Article type
Year
Abstract
Background:
Individual patient data (IPD) are considered to be a promising source of information and can offer additional input to traditional systematic reviews. Nonetheless, more empirical evidence is needed to make clear the whole range of benefits from IPD compared to aggregate data (AGD).
Objectives:
To describe the case of a recently published review on blood pressure targets (DOI:10.1002/14651858.CD010315.pub2), in which the results for primary outcomes changed substantially after using IPD.
Methods:
In October 2017 a Cochrane systematic review on hypertension was published, including IPD from five clinical trials (AASK, ACCORD, HOT, PAST-BP, SPRINT; n = 6775 participants) and AGD from one study (SPS3, n = 3020 participants). Once we obtained IPD for the SPS3 trial from a public repository, two review authors independently extracted the data and re-calculated all primary outcome estimates in order to update the meta-analysis. We checked the forest plots before and after the update looking for major differences.
Results:
According to the GRADE assessment, there were four primary variables in the protocol rated as key outcomes: total mortality, serious adverse events (SAE), total cardiovascular events and cardiovascular mortality. IPD analysis led to the exclusion of normotensive patients in the SPS3 trial (311 out of 3020 patients).
In terms of the whole meta-analysis, we observed relevant changes in the results for some of these outcomes (Figure 1). No differences in SAE between interventions were confirmed, with a greater number of participants identified (572 instead of 93) based on IPD. The mean estimate for cardiovascular mortality changed from 0.96 with AGD (95% confidence interval (CI) 0.77 to 1.21) to 1.03 with IPD (95% CI 0.82 to 1.29). For total cardiovascular events, statistical significance was lost after considering IPD as the source of data (0.89, 95% CI 0.80 to 1.00) compared to the previous situation with AGD (0.87, 95% CI 0.78 to 0.98).
Conclusions:
In our example, additional IPD had a large effect on the final results for several main outcomes. In particular, an IPD strategy seems to be successful in identifying unpublished adverse events. Using IPD in a systematic review can improve the precision of meta-analysis estimates, lower the risk of bias and benefit patients in a practical way.
Patient or healthcare consumer involvement:
None.
Individual patient data (IPD) are considered to be a promising source of information and can offer additional input to traditional systematic reviews. Nonetheless, more empirical evidence is needed to make clear the whole range of benefits from IPD compared to aggregate data (AGD).
Objectives:
To describe the case of a recently published review on blood pressure targets (DOI:10.1002/14651858.CD010315.pub2), in which the results for primary outcomes changed substantially after using IPD.
Methods:
In October 2017 a Cochrane systematic review on hypertension was published, including IPD from five clinical trials (AASK, ACCORD, HOT, PAST-BP, SPRINT; n = 6775 participants) and AGD from one study (SPS3, n = 3020 participants). Once we obtained IPD for the SPS3 trial from a public repository, two review authors independently extracted the data and re-calculated all primary outcome estimates in order to update the meta-analysis. We checked the forest plots before and after the update looking for major differences.
Results:
According to the GRADE assessment, there were four primary variables in the protocol rated as key outcomes: total mortality, serious adverse events (SAE), total cardiovascular events and cardiovascular mortality. IPD analysis led to the exclusion of normotensive patients in the SPS3 trial (311 out of 3020 patients).
In terms of the whole meta-analysis, we observed relevant changes in the results for some of these outcomes (Figure 1). No differences in SAE between interventions were confirmed, with a greater number of participants identified (572 instead of 93) based on IPD. The mean estimate for cardiovascular mortality changed from 0.96 with AGD (95% confidence interval (CI) 0.77 to 1.21) to 1.03 with IPD (95% CI 0.82 to 1.29). For total cardiovascular events, statistical significance was lost after considering IPD as the source of data (0.89, 95% CI 0.80 to 1.00) compared to the previous situation with AGD (0.87, 95% CI 0.78 to 0.98).
Conclusions:
In our example, additional IPD had a large effect on the final results for several main outcomes. In particular, an IPD strategy seems to be successful in identifying unpublished adverse events. Using IPD in a systematic review can improve the precision of meta-analysis estimates, lower the risk of bias and benefit patients in a practical way.
Patient or healthcare consumer involvement:
None.