Lost to follow-up: Where did the lost patients go?

Loss to follow-up is a common issue encountered in clinical trials, particularly in those looking at long-term outcomes. Lost to follow-up (LTFU) is a term used to describe individuals who drop out of medical research studies before completion, or who otherwise become unreachable to the study investigators. Loss to follow-up is related to attrition, a term describing the loss of consented participants during the course of a study, and is essentially the opposite of patient retention.

Reasons for loss to follow-up

A participant could become lost to follow up due to patient-related factors, some of which include moving, changing their contact information (phone number, etc.) without informing the investigators, inability to continue the study. Study-related factors and institutional factors, such as staff shortages and resulting miscommunication, technical problems with scheduling software or reminders, or overly burdensome protocols are other possible causes of loss to follow-up.

Importance of follow-up in medical research

In medical research studies, follow-up is important for tracking and demonstrating changes due to interventions or treatments over time. Following up with participants provides additional insight into how each individual responds in different environments, and reveals information that might otherwise remain unknown without proper follow-up. Thus, it allows researchers to draw better conclusions and understand the implications of their findings from these types of studies more thoroughly.

Why is loss to follow-up important?

Loss to follow up is important because it can lead to problems for the validity of the study results, besides being a potential indicator of negative participant experiences. Many studies, such as randomized placebo-controlled trials or longitudinal studies, are designed in such a way that information is needed on all trial participants in order to assess the efficacy of the study drug and key safety parameters with optimum accuracy. Missing part of this crucial data can jeopardize the strength of the data and results.

In what study design is loss to follow-up generally more of a concern?

Loss to follow-up generally poses a relatively greater risk in clinical trials studying-long term outcomes, as the extended study durations present more opportunity for drop-outs or other circumstances leading to LTFU to occur. In some drug effectiveness studies, patients are monitored over a number of years rather than just weeks or months. Studies involving arduous protocols or procedures, excessive travel time, or other burdensome aspects are also more likely to have higher loss to follow-up. For this reason, there is an increasing focus on the design of patient-centric clinical trials.

Effects of loss to follow-up on study power and potential bias

Researchers determine sample size for a clinical trial intentionally, based on various factors, in order to optimize the potential of the study to provide strong, statistically valid results. Statistical power describes how well a study can detect a difference between the two groups being tested, and is directly related to the number of participants involved (sample size). Each time a participant becomes lost to follow-up, the lost patient exerts an effect on the study’s power by reducing sample size.

LTFU has the potential to confound study results even further by introducing a type of selection bias, particularly in the case where there is a common factor amongst the participants who are lost to follow up (for example, they belong to the same study arm or have a unifying demographic characteristic).[1] In such a case, the groups - who were originally randomized in order to minimize variability between prognostic factors (such as age, sex, co-illnesses, etc.) - become less similar, making statistical analysis less effective.

How do you prevent loss to follow up bias?

To avoid potential bias introduced due to loss of data resulting from loss to follow up, the first line of defense is to try to prevent the attrition in the first place! [2] Some strategies include employing retention strategies throughout all phases of the study, such as:

• Designing patient-centric studies and reducing patient burden, for example by collecting only necessary data and minimizing travel requirements
• Offering transportation cost reimbursement for participants attending appointments
• Sending out reminders before each visit - which can be easily automated
• Remote monitoring in order to follow-up rapidly after missed appointments or failed contact
• Paying attention to staff-patient interactions, ensuring to listen to patients and their concerns, treat them with respect, be attentive, and be open to feedback

In general, some degree of attrition is normal and should even be expected. A good starting to point is to assess the sensitivity/robustness of the study design to loss to follow-up by simulating a worst-case scenario.[1] For study designs which are found to be sensitive to the effects of LTFU, considerations should be taken before beginning the study.

For studies wherein higher loss to follow-up is predictable and seems inevitable (for example, longitudinal studies for life-threatening illnesses), another strategy to prevent the loss to follow-up from causing problems with study quality is to over-enroll. This is a potentially costly strategy that should be chosen only when deemed appropriate.

In the case that you do end up with higher than expected loss to follow-up, there are a few methods to account for the effects and any potential bias introduced. We quickly overview two potential strategies below.

How do you mitigate or account for the effects of loss to follow-up?

Researchers are strongly advised to critically assess the effects of loss to follow-up and attrition in trials, and acknowledge any potential influences while analyzing data.

Baer et al. proposed the “LTFU-aware fragility indices” as a way to rigorously assess the outcomes of patients lost to follow up.[3] Intention-to-treat analysis is another potential method wherein participants are counted in the group they were assigned to, despite the missing data.[4] The latter method should be used cautiously as it has the potential to introduce inaccuracies; this further reinforces the idea that loss to follow-up is best mitigated by taking measures to prevent it in the first place.

How many patients can be lost to follow-up without affecting trial results?

There isn't a generalizable rule here, as individual clinical trials can differ substantially in their sensitivity to loss to follow-up, as discussed above. As a rough guideline, losing 5% of patients to follow up is considered unlikely to adversely impact overall trial results. Losses of 20% are typically likely to introduce bias or other issues with the robustness and scientific validity of study results.[5] However, this strongly depends on specifics of the trial, including factors such as sample size and homogeneity within and between study groups. The best way to determine the acceptable loss to follow-up is to conduct a sensitivity analysis.