Why did you choose the individual participant data approach for this systematic review?

Often, traditional meta-analyses aggregate data from clinical trials that may include different cohorts, use slightly different treatments or measure outcomes via different methods.

Combining trial data in this scenario can lead to results that are difficult to interpret—making them not particularly clinically relevant.

Additionally, combined trial results in a traditional meta-analysis are a blunt tool, only providing the treatment effect in an overall population.

It’s not possible to explore differences in outcomes within subgroups of patient types (that is, patients who are elderly, obese or with multiple comorbidities).

An individual participant data meta- analysis is a method that combines the complete raw datasets of the detailed patient-level data from highly similar trials into one mega-database.

This approach not only increases the statistical power for analyses, but also allows for results to be statistically adjusted for factors that may confound results and permits the exploration of how a specific treatment may be more or less effective in different subgroups of patient types.

This type of analysis can provide clinically relevant answers to questions that guide implementation and service planning.

For example, ‘In a resource-poor environment, can I target the types of patient that this treatment works most effectively for? Are there some types of patients that I could choose not to treat because the treatment is least beneficial for them?’

Professor Monika Fagevik Olsén

How much did preoperative physiotherapy education with breathing exercise training reduce the incidence of postoperative pulmonary complications?

Combining data from these two very similar trials found that a single preoperative physiotherapy session halved the risk of developing a postoperative pulmonary complication (PPC).

The combined sample size of 800 participants provided strong statistical power with a 95% confidence interval indicating that this type of physiotherapy treatment is highly effective (OR 0.53; 95% CI 0.34 to 0.85).

At the very least, preoperative physiotherapy reduced a patient’s risk of a PPC after abdominal surgery by 15 per cent.

At best, it reduced PPC risk by 66 per cent.

This is a large significant effect and a clinically important outcome for patients and hospitals.

What about hospital length of stay and mortality?

Previously, individual trials were not large enough to provide confident estimates of effects in regards to secondary outcomes such as length of stay and mortality.

Hospital length of stay varies greatly between patients and hospitals.

A single trial would need 2000 participants to be certain of detecting a clinically significant one-day reduction in length of stay due to a physiotherapy intervention following abdominal surgery.

Despite the increased sample size and statistical power in our individual participant meta-analysis, we were unable to determine with certainty that preoperative physiotherapy reduces length of stay.

We detected a one-day difference favouring the treatment group—although this value ranged from a possible two-day reduction to a small 0.3 increase in length of stay in the overall population treated.

However, conducting this individual participant data meta-analysis enabled us to explore effects on length of stay in different patient subgroups.

Here we detected a significant three-day reduction in hospital stay in patients with multiple comorbidities provided with preoperative physiotherapy.

This effect could be as large as six days or as small as 0.3 days.

We can now say, with increased certainty, that preoperative physiotherapy reduces the length of stay for comorbid patients.

Thankfully, postoperative mortality occurs rarely; less than one per cent of patients die within 30 days of surgery.

However, by 12 months, approximately 10 per cent of patients having major abdominal surgery will be deceased.

We detected an absolute three per cent drop in mortality favouring the treatment group (10 per cent versus 7 per cent).

Nevertheless, this small but clinically significant difference was not statistically significant.

A study would require 2700 patients to be certain that this difference is true.

Dr Ianthe Boden

How stable were the treatment effects across different postoperative pulmonary complication definitions, including pneumonia?

We tested the effect of preoperative physiotherapy on four different definitions of postoperative respiratory compromise, including pneumonia.

Regardless of the definition, the effect of preoperative physiotherapy was the same—the risk was halved.

How much did the treatment effects vary within clinically relevant subgroups?

Preoperative physiotherapy was significantly more effective in patients with any of the following characteristics: multiple comorbidities, abnormal body weight, under 45 years, current smoker or having upper abdominal surgery.

In these types of patients, preoperative physiotherapy reduced PPC by 60 to 90 per cent.

Preoperative physiotherapy was least effective in patients having operations that are less than three hours long.

>> Monika Fagevik Olsén is a Professor at Gothenburg University and senior consultant physiotherapist at the Sahlgrenska University Hospital, Sweden.

>> Ianthe Boden is an NHMRC research fellow at the University of Tasmania and surgical clinical lead physiotherapist at the Launceston General Hospital.

>> Click here to read the paper.