, Hung WF, Sung ML, Yen D. The design

and evaluation of clinical decision support systems in the area of

pharmacokinetics. Medical Informatics and the Internet in Medicine

2004;29(3-4):239–51.

Kroese 2005 {published data only}

Kroese WL, Avery AJ, Savelyich BS, Brown NS, Schers H, Howard

R, et al.Assessing the accuracy of a computerized decision support

system for digoxin dosing in primary care: an observational study.

Journal of Clinical Pharmacy and Therapeutics 2005;30(3):279–83.

Manotti 2001 mainten {published data only}

Manotti C, Moia M, Palareti G, Pengo V, Ria L, Dettori AG. Effect

of computer-aided management on the quality of treatment in

anticoagulated patients: a prospective, randomized, multicenter

trial of APROAT (Automated PRogram for Oral Anticoagulant

Treatment). Haematologica 2001;86(10):1060–70.

McDonald 1980 {published data only}

McDonald CJ, Wilson GA, McCabe GP, Jr. Physician response to

computer reminders. JAMA 1980;244(14):1579–81.



McMichael 1993 {published data only}

McMichael J, Lieberman R, Doyle H, McCauley J, Fung J, Starzl

TE. An intelligent and cost-effective computer dosing system for

individualizing FK506 therapy in transplantation and autoimmune

disorders. Journal of Clinical Pharmacology 1993;33(7):599–605.

Murchie 1989 {published data only}

Murchie CJ, Kenny GN. Comparison among manual, computer-

assisted, and closed-loop control of blood pressure after cardiac

surgery. Journal of Cardiothoracic Anesthesia 1989;3(1):16–19.

Nieuwenhuyze 1995 {published data only}

van den Nieuwenhuyzen MC, Engbers FH, Burm AG, Vletter AA,

van Kleef JW, Bovill JG. Computer-controlled infusion of

alfentanil versus patient-controlled administration of morphine for

postoperative analgesia: a double-blind randomized trial. Anesthesia

and Analgesia 1995;81(4):671–9.

Nightingale 2000 {published data only}

Nightingale PG, Adu D, Richards NT, Peters M. Implementation

of rules based computerised bedside prescribing and

administration: intervention study. BMJ 2000;320(7237):750–3.

Peck 1973 {published data only}

Peck CC, Sheiner LB, Martin CM, Combs DT, Melmon KL.

Computer-assisted digoxin therapy. New England Journal of

Medicine 1973;289(9):441–6.

Peters 1996 {published data only}

Peters A, Kerner W. Analytical design and clinical application of an

intelligent control system for insulin treatment) (Analytical design

and clinical application of an intelligent control system for insulin

treatment [Analytisches Design und klinische Anwendung eines

lernfähigen Regelsystems für die Pharmakotherapie mit Insulin].

Biomedizinische Technik 1996;41(1-2):2–13.

Peterson 1986 {published data only}

Peterson CM, Jovanovic L, Chanoch LH. Randomized trial of

computer-assisted insulin delivery in patients with type I diabetes

beginning pump therapy. American Journal of Medicine 1986;81(1):

69–72.

Peterson 2005 {published data only}

Peterson JF, Kuperman GJ, Shek C, Patel M, Avorn J, Bates DW.

Guided prescription of psychotropic medications for geriatric

inpatients. Archives of Internal Medicine 2005;165(7):802–7.

Poller 1993 {published data only}

Poller L, Wright D, Rowlands M. Prospective comparative study of

computer programs used for management of warfarin. Journal of

Clinical Pathology 1993;46(4):299–303.

Rood 2005 {published data only}

Rood E, Bosman RJ, van der Spoel JI, Taylor P, Zandstra DF. Use

of a computerized guideline for glucose regulation in the intensive

care unit improved both guideline adherence and glucose

regulation. Journal of the American Medical Informatics Association

2005;12(2):172–80.

Rotman 1996 {published data only}

Rotman BL, Sullivan AN, McDonald TW, Brown BW, DeSmedt P,

Goodnature D, et al.A randomized controlled trial of a computer-

based physician workstation in an outpatient setting:

implementation barriers to outcome evaluation. Journal of the

American Medical Informatics Association 1996;3(5):340–8.

Ryff-de Leche 1992 {published data only}

Ryff-de Leche A, Engler H, Nutzi E, Berger M, Berger W. Clinical

application of two computerized diabetes management systems:

comparison with the log-book method. Diabetes Research 1992;19

(3):97–105.

Strack 1985 {published data only}

Strack T, Bergeler J, Beyer J, Hutten H. Computer assisted

conventional insulin therapy. Life Support Systems 1985;3 Suppl 1:

568–72.

Tamblyn 2003 {published data only}

Tamblyn R, Huang A, Perreault R, Jacques A, Roy D, Hanley J, et

al.The medical office of the 21st century (MOXXI): effectiveness of

computerized decision-making support in reducing inappropriate

prescribing in primary care. CMAJ 2003 Sep 16;169(6):549–56.


White KS, Lindsay A, Pryor TA, Brown WF, Walsh K. Application

of a computerized medical decision-making process to the problem

of digoxin intoxication. Journal of the American College of

Cardiology 1984;4(3):571–6.

Willcourt 1994 {published data only}

Willcourt RJ, Pager D, Wendel J, Hale RW. Induction of labor with

pulsatile oxytocin by a computer-controlled pump. American

Journal of Obstetrics and Gynecology 1994;170(2):603–8.

References to studies awaiting assessment

Evans 1998 {published data only}

Evans WE, Relling MV, Rodman JH, Crom WR, Boyett JM, Pui

CH. Conventional compared with individualized chemotherapy for

childhood acute lymphoblastic leukemia. New England Journal of

Medicine 1998;338:499–505.

Fernández de Gatta MD {published data only}

Fernández de Gatta MD, Calvo MV, Hernández JM, Caballero D,

San Miguel JF, Domínguez-Gil. Cost-effectiveness analysis of

serum vancomycin concentration monitoring in patients with

hematologic malignancies. Clinical Pharmacology & Therapeutics

1996;60:332–40.

Le Meur 2007 {published data only}

Le Meur Y, Büchler M, Thierry A, Caillard S, Villemain F, Lavaud

S, et al.Individualized mycophenolate mofetil dosing based on drug

exposure significantly improves patient outcomes after renal

transplantation. American Journal of Transplantation 2007;7:

2496–503.

Mihahlovic 2003 {published data only}

Mihahlovic GS, Milovanovic DR, Jankovic SM. Comparison of

efficacy and safety between individualized and empiric dose

regimen of amitriptyline in the treatment of major depressive

episode. Psychiatry and Clinical Neurosciences 2003;57:580–5.

van Lent-Evers 1999 {published data only}

van Lent-Evers NA, Mathôt RA, Geus WP, van Hout BA, Vinks

AA. Impact of goal-oriented and model-based clinical

pharmacokinetic dosing of aminoglycosides on clinical outcome: a

cost-effectiveness analysis. Journal of Therapeutic Drug Monitoring

1999;21(1):63–73.

Additional references



Baldwin 1995

Baldwin L. Calculating drug doses. British Medical Journal 1995;

310(6988):1154.

Durieux 2005

Durieux P. Electronic medical alerts--so simple, so complex. New

England Journal of Medicine 2005;352(10):1034–6.

Garcia 2002

Garcia JM, Martin JLR, Subirana M, Gich I. Self management for

oral anticoagulation. Cochrane Database of Systematic Reviews 2002,

Issue 4. [DOI: 10.1002/14651858.CD003839]

Garg 2005

Garg AX, Adhikari NK, McDonald H, Rosas-Arellano MP,

Devereaux PJ, Beyene J, Sam J, Haynes RB. Effects of

computerized clinical decision support systems on practitioner

performance and patient outcomes: a systematic review. Journal of

the American Medical Association 2005;293(10):1223–38.

Kaushal 2006

Kaushal R, Jha AK, Franz C, Glaser J, Shetty KD, Jaggi T,

Middleton B, Kuperman GJ, Khorasani R, Tanasijevic M, Bates

DW, Brigham and Women’s Hospital CPOE Working Group.

Return on investment for a computerized physician order entry

system. Journal of the American Medical Informatics Association

2006;13(3):261–3.

Kawamoto 2005

Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving

clinical practice using clinical decision support systems: a

systematic review of trials to identify features critical to success.

British medical journal 2005;330(7494):765.

Kuperman 2003

Kuperman GJ, Gibson RF. Computer physician order entry:

benefits, costs, and issues. Annals of internal medicine 2003;139(1):

31–9.

McDonald 1976

McDonald CJ. Protocol-based computer reminders, the quality of

care and the non-perfectability of man. New England Journal of

Medicine 1976;295(24):1351–5.

National Asthma 2002

National Asthma Education and Prevention Program. Expert Panel

Report: guidelines for the diagnosis and management of asthma

update on selected topics-2002. Journal of Allergy and Clinical

Immunology 2002;110:S141–219.

Nies 2006

Nies J, Colombet I, Degoulet P, Durieux P. Determinants of

Success for Computerized Clinical Decision Support Systems

Integrated in CPOE Systems: a Systematic Review. AMIA Annual

Symposium Proceedings. 2006:594–8.

Rolfe 1995

Rolfe S, Harper NJ. Ability of hospital doctors to calculate drug

doses. British Medical Journal 1995;310(6988):1173–4.

Walton 2001

Walton RT, Harvey E, Dovey S, Freemantle N. Computerised

advice on drug dosage to improve prescribing practice. Cochrane

Database of Systematic Reviews 2001, Issue 1. [DOI: 10.1002/

14651858.CD002894]∗ Indicates the major publication for the study



C H A R A C T E R I S T I C S O F S T U D I E S

Characteristics of included studies [ordered by study ID]

Ageno 1998

Methods Design: RCT

Unit of allocation: Patient

Unit of analysis: Episode of care

Power calculation: NC

Concealment of allocation: NC

Follow-up of professionals: NC

Follow-up of patients: Done

Blinded assessment of primary outcome: Done

Blinded measurement of primary outcome: NC

Reliable outcome: Done

Protection against contamination: Not done

Participants Profession: Mixed (Physicians+nurses)

Level of training: Accredited/licensed

Clinical specialty: Other, anticoagulant clinic

Country: Canada

Patients: 101 outpatients on long-term oral anticoagulant therapy after mechanical heart valve replacement

Interventions Prediction rules, computer-assisted group (n=50) vs control group (n=51)

Location of care: outpatient

Clinical problem: long-term warfarin therapy

Computer advice: Given in real time

CDSS integration in CPOE: No

Starter: User-initiated

Type of intervention: Direct intervention

Outcomes proportion of doses adjustments, proportion of INR measurements w/in therapeutic range

Notes

Risk of bias

Item Authors’ judgement Description

Allocation concealment? Unclear B - Unclear



Begg 1989

Methods Design: RCT


Unit of analysis: Patient


Concealment of allocation: Done



Blinded assessment of primary outcome: NC

Blinded measurement of primary outcome: Not done



Participants Profession: Physicians

Level of training: NC

Clinical specialty: NC

Country: New Zealand

Patients: 50 hospital inpatients (intensive care unit excluded)

Interventions Pharmacokinetic model, computer-assisted group (n= 24) vs control group (n=26)

Location of care: Inpatient care

Clinical problem: aminoglycoside


CDSS integration in CPOE NC


Type of intervention: NC

Outcomes total dose, serum drug concentration, percentage of patients w/in drug therapeutic range, deaths, positive

events

Notes

Risk of bias


Allocation concealment? Yes A - Adequate



Burton 1991

Methods Design: CCT



Power calculation :NC


Follow-up of professionals: Done



Blinded measurement of primary outcome: Done


Protection against contamination: NC




Country: United States of America

Patients: 147 patients treated with aminoglycosides

Interventions Dose advice based on Bayesian pharmacokinetic model (n=72) vs usual care (n=75)





Starter: NC


Outcomes Initial dose, maintenance dose, proportion of patients with toxic drug levels, serum drug concentration,

deaths, adverse reactions, length of stay, positive events

Notes

Risk of bias





Carter 1987

Methods Design: RCT











Participants Profession: Mixed (physicians+pharmacists)




Patients: 65 adult inpatients receiving warfarin sodium

Interventions Pharmacokinetic concepts, analog-computer program (n=31) vs empiric dosing (n=34)


Clinical problem: initiation of warfarin therapy



Starter: NC

Type of intervention: Indirect intervention

Outcomes maintenance dose, time to stabilization

Notes

Casner 1993

Methods Design: RCT



Power calculation: Not done



Follow-up of patients: Not done









Patients: 35 patients with diagnoses of asthma or obstructive pulmonary disease



Casner 1993 (Continued)

Interventions Suggestion based on linear one compartment model (n=17) vs usual care (n=18)


Clinical problem: theophylline maintenance for asthma


CDSS integration in CPOE: Yes



Outcomes proportion of patients with toxic drug levels, serum drug concentration, length of stay

Notes

Risk of bias



Chertow 2001

Methods Design: CT (alternating time series design with four consecutive 2-month period)




Concealment of allocation Not done









Clinical specialty: Other mixed


Patients: 17828 inpatients with renal insufficiency

Interventions CDSS periods (n=7887 patients) vs control periods (n=9941 patients)


Clinical problem: Renal insufficiency



Starter: System-initiated


Outcomes proportion of appropriate orders, length of stay



Chertow 2001 (Continued)

Notes

Risk of bias


Allocation concealment? No C - Inadequate

Destache 1990

Methods Design: RCT



Power calculation: Done








Participants Profession: Mixed physicians+clinical pharmacists)

Level of training: Mixed

Clinical specialty: Internal medicine, Surgery ICU


Patients: 145 patients treated with aminoglycosides for infection.

Interventions Patients whose doctors accepted recommendations based on a one compartment Bayesian pharmacoki-

netic

model (n=75) vs those of doctors who did not (n=70).





Starter: NC


Outcomes number of doses adjustments, proportion of patients with signs of toxic drug levels, proportion of patients

w/in drug therapeutic range, deaths, adverse reactions, length of stay

Notes

Risk of bias




Destache 1990 (Continued)


Fitzmaurice 2000

Methods Design: RCT










Protection against contamination: Done

Participants Profession: Mixed (Physicians+Nurses)


Clinical specialty: General/family practice

Country: England

Patients: 224 outpatients with cardiovascular disease

Interventions CDSS group (n=122) vs routine care (n=102)

Location of care: Community based care

Clinical problem: Warfarin adjustment for long-term therapy


CDSS integration in CPOE: NC



Outcomes proportion of INR measurements w/in therapeutic range, deaths, adverse reactions

Notes There was 2 levels of randomization. Practices were randomly tagged as intervention or control practices.

Then, in intervention practices, patients were individually randomized to intervention or control. We

didn’t analyzed ’control practices’ because of a potential unit of analysis error.

Risk of bias





Gonzalez 1989

Methods Design: RCT















Patients: 82 patients with asthma treated with aminophylline

Interventions Bayesian one compartment pharmacokinetic model (n=37) vs population based guidelines (n=30)


Clinical problem: theophylline





Outcomes Initial dose, maintenance dose, serum drug concentration, adverse reactions

Notes

Risk of bias





Hickling 1989

Methods Design: RCT











Participants Profession: NC


Clinical specialty: Other (Intensive care)

Country: New Zealand

Patients: 32 ICU patients who required aminoglycoside therapy for serious life threatening infections

Interventions Pharmacokinetic model, computer-assisted group (n=15) vs control group (n=17)


Clinical problem: aminoglycoside (gentamicin or tobramycin)



Starter: NC


Outcomes serum drug concentration, proportion of patients w/in drug therapeutic range

Notes

Risk of bias





Hurley 1986

Methods Design: RCT













Clinical specialty: Other (Emergency)

Country: Australia

Patients: 91 patients admitted to hospital with asthma

Interventions Doctors given estimate of theophylline clearance based on one compartment linear pharmacokinetic

model

(n=48) vs usual care based on theophylline levels (n=43). Computer gave advice on dose each day based

on

estimates of theophylline clearance.


Clinical problem: theophylline





Outcomes Initial dose, maintenance dose, serum drug concentration, deaths, adverse reactions, length of stay

Notes

Risk of bias





Lesourd 2002

Methods Design: RCT









Reliable outcome: NC




Clinical specialty: Obstetrics and gynaecology

Country: France

Patients: 164 women undergoing ovarian stimulation to treat infertility

Interventions CDSS group (n=82) vs control group (n=82)

Location of care: Outpatient care

Clinical problem: ovarian stimulation by gonadotropins



Starter: NC


Outcomes total dose, adverse reactions, positive events

Notes

Risk of bias





Manotti 2001

Methods Design: CCT














Country: Italy

Patients: 335 patients on oral anticoagulants

Interventions Computer-aided dosing (n=145) vs manual dosing (n=190)


Clinical problem: initiation of oral anticoagulant therapy (warfarin and acenocoumarol)





Outcomes Proportion of stable patients

Notes

Risk of bias





Mungall 1994

Methods Design: RCT




Concealment of allocation NC







Participants Profession: Mixed (Physicians + pharmacists)


Clinical specialty: Other (Coronary care unit)


Patients: 51 patients needing anticoagulation with heparin after myocardial infarction

Interventions Bayesian computer generated starting doses (n=25) vs doctors using nomogram (n=26)


Clinical problem: heparin adjustment



Starter: NC


Outcomes total dose, adverse reactions

Notes

Risk of bias





Poller 1998

Methods Design: RCT



Power calculaton: NC










Clinical specialty: Not applicable

Country: Europe (5 centres)

Patients: 79 inpatients needing anticoagulant therapy

Interventions Computer-generated-dose group (n=39) or traditional-dose group (n=40)


Clinical problem: Warfarin therapy maintenance

Computer advice: NC


Starter: NC


Outcomes proportion of doses adjustments, proportion of time spent w/in target (INR)

Notes

Risk of bias





Poller 1998a

Methods Design: RCT













Clinical specialty: Not applicable

Country: Europe (5 centres)

Patients: 175 outpatients needing anticoagulant therapy

Interventions Computer-generated-dose group (n=83) or traditional-dose group (n=92)


Clinical problem: Warfarin therapy stabilisation

Computer advice: NC


Starter: NC


Outcomes proportion of doses adjustments, proportion of time spent w/in target (INR)

Notes

Risk of bias





Rodman 1984

Methods Design: RCT



Power calculation Not done






Reliable outcome: NC


Participants Profession: NC




Patients: 20 patients admitted to medical ICU or coronary care unit needing lignocaine therapy

Interventions Advice on initial therapy using individualised linear two compartment pharmacokinetic model (n=9) vs

usual

care (n=11)


Clinical problem: lidocaine therapy





Outcomes initial dose, maintenance dose, total dose, serum drug concentration, adverse reactions

Notes

Risk of bias





Ruiz 1993

Methods Design: RCT











Participants Profession: Nurses


Clinical specialty: cardiac surgery

Country: Spain

Patients: 60 patients needing post-operative control of blood pressure with sodium nitroprusside.

Interventions Fuzzy logic controlled pump with arterial pressure sensor (n=40) vs usual care (n=20). Computer calculated

the appropriate dose and gave the drug using a pump. Clinical staff could adjust dose if necessary.


Clinical problem: Sodium nitroprusside infusion rate


CDSS integration in CPOE: NC Starter: System-initiated


Outcomes maintenance dose, proportion of time spent w/in target (mean arterial pressure )

Notes

Risk of bias





Theil 1993 fentanyl

Methods Design: RCT








Baseline measurement of primary outcome: Not done





Clinical specialty: Other (Anesthesia)


Patients: 24 patients undergoing cardiac surgery with continuous infusion of intra-venous anaesthetics.

Interventions Computer controlled pump using pharmacokinetic model to achieve target serum level (n=12) vs infusion

controlled by doctor (n=12).


Clinical problem: Fentanyl





Outcomes Initial dose, maintenance dose, total dose, number of doses adjustments, serum drug concentration

Notes

Risk of bias





Theil 1993 midazolam

Methods Design: RCT







Blinded assessment: Done

Baseline measurement: Done

Reliable outcomes: Done

Protection against contamination: Done



Clinical specialty: Other (Anesthesia)


Patients: 24 patients undergoing cardiac surgery with continuous infusion of intra-

venous anaesthetics.

Interventions Computer controlled pump using pharmacokinetic model to achieve target serum level (n=12) vs infusion

controlled by doctor (n=12)


Clinical problem: Fentanyl





Outcomes serum drug concentration

Notes

Risk of bias





Vadher 1997

Methods Design: RCT











Participants Profession: Mixed (Physicians+Nurses)

Level of training: In training


Country: United Kingdom

Patients: 148 inpatients requiring start of warfarin therapy

Interventions CDSS group (n=72 ) vs control group (n=76)

Location of care: Mixed

Clinical problem: Warfarin therapy initiation

Computer advice: NC


Starter: NC


Outcomes time to reach therapeutic range, time to stabilization, deaths, adverse reactions

Notes

Risk of bias





Vadher 1997 pop 1

Methods Design: RCT











Participants Profession: Mixed (Physicians + Nurses)


Clinical specialty: Other (Cardiology)


Patients: patients requiring anticoagulation for DVT, pulmonary embolu, atrial fibrillation



Clinical problem: Warfarin long term therapy (therapeutic range 2-3)





Outcomes maintenance dose, adverse reactions

Notes

Risk of bias





Vadher 1997 pop2

Methods Design: RCT











Participants Profession: Mixed (Physicians + Nurses)


Clinical specialty: Other (Cardiology)


Patients: patients reuiqring anticoagulation for heart valve disease, valve replacement or recurrrent throm-

boembolism



Clinical problem: Warfarin long term (therapeutic range 3-4,5)





Outcomes maintenance dose, adverse reactions

Notes

Risk of bias





Verner 1992

Methods Design: RCT




Concealment of allocation: Not done









Clinical specialty: Internal medicine

Country: Israel

Patients: 25 patients needing aminophylline therapy for acute asthma

Interventions Computer suggested dose based on individualised pharmacokinetic model to doctor (n=10) vs usual care

(n=15)


Clinical problem: Theophylline

Computer advice: NC




Outcomes Initial dose, serum drug concentration, proportion of time spent w/in drug therapeutic range, length of

stay

Notes

Risk of bias


Allocation concealment? No C - Inadequate



White 1987

Methods Design: RCT







Blinded assessment of primary outcome: Not done


Reliable outcome:


Participants Profession: Nurses


Clinical specialty: Other (anticoagulant clinic)


Patients: 75 patients requiring anticoagulation with warfarin

Interventions Initial dose suggested by Bayesian computer pharmacokinetic and pharmacodynamic model (n=39) vs

usual care (n=36)


Clinical problem: Warfarin initiation





Outcomes Proportion of time spent w/in target (prothrombin ratio), time to reach therapeutic range, time to stabi-

lization, length of stay

Notes

Risk of bias





White 1991

Methods Design: RCT











Participants Profession: Mixed (Physicians + Pharmacists)




Patients: 50 patients needing anticoagulation with warfarin (long-term oral therapy)

Interventions Maintenance dose suggested by Bayesian computer pharmacokinetic model (n=24) vs usual care (n=26)


Clinical problem: Long term warfarin adjustment





Outcomes Proportion of patients w/in target (final prothrombin time)

Notes

Risk of bias



Abbreviations

CDSS - Computer decision support system

CPOE - Computer physician order entry

ICU - Intensive Care Unit

RCT - Randomized controlled trial



Characteristics of excluded studies [ordered by study ID]

Study Reason for exclusion

Abbrecht 1982 - Computer controlled pump not under physician control

Alvis 1985 - Design

Bury 2005 - Not computerized drug dosage

Carter 1987 linear - Not computerized drug dosage

Chiarelli 1990 - Patient aid not under physician control

Collins 2004 - Design

Fihn 1994 - Outcomes did not fit inclusion criteria

Fitzmaurice 1996 - Absence of relevant data for primary outcome

Fitzmaurice 1998 - Design

Hobbs 1996 - Outcomes did not fit inclusion criteria

Horn 2002 - Design

Hwang 2004 - Design

Kroese 2005 - Design

Manotti 2001 mainten - Absence of relevant data for primary outcome

McDonald 1980 - Dose prescribing rather than drug dosage

McMichael 1993 - No professional behaviour change or patient outcomes

Murchie 1989 - Absence of relevant data for primary outcome

Nieuwenhuyze 1995 - Computer controlled infusion not under physician control

Nightingale 2000 - Dose prescribing rather than drug dosage

Peck 1973 - Absence of relevant data for primary outcome

Peters 1996 - Patient aid not under physician control

Peterson 1986 - Patient aid not under physician control



(Continued)

Peterson 2005 - Not drug dosage

Poller 1993 - Outcomes did not fit inclusion criteria

Rood 2005 - Absence of relevant data for primary outcome

Rotman 1996 - Not computerized drug dosage

Ryff-de Leche 1992 - Infusion not under physician control

Strack 1985 - Design

Tamblyn 2003 - Not computer drug dosage

White 1984 - Outcomes did not fit inclusion criteria

Willcourt 1994 - Computer controlled infusion not under physician control



D A T A A N D A N A L Y S E S

Comparison 1. Dose of drug used

Outcome or subgroup titleNo. of

studies

No. of

participants Statistical method Effect size

1 Dose administered to the patient 12 Std. Mean Difference (IV, Random, 95% CI) Subtotals only

1.1 initial 5 345 Std. Mean Difference (IV, Random, 95% CI) 1.12 [0.33, 1.92]

1.2 maintenance 8 611 Std. Mean Difference (IV, Random, 95% CI) 0.19 [-0.10, 0.48]

1.3 total 4 280 Std. Mean Difference (IV, Random, 95% CI) 0.43 [-0.29, 1.16]

2 Number of doses adjustments 2 169 Std. Mean Difference (IV, Random, 95% CI) 0.26 [-0.46, 0.98]

Comparison 2. Serum concentrations and therapeutic range


studies

No. of


1 Serum concentrations 8 440 Std. Mean Difference (IV, Random, 95% CI) 1.12 [0.43, 1.82]

1.1 Theophylline 4 201 Std. Mean Difference (IV, Random, 95% CI) 0.41 [-0.20, 1.02]

1.2 Lidocaine 1 20 Std. Mean Difference (IV, Random, 95% CI) 1.32 [0.33, 2.32]

1.3 Aminoglycoside 3 219 Std. Mean Difference (IV, Random, 95% CI) 2.22 [0.04, 4.40]

2 Percentage of patients within

therapeutic range

3 212 Odds Ratio (IV, Fixed, 95% CI) 1.38 [0.71, 2.71]

3 Toxic Drug Levels 4 348 Risk Ratio (M-H, Random, 95% CI) 0.45 [0.30, 0.70]

Comparison 3. Physiological parameters


studies

No. of


1 Mean proportion of time spent

within target

2 135 Std. Mean Difference (IV, Random, 95% CI) 1.62 [-0.35, 3.59]



Comparison 4. Time to achieve therapeutic control


studies

No. of


1 Time to achieve therapeutic

range

2 223 Std. Mean Difference (IV, Random, 95% CI) -0.22 [-0.69, 0.26]

2 Time to stabilization 3 281 Std. Mean Difference (IV, Random, 95% CI) -0.55 [-1.03, -0.08]

Comparison 5. Clinical events


studies

No. of


1 Death 6 789 Risk Ratio (M-H, Random, 95% CI) 0.81 [0.37, 1.81]

2 Adverse reactions 10 Risk Ratio (M-H, Random, 95% CI) Totals not selected

3 Improvement 3 Risk Ratio (M-H, Random, 95% CI) Totals not selected

Comparison 6. Health care costs


studies

No. of


1 Length of stay 6 518 Std. Mean Difference (IV, Random, 95% CI) -0.35 [-0.52, -0.17]



Analysis 1.1. Comparison 1 Dose of drug used, Outcome 1 Dose administered to the patient.

Review: Computerized advice on drug dosage to improve prescribing practice

Comparison: 1 Dose of drug used

Outcome: 1 Dose administered to the patient

Study or subgroup Intervention Control Std. Mean Difference Weight Std. Mean Difference

N Mean(SD) N Mean(SD) IV,Random,95% CI IV,Random,95% CI

1 initial

Verner 1992 10 437 (71) 10 167 (23) 10.4 % 4.90 [ 2.99, 6.81 ]

Burton 1991 72 238 (64.8) 75 230 (49.7) 25.1 % 0.14 [ -0.19, 0.46 ]

Gonzalez 1989 37 4.2 (2.4) 30 3.8 (2.4) 23.9 % 0.16 [ -0.32, 0.65 ]

Hurley 1986 48 250 (41.5) 43 227 (41.5) 24.4 % 0.55 [ 0.13, 0.97 ]

Rodman 1984 9 82.68 (18.15) 11 42.27 (12.8) 16.0 % 2.51 [ 1.27, 3.75 ]

Subtotal (95% CI) 176 169 100.0 % 1.12 [ 0.33, 1.92 ]

Heterogeneity: Tau2 = 0.63; Chi2 = 36.48, df = 4 (P<0.00001); I2 =89%

Test for overall effect: Z = 2.77 (P = 0.0056)

2 maintenance

Burton 1991 72 272 (92.5) 75 261 (75.8) 16.0 % 0.13 [ -0.19, 0.45 ]

Carter 1987 20 7.16 (4.41) 19 7.82 (3.2) 10.3 % -0.17 [ -0.80, 0.46 ]

Gonzalez 1989 37 0.6 (0.2) 30 0.4 (0.2) 12.3 % 0.99 [ 0.48, 1.50 ]

Hurley 1986 48 831 (198) 43 698 (198) 14.0 % 0.67 [ 0.24, 1.09 ]

Rodman 1984 9 29.24 (15.93) 11 31.24 (7.59) 7.1 % -0.16 [ -1.04, 0.72 ]

Ruiz 1993 40 4.2 (2.8) 20 5.2 (2.6) 11.8 % -0.36 [ -0.90, 0.18 ]

Vadher 1997 pop 1 44 5 (2.67) 37 5 (3) 13.7 % 0.0 [ -0.44, 0.44 ]

Vadher 1997 pop2 46 6.5 (2.89) 60 6 (2.56) 14.8 % 0.18 [ -0.20, 0.57 ]

Subtotal (95% CI) 316 295 100.0 % 0.19 [ -0.10, 0.48 ]

Heterogeneity: Tau2 = 0.11; Chi2 = 20.83, df = 7 (P = 0.004); I2 =66%


3 total

Begg 1989 22 312 (79.7) 23 203 (62.3) 24.2 % 1.50 [ 0.83, 2.17 ]

Lesourd 2002 82 860 (382) 82 938 (516) 28.9 % -0.17 [ -0.48, 0.14 ]

Mungall 1994 25 1290 (430) 26 1190 (260) 25.9 % 0.28 [ -0.27, 0.83 ]

Rodman 1984 9 39.68 (21.09) 11 35.63 (14) 21.0 % 0.22 [ -0.66, 1.11 ]

Subtotal (95% CI) 138 142 100.0 % 0.43 [ -0.29, 1.16 ]



-10 -5 0 5 10

Favours control Favours intervention



Analysis 1.2. Comparison 1 Dose of drug used, Outcome 2 Number of doses adjustments.


Comparison: 1 Dose of drug used

Outcome: 2 Number of doses adjustments



Destache 1990 75 1.14 (0.97) 70 0.64 (0.83) 62.1 % 0.55 [ 0.22, 0.88 ]

Theil 1993 fentanyl 12 1.2 (1.8) 12 1.5 (0.8) 37.9 % -0.21 [ -1.01, 0.59 ]

Total (95% CI) 87 82 100.0 % 0.26 [ -0.46, 0.98 ]



-4 -2 0 2 4

Favours intervention Favours control

Analysis 2.1. Comparison 2 Serum concentrations and therapeutic range, Outcome 1 Serum

concentrations.


Comparison: 2 Serum concentrations and therapeutic range

Outcome: 1 Serum concentrations



1 Theophylline

Casner 1993 17 14.8 (4.4) 18 12.6 (4.1) 12.9 % 0.51 [ -0.17, 1.18 ]

Gonzalez 1989 37 14.6 (3.1) 30 11.4 (3.9) 13.6 % 0.91 [ 0.40, 1.42 ]

Hurley 1986 37 16.1 (5.2) 37 17.9 (7) 13.8 % -0.29 [ -0.75, 0.17 ]

Verner 1992 10 17 (5.06) 15 13.6 (5.8) 12.2 % 0.60 [ -0.22, 1.42 ]

Subtotal (95% CI) 101 100 52.4 % 0.41 [ -0.20, 1.02 ]



2 Lidocaine

Rodman 1984 9 5.3 (0.9) 11 3.7 (1.33) 11.3 % 1.32 [ 0.33, 2.32 ]

-10 -5 0 5 10


(Continued . . . )



(. . . Continued)Study or subgroup Intervention Control Std. Mean Difference Weight Std. Mean Difference


Subtotal (95% CI) 9 11 11.3 % 1.32 [ 0.33, 2.32 ]

Heterogeneity: not applicable


3 Aminoglycoside

Begg 1989 22 6.49 (0.39) 23 4.27 (0.52) 10.3 % 4.73 [ 3.55, 5.91 ]

Burton 1991 72 5.3 (1.8) 75 4.4 (1.7) 14.2 % 0.51 [ 0.18, 0.84 ]

Hickling 1989 13 7.45 (1.44) 14 5.14 (1.35) 11.8 % 1.61 [ 0.72, 2.49 ]

Subtotal (95% CI) 107 112 36.3 % 2.22 [ 0.04, 4.40 ]



Total (95% CI) 217 223 100.0 % 1.12 [ 0.43, 1.82 ]



-10 -5 0 5 10


Analysis 2.2. Comparison 2 Serum concentrations and therapeutic range, Outcome 2 Percentage of

patients within therapeutic range.



Outcome: 2 Percentage of patients within therapeutic range

Study or subgroup Intervention Control Odds Ratio Weight Odds Ratio

n/N n/N IV,Fixed,95% CI IV,Fixed,95% CI

Begg 1989 6/22 0/23 5.2 % 18.52 [ 0.97, 351.82 ]

Destache 1990 23/71 22/69 89.8 % 1.02 [ 0.50, 2.08 ]

Hickling 1989 13/13 8/14 5.0 % 20.65 [ 1.03, 415.43 ]

Total (95% CI) 106 106 100.0 % 1.38 [ 0.71, 2.71 ]

Total events: 42 (Intervention), 30 (Control)

Heterogeneity: Chi2 = 6.79, df = 2 (P = 0.03); I2 =71%


0.01 0.1 1 10 100




Analysis 2.3. Comparison 2 Serum concentrations and therapeutic range, Outcome 3 Toxic Drug Levels.



Outcome: 3 Toxic Drug Levels

Study or subgroup Intervention Control Risk Ratio Weight Risk Ratio

n/N n/N M-H,Random,95% CI M-H,Random,95% CI

Burton 1991 12/72 30/75 53.4 % 0.42 [ 0.23, 0.75 ]

Casner 1993 1/17 0/18 1.9 % 3.17 [ 0.14, 72.80 ]

Hurley 1986 9/48 16/43 36.9 % 0.50 [ 0.25, 1.02 ]

White 1987 2/39 6/36 7.8 % 0.31 [ 0.07, 1.43 ]

Total (95% CI) 176 172 100.0 % 0.45 [ 0.30, 0.70 ]


Heterogeneity: Tau2 = 0.0; Chi2 = 1.89, df = 3 (P = 0.60); I2 =0.0%


0.01 0.1 1 10 100

Favours experimental Favours control

Analysis 3.1. Comparison 3 Physiological parameters, Outcome 1 Mean proportion of time spent within

target.


Comparison: 3 Physiological parameters

Outcome: 1 Mean proportion of time spent within target



Ruiz 1993 40 72.8 (6.7) 20 51.2 (10.3) 49.0 % 2.65 [ 1.92, 3.37 ]

White 1987 39 58 (23) 36 42 (27) 51.0 % 0.63 [ 0.17, 1.10 ]

Total (95% CI) 79 56 100.0 % 1.62 [ -0.35, 3.59 ]



-10 -5 0 5 10




Analysis 4.1. Comparison 4 Time to achieve therapeutic control, Outcome 1 Time to achieve therapeutic

range.


Comparison: 4 Time to achieve therapeutic control

Outcome: 1 Time to achieve therapeutic range



Vadher 1997 76 3 (2.92) 72 3 (2.46) 55.8 % 0.0 [ -0.32, 0.32 ]

White 1987 39 3.2 (1.6) 36 4.5 (3.4) 44.2 % -0.49 [ -0.95, -0.03 ]

Total (95% CI) 115 108 100.0 % -0.22 [ -0.69, 0.26 ]



-4 -2 0 2 4


Analysis 4.2. Comparison 4 Time to achieve therapeutic control, Outcome 2 Time to stabilization.


Comparison: 4 Time to achieve therapeutic control

Outcome: 2 Time to stabilization



Carter 1987 31 6.8 (1.26) 34 8.42 (3.47) 30.9 % -0.60 [ -1.10, -0.10 ]

Vadher 1997 76 7 (3.75) 72 9 (15.3) 38.5 % -0.18 [ -0.50, 0.14 ]

White 1987 36 5.7 (1.7) 32 9.4 (5.2) 30.6 % -0.97 [ -1.47, -0.46 ]

Total (95% CI) 143 138 100.0 % -0.55 [ -1.03, -0.08 ]



-4 -2 0 2 4




Analysis 5.1. Comparison 5 Clinical events, Outcome 1 Death.


Comparison: 5 Clinical events

Outcome: 1 Death

Study or subgroup Intervention Control Risk Ratio Weight Risk Ratio


Begg 1989 1/22 5/23 12.3 % 0.21 [ 0.03, 1.65 ]

Burton 1991 1/68 3/68 10.7 % 0.33 [ 0.04, 3.13 ]

Destache 1990 14/75 7/70 38.4 % 1.87 [ 0.80, 4.35 ]

Fitzmaurice 2000 3/122 3/102 18.6 % 0.84 [ 0.17, 4.05 ]

Hurley 1986 0/48 2/43 6.4 % 0.18 [ 0.01, 3.64 ]

Vadher 1997 2/72 2/76 13.7 % 1.06 [ 0.15, 7.30 ]

Total (95% CI) 407 382 100.0 % 0.81 [ 0.37, 1.81 ]




0.5 0.7 1 1.5 2




Analysis 5.2. Comparison 5 Clinical events, Outcome 2 Adverse reactions.



Outcome: 2 Adverse reactions

Study or subgroup Intervention Control Risk Ratio Risk Ratio


Burton 1991 4/72 7/75 0.60 [ 0.18, 1.95 ]

Destache 1990 6/75 10/70 0.56 [ 0.21, 1.46 ]

Fitzmaurice 2000 3/122 3/102 0.84 [ 0.17, 4.05 ]

Gonzalez 1989 4/37 2/30 1.62 [ 0.32, 8.26 ]

Mungall 1994 0/25 6/26 0.08 [ 0.00, 1.35 ]

Rodman 1984 0/9 0/11 0.0 [ 0.0, 0.0 ]

Vadher 1997 6/72 5/76 1.27 [ 0.40, 3.97 ]

Vadher 1997 pop 1 3/37 3/44 1.19 [ 0.26, 5.54 ]

Vadher 1997 pop2 4/37 4/44 1.19 [ 0.32, 4.43 ]

White 1987 0/39 3/36 0.13 [ 0.01, 2.47 ]

0.5 0.7 1 1.5 2


Analysis 5.3. Comparison 5 Clinical events, Outcome 3 Improvement.



Outcome: 3 Improvement

Study or subgroup Intervention Control Risk Ratio Risk Ratio


Begg 1989 9/22 7/23 1.34 [ 0.61, 2.98 ]

Burton 1991 18/68 19/68 0.95 [ 0.55, 1.64 ]

Lesourd 2002 15/82 13/82 1.15 [ 0.59, 2.27 ]

0.5 0.7 1 1.5 2




Analysis 6.1. Comparison 6 Health care costs, Outcome 1 Length of stay.


Comparison: 6 Health care costs

Outcome: 1 Length of stay



Burton 1991 72 13 (11.15) 75 17.6 (11.15) 28.4 % -0.41 [ -0.74, -0.08 ]

Casner 1993 17 11.4 (21.6) 18 8.8 (15.4) 6.9 % 0.14 [ -0.53, 0.80 ]

Destache 1990 75 13.4 (11.3) 70 18.5 (22.4) 28.3 % -0.29 [ -0.62, 0.04 ]

Hurley 1986 48 6.3 (4.5) 43 8.7 (6.7) 17.5 % -0.42 [ -0.84, -0.01 ]

Verner 1992 10 4.4 (0.5) 15 4.4 (0.5) 4.7 % 0.0 [ -0.80, 0.80 ]

White 1987 39 13 (8) 36 20 (15) 14.2 % -0.58 [ -1.05, -0.12 ]

Total (95% CI) 261 257 100.0 % -0.35 [ -0.52, -0.17 ]

Heterogeneity: Tau2 = 0.0; Chi2 = 4.14, df = 5 (P = 0.53); I2 =0.0%


-4 -2 0 2 4


A P P E N D I C E S

Appendix 1. Search strategy

Search strategy - new

1. Drug Therapy, Computer-Assisted/

2. Hospital Information Systems/

3. exp Computer Systems/

4. computer$.tw.

5. Decision Support Systems, Clinical/

6. or/1-5

7. (advice or decision).tw.

8. decision making/

9. 7 or 8

10. ad.fs.

11. Prescriptions, Drug/

12. prescrib$.tw.

13. Drug Therapy/

14. dt.fs.

15. or/10-14

16. 6 and 15



17. 16 and 9

18. randomized controlled trial.pt.

19. controlled clinical trial.pt.

20. intervention studies/

21. experiment$.tw.

22. (time adj series).tw.

23. (pre test or pretest or (posttest or post test)).tw.

24. random allocation/

25. impact.tw.

26. intervention?.tw.

27. chang$.tw.

28. evaluation studies/

29. evaluat$.tw.

30. effect?.tw.

31. comparative study.pt.

32. or/18-31

33. 17 and 32

34. limit to review

35. 33 not 34

36. meta-analysis.pt.

37. 35 not 36

38. limit 37 to yr=1996-2006

Search strategy - old

1.Computer systems/

2.Artificial intelligence/

3.1 or 2

4.(prescr* Or drug therapy)

5.3 and 4

6.comparative study/

7.clinical trials/

8.6 or 7

9.5 and 8

F E E D B A C K

Feedback from Andrew Herxheimer, 16 September 2008

Summary

Computerized advice on drug dosage to improve prescribing practice

1. The review is interesting and useful, but the ’broad brush’ approach ignores some important details.

2. The hypothesis that Decisions on drug dosage based on computer advice lead to fewer unwanted effects is too general. Analysis 5.2

shows a meta-analysis of unspecified adverse reactions to unspecified drugs. It makes no sense to assume that these 11 studies are

homogeneous, let alone comparable. A meta-analysis is inappropriate. It seems unreasonable to expect computer help with dosing to

affect all adverse reactions similarly. The adverse reactions need to be looked at - what was the drug, what was the reaction, at what

point during treatment did it occur, might it have been predicted or prevented? Of course many or most of the reports of these

studies will have included little detail; I ask the review authors to tell us what the reports did include and what they would have

wished them to include.



3. The studies should be grouped by the drugs used and the effect aimed at - anticoagulants, aminoglycosides, theophylline,

lidocaine, fentanyl/midazolam, nitroprusside, etc. In the Characteristics of included studies would be better arranged chronologically

(within drugs); the alphabetic order of first authors’ names is a distracting irrelevance.

4. It should be noted that lidocaine and theophylline are now obsolete.

Reply

We thank Andrew Herxheimer for his feedback. His comments deal mainly with the great clinical diversity between studies included

in this review. This diversity was inherent to computer advice on drug dosage, which has been evaluated in several clinical fields for

different drugs on different outcomes. This resulted in a great heterogeneity in all comparisons analysed in this review. We discussed

this heterogeneity widely in the manuscript (see the Results and Discussion sections).

We agree that heterogeneity could be a specific problem for comparison 5.2. According to the Cochrane handbook for systematic

reviews of intervention, chapter 14.6, we reanalysed the 11 concerned studies, in order to answer these specific questions:

• Are definitions of reported adverse effects given?

• Were the methods used for monitoring adverse effects reported? And how? (prospective or routine monitoring, spontaneous

reporting, questionnaire or diary, systematic survey of patients?

• Were any patients excluded from the adverse effect analysis

• What was the length of follow up?

Results are presented in the following table.

Assessment of the quality of evidence on adverse reactions

Study Drug Definition Monitoring

specified a pri-

ori

Type Methods used

for monitoring

Exclusion of

patients

Maxi-

mum length of

follow-up

Burton 1991 Aminoglyco-

sides

Yes Yes Nephrotoxic-

ity

Routine moni-

toring

None In-hospital

Destache

1990

Aminoglyco-

sides

Yes Yes Nephrotoxic-

ity

NR Yes (55/200

patients)

In-hospital

Fitzmaurice

2000

Anticoagu-

lants

No Yes Thrombosis,

haemorrage

NR None 12 months

Gonzalez

1989

Amino-

phylline

Yes No Nausea, vomit-

ing

NR Yes (15/82 pa-

tients)

In-hospital

Mungall 1994 Heparin Yes Yes Clinical events

(recur-

rent chest pain,

reoc-

clusion, stroke,

CHF, car-

diac arrest) and

bleeding events

(which were re-

ported

separetely)

NR None In-hospital



(Continued)

Rodman 1984 Lidocaine No No Toxic response

requiring

reduced dosage

or discontinu-

ation of lido-

caine

NR None In-hospital

Vadher 1997 Warfarin No Yes Thrombosis,

haemorrage

NR None 13 months

Vadher 1997

pop1

Warfarin No Yes Thrombosis,

haemorrage

NR None 3 months

Vadher 1997

pop2

Warfarin No Yes Thrombosis,

haemorrage

NR None 6 months

White 1987 Warfarin Yes Yes Bleeding com-

plications

NR None In-hospital

After reviewing all the studies, we decided to exclude one study (Lesourd 2002) from comparison 5.2 because the event we considered

was not defined as an adverse reaction by the authors (cancelled cycles of ovarian stimulation).

Moreover, as suggested by the reviewer, we suppressed the meta-analysis for comparison 5.2.

We disagree with the reviewer who considered that Lidocaine and Theophylline are now obsolete. Theophylline still appears in stage

4 of the UK British Thoracic Association guidelines albeit as slow release form and as intravenous in acute severe asthma as alternative

treatment. It also still appear in French recommendations for management of asthma. We acknowledged in the discussion section that

Theophylline is not a first-choice drug. However, if there were a safer, better way of administering it (ie computerized dosing) then it

might be more widely used. It probably does still have a role orally in COPD although it has always been the case that getting the best

effect from it required therapeutic drug monitoring. Computer advice could minimize the need for testing as with warfarin.

Like theophylline, lidocaine is still referenced in large print in the British National Formulary (BNF). Lidocaine is still suggested as

first choice for emergency use for ventricular arrhythmias in the BNF.

Finally, we will consider the change of the classification of studies in the next update of the review.

Contributors

Feedback from: Dr. Andrew Herxheimer, London, UK

Response from: Dr Pierre Durieux, Paris, France.

Feedback from Sylvain Goutelle, 19 July 2010

Summary

This review is very interesting and useful for further development of computer-assisted individualization of drug dosage regimens.

However, it seems that the search strategy did not retrieve some relevant references in the field. I identified five studies which have not

been cited in this article (even in the references excluded from the review), while their subjects, designs and methods appear to match

the selection criteria (all those studies are classified as randomized clinical trials or controlled studies in Medline).

Van Lent-Evers et al performed a cost-effectiveness analysis of Bayesian adaptive control of aminoglycoside dosing [1]. The method was

implemented in a pharmacokinetic software. A similar computerized method was used by del Mar Fernandez de Gatta et al to assess

vancomycin dosage individualization in patients with hematologic diseases [2]. Le Meur and colleagues compared a concentration-

controlled strategy versus a fixed-dose strategy in renal transplant patients treated with mycophenolate-mofetil. The concentration-



controlled method used a computer program for Bayesian calculation of mycophenolic acid AUC and adjustment of MMF dose

[3]. In a psychiatric institution, Mihajlovic et al compared empiric dosage regimen of amitriptyline with Bayesian individualization

supported by a specific computer program [4]. Finally, in children with acute lymphoblastic leukemia, Evans et al compared conventional

chemotherapy based on body surface area to an individualized therapy based on Bayesian estimation of drug clearances [5].

In summary, Bayesian methods supported by various programs were used in all of those five studies to individualize the dosage regimens

of drugs, and such individualized therapies were compared with usual strategies. All these studies indicated that computer-assisted

individualization may have some benefits. I think that those studies should have been considered in the review by Durieux et al. Their

incorporation might modify some of the author’s conclusions.

Of note, I did not perform a systematic search, so other studies might have been missed. The authors should consider revising their

search strategy for the next update of this review.

References

1. N.A.E.M. van Lent-Evers, R.A.A. Mathôt, W.P. Geus,B.A. van Hout, and A.A.T.M.M. Vinks. Impact of goal-oriented and

model-based clinical pharmacokinetic dosing of aminoglycosides on clinical outcome: a cost-effectiveness analysis. Therapeutic Drug

Monitoring 1999;21:63-73

2. M. del Mar Fern?ndez de Gatta, M.V. Calvo, J.M. Hern?ndez, D. Caballero, J.F. San Miguel, and A. Dom?nguez-Gil. Cost-

effectiveness analysis of serum vancomycin concentration monitoring in patients with hematologic malignancies. Clinical Pharmacology

& Therapeutics 1996;60:332-340

3. Y. Le Meur, M. B?chler, A. Thierry, S. Caillard, F. Villemain, S. Lavaud, I. Etienne, P.-F. Westeel, B. H. de Ligny, L. Rostaing, E.

Thervet, J. C. Szelag, J.-P. R?rolle, A. Rousseau, G. Touchard, and P. Marquet. Individualized mycophenolate mofetil dosing based on

drug exposure significantly improves patient outcomes after renal transplantation. American Journal of Transplantation 2007;7:2496-

2503

4. G.S. Mihahlovic, D.R. Milovanovic, and S.M. Jankovic. Comparison of efficacy and safety between individualized and empiric

dose regimen of amitriptyline in the treatment of major depressive episode. Psychiatry and Clinical Neurosciences 2003;57:580-585

5. W.E. Evans, M.V. Relling, J.H. Rodman, W.R. Crom, J.M. Boyett, and C.H. Pui. Conventional compared with individualized

chemotherapy for childhood acute lymphoblastic leukemia. New England Journal of Medicine 1998;338:499-505

Reply

Thank you for your feedback. We have added these five studies to the Studies Awaiting Classification. We will fully assess them as part

of our next update and also modify our search strategy so that the studies are identified.

Contributors

Sylvain Goutelle

Pierre Durieux

W H A T ’ S N E W

Last assessed as up-to-date: 13 May 2008.

Date Event Description

8 September 2010 Feedback has been incorporated Feedback provided on searching and addtional studies. Studies added to

“Studies awaiting Assessment”.

8 September 2010 Amended Five new studies added to “Studies Awaiting Assessment” in response to

feedback received.



H I S T O R Y

Review first published: Issue 1, 2001

Date Event Description

12 November 2008 Amended Change in contact details

15 May 2008 New citation required and conclusions have changed Substantive amendment

15 May 2008 New search has been performed New searches, new studies, updated results

14 May 2008 New citation required and conclusions have changed New team of authors, updated search, updated con-

clusions.

10 April 2008 Amended Converted to new review format.

C O N T R I B U T I O N S O F A U T H O R S

PD, IC, LT and JN prepared the protocol. All authors applied the inclusion criteria, assessed the quality and extracted the data for

the included studies. PD and LT conducted the quantitative analyses and qualitative analyses. PD drafted the manuscript with input

from LT and IC. BB, MR, JN provided comments on the manuscript. RW conducted the initial review and provided comments on

the revised manuscript.

D E C L A R A T I O N S O F I N T E R E S T

During the review, JN was a PhD student sponsored by MEDASYS SA.

I N D E X T E R M SMedical Subject Headings (MeSH)

∗Drug Therapy, Computer-Assisted; ∗Physician’s Practice Patterns; Medication Errors [prevention & control]; Randomized Controlled

Trials as Topic

MeSH check words

Humans



computerized advice on drug dosage to improve prescribing practice

Documents