ObjectiveMeta-analyses of observational studies are frequently published in the literature, but they are generally considered suboptimal to those involving randomised controlled trials (RCTs) only. This is due to the increased risk of biases that observational studies may entail as well as because of the high heterogeneity that might be present. In this article, we highlight aspects of meta-analyses with observational studies that need more careful consideration in comparison to meta-analyses of RCTs.MethodsWe present an overview of recommendations from the literature with respect to how the different steps of a meta-analysis involving observational studies should be comprehensively conducted. We focus more on issues arising at the step of the quantitative synthesis, in terms of handling heterogeneity and biases. We briefly describe some sophisticated synthesis methods, which may allow for more flexible modelling approaches than common meta-analysis models. We illustrate the issues encountered in the presence of observational studies using an example from mental health, which assesses the risk of myocardial infarction in antipsychotic drug users.ResultsThe increased heterogeneity observed among studies challenges the interpretation of the diamond, while the inclusion of short exposure studies may lead to an exaggerated risk for myocardial infarction in this population.ConclusionsIn the presence of observational study designs, prior to synthesis, investigators should carefully consider whether all studies at hand are able to answer the same clinical question. The potential for a quantitative synthesis should be guided through examination of the amount of clinical and methodological heterogeneity and assessment of possible biases.

No abstract

Background: Treatment free remission (TFR) is now a realistic goal of treatment for CML. Approximately 50% of patients (pts) who discontinue tyrosine kinase inhibitors (TKI) after achieving deep molecular responses (DMR) are able to remain off treatment without losing major molecular response (MR3). Data from the largest available TKI stopping trial, EURO-SKI, showed that the most important variable associated with prolonged TFR is the duration of DMR. However, to date no clinical tool exists to guide clinicians and patients in predicting the likelihood of success of TFR attempt. Methods: We performed a retrospective analysis of clinical data from 172 pts with CML in whom treatment was discontinued in 6 hospital centres in order to identify factors associated with TFR. Data analysis started with a training set (TS) derived from pts treated at a single centre which was then validated on a validation set (VS) derived from the 5 other centres. Eligibility criteria included diagnosis of CML in chronic phase, a minimum duration of treatment with TKI of 3 years and discontinuation of TKI after achievement of confirmed ≥MR4. Patients diagnosed in accelerated phase CML and/or who underwent prior allogeneic stem cell transplant were excluded. Kaplan-Meier method was used for univariate analysis, with log-rank test for group comparison. A Cox proportional hazards model was employed with the purpose of choosing the most influential prognostic predictors on the probability of TFR in MR3 (pTFR3) and TFR in MR4 (pTFR4) on the TS. Variables with a p-value ≤0.1 entered in the multivariate analysis (MVA). Proportional hazard assumptions were tested for the final model. A prognostic TFR score was built from the combination of the predictors identified by the Cox model and validated on the VS. Results: The TS included 118 pts, while the VS 54 pts (Table 1). In the TS, the 2-year pTFR3 was 67.4% (95% CI 66.5-68.3%) and the 2-y pTFR4 was 56.8% (95% CI, 55.9-57.7%). The median time to MR3 loss was 3.8 months (range 1-31), and for MR4 loss was 3.2 months (range 0.8-24.5). After loss of MR4, the 1-year probability of MR3 loss was 77% (95% CI, 70.8-73.2%). However, 10 pts (8.5%) resumed TKI after MR4 loss and were not evaluable for time to loss of MR3. In univariate analysis, the variables most significantly associated with higher pTFR3 and pTFR4 were age at diagnosis >40 years (p=0.029 and p=0.002), absence of previous TKI resistance (p=0.003 and p= 0.068), longer duration of MR4 (p=0.003 and p<0.0001) and ≥MR4.5 at stopping (p=0.026 and p= 0.004). Variables entered into the MVA were age at diagnosis, BCR-ABL1 transcript type, Sokal score, dose of TKI at stopping, previous TKI resistance, duration of MR4 at stopping, depth of response at stopping. The Cox model suggested the inclusion of the following variables, for both pTFR3 and pTFR4: duration of MR4, previous TKI resistance, age at diagnosis and transcript type. Using these variables we developed a predictive score (Figure 1a), which was able to identify a good risk population (2-y pTFR3 81.8%, 2-y pTFR4 80%); intermediate (66.6% and 61.5%) and poor risk (42.3% and 30.8%) (overall log-rank test p=0.00092 and p <0.0001 for pTFR3 and pTFR4, respectively)(Figure 1b). The score was tested on the VS of 54 pts. In this population, the overall 2-y pTFR3 and pTFR4 were 61.3% (95% CI, 59.8-62.7%) and 42.6% (95% CI, 41.2-44%), respectively. Despite the small sample size, our score was still able to predict different 2-y TFR probabilities (Figure 1c) in the three risk groups. Of the pts who lost MR3 in the TS (n=39), 37 regained ≥MR3 after resuming TKI; 1 patient did not restart TKI and died from an unrelated cause; 1 patient had only 2 months follow-up after TKI resumption. In the VS, 15 of 21 pts losing MR3 achieved ≥MR3 again after TKI resumption; 3 pts had a follow-up <3 months after MR3 loss, 2 were lost to follow-up and 1 had not yet re-gained MR3 6 months after restarting TKI. In both cohorts no case of disease progression had occurred at last follow-up. Conclusions:This retrospective study confirms the safety of TFR attempt and identifies variables strongly associated with prolonged TFR. The resulting predictive score presented here, if validated in larger patient cohorts, might help in tailoring the choice of TKI discontinuation to the individual patient. Also, most pts who lose MR4 inevitably lose MR3, suggesting the importance of a more intense monitoring strategy in this subgroup. Disclosures Claudiani: Pfizer: Honoraria; Incyte: Honoraria. Byrne:Ariad/Incyte: Honoraria, Speakers Bureau. Rothwell:Incyte: Speakers Bureau; Novartis: Honoraria, Other: advisory board; Pfizer: Speakers Bureau. Copland:Incyte: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau; Pfizer: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Novartis: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau; Astellas: Honoraria, Speakers Bureau; Cyclacel: Research Funding; Bristol-Myers Squibb: Honoraria, Membership on an entity's Board of Directors or advisory committees, Speakers Bureau. Clark:Ariad/Incyte: Honoraria; Pfizer: Honoraria, Research Funding; BMS: Honoraria, Research Funding; Novartis: Honoraria, Research Funding. Milojkovic:BMS: Honoraria, Speakers Bureau; Pfizer: Honoraria, Speakers Bureau; Incyte: Honoraria, Speakers Bureau; Novartis: Honoraria, Speakers Bureau. Apperley:Pfizer: Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau; Novartis: Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau; Incyte: Membership on an entity's Board of Directors or advisory committees, Research Funding, Speakers Bureau; Bristol Myers Squibb: Membership on an entity's Board of Directors or advisory committees, Speakers Bureau.