Skip navigation

Do weight-reducing drugs work?

Orlistat [1]
Results
Sibutramine [2]
Comment
Pointy-head point
Obesity is a health problem all of us know about, and which concerns many people. A tablet that helped lose weight is something many of us would welcome, and some drug therapies are available. A frequently-asked question is whether anti-obesity drugs work, and how well they work?

We now have two excellent reviews from Health Technology Assessment, which can be downloaded in full from the Internet at no cost [1,2]. Bandolier thought it might be useful to provide a thumbnail of the results to try and answer the frequently-asked questions.

Orlistat [1]


This review sought randomised trials of any duration of therapy or length of follow up (with a minimum duration of one year from company submissions to NICE). Participants had to be overweight or obese, or who wished to maintain weight loss having previously been overweight or obese. Trials of people with eating disorders, or with healthy weight were excluded. Orlistat could be combined with other strategies like dietary restriction or behavioural programmes. Controls could be placebo, other anti-obesity agents, or alternative interventions like dietary regimens or physical activity or behavioural modification. Outcomes were obesity or overweight status, like body weight, fat content or fat distribution.

Results


Four trials with a one-year outcome had 1274 patients given orlistat 120 mg three times a day and 837 patients given placebo. At baseline BMI was in the range of 28 to over 40 kg/sq metre. The overall weight loss was 2.9 kg (95% confidence interval 2.2 to 3.6 kg).

In four trials, at least 5% reduction in body weight at one year was achieved by 661/1144 (58%) of people on orlistat 120 mg three times a day compared with 225/705 (32%) on placebo (Figure 1). The relative benefit was 1.6 (1.5 to 1.8) and the number needed to treat compared with placebo was 3.9 (3.3 to 4.6) (Table 1).

Figure 1: At least 5% weight reduction at one year with orlistat




Table 1: Calculating the NNT the correct way and the wrong way

 

Correct way

Wrong way

Number/Total (%)

Number/Total (%)

Treatment

Outcome

Treatment

Placebo

NNT
(95%CI)

Treatment

Placebo

NNT
(95%CI)

Orlistat 360 mg daily At least 5% weight reduction at one year 661/1144
(58)
225/708
(32)
3.9 (3.3 to 4.6) 201/400
(50)
119/400
(30)
4.9 (3.7 to 7.2)
Orlistat 360 mg daily At least 10% weight reduction at one year 460/1388
(33)
146/951
(15)
5.6 (4.7 to 6.9) 140/500
(28)
70/500
(14)
7.1 (5.3 to 11)
Sibutramine 10 mg daily At least 5% weight reduction at six months 130/205
(63)
52/202
(26)
2.7 (1.9 to 3.5) 133/200
(67)
60/200
(30)
2.7 (2.2 to 3.7)

In five trials, at least 10% reduction in body weight at one year was achieved by 460/1388 (33%) of people on orlistat 120 mg three times a day compared with 146/951 (15%) on placebo (Figure 2). The relative benefit was 1.9 (1.6 to 2.3) and the number needed to treat compared with placebo was 5.6 (4.7 to 6.9) (Table 1).

Figure 2: At least 10% weight reduction at one year with orlistat



Gastrointestinal adverse effects were mentioned in most trials, but no analysis is given, nor probably was possible. Adverse event withdrawals in four trials at one year were no different for orlistat (97/1178, 8%) or placebo (44/739, 6%). In an economic analysis the report tells us that treating 100 patients with orlistat for two years will cost about £73,000. The cost per quality adjusted life year gained is about £46,000 with a range of £19,000 to £55,000.

Sibutramine [2]


The format of the review of sibutramine, including trial inclusion, exclusion, and outcomes was broadly similar to that for orlistat. Many trials were of short duration.

The most useful outcome was that of at least 5% reduction in body weight at six months. In two trials that reported it, this was achieved by 130/205 (63%) of patients on 10 mg sibutramine daily, compared with 52/202 (26%) on placebo. The relative benefit was 2.5 (1.9 to 3.2) and the number needed to treat compared with placebo was 2.7 (2.2 to 3.5) (Table 1). No clear analysis of adverse events seemed possible. Estimates of cost per quality of life gained varied from £5,700 to £35,200, though with reservations.

Comment


Obviously there is much, much more in these reports than is abstracted here. The great thing is that those of us with an interest can download them from the HTA Internet site and read them at our leisure and make up our own minds.

On the face of it these are reasonable clinical results. But adverse events have to be weighed against efficacy, and, as usual, this seems to be a bit of a black hole. There are other methodological issues. One, for example, is that of how patients were recruited. This was sometimes through advertisements, sometimes through hospitals, but there is a residue of uncertainty over whether the people in these trials were like our patients. Another involves what to do about withdrawals. Some trials used a method where the last actual measurement was carried forward for missing data. In this situation, this may not be a conservative way of doing things, and may overstate treatment efficacy.

Then there is the uncertainty over cost effectiveness. The imputed costs per quality of life year gain were high. Perhaps there are better ways of spending money to help overweight people who want to lose weight.

Pointy-head point


Of interest to all those learning or teaching critical appraisal skills, this concerns calculating NNTs. In both these reports dichotomous outcomes like the number of patients with a particular level of weight loss at a particular time was reported as percentages. The figures were used (X/100 for treatment, Y/100 for placebo for trial 1, A/100, B/100 for trial 2) in RevMan to calculate relative risks.

But the numbers of patients were different from 100, and different trials had different numbers of patients. The weight we give each trial is more or less dependent on the number of patients. So the method used had incorrect weighting.

Actually relative risks are about right, though confidence intervals will not be. NNTs can be wrong, as Table 1 demonstrates by showing the calculations in the right and wrong way. If, as with sibutramine, there is not much difference in total numbers, the effect will be slight. Where the difference is large, and especially where trial sizes are very different, the effect will be greater, as with orlistat.

Bandolier doesn't wish to be picky about two good reports, but there is a lesson for us. We all make mistakes. This mistake was not picked up by the authors, their advisors, their reviewers, or the companies making the products (as best we can tell). Crass mistakes will get through. In the end it is our responsibility to know enough of the basics to pick up mistakes made in good faith.

References:

  1. S O'Meara et al. A rapid and systematic review of the clinical effectiveness and cost-effectiveness of orlistat in the management of obesity. Health Technology Assessment 2001 5:18 (available on the Internet at http://www.ncchta.org/HTAPUBS.HTM )
  2. S O'Meara et al. The clinical effectiveness and cost-effectiveness of sibutramine in the management of obesity: a technology assessment. Health Technology Assessment 2002 6:6 (available on the Internet at http://www.ncchta.org/HTAPUBS.HTM )
previous or next story in this issue