This systematic review and meta-analysis of studies that traced patients who were LTFU in ART programmes in resource-limited settings showed that the outcome of over a third of patients remained unknown. All studies except one were conducted in sub-Saharan Africa and no study was done in children. Among African adults who were LTFU after starting ART and successfully traced, the combined mortality was 46%. Mortality ranged from 12% to 87% across studies, and was inversely associated with the rate of LTFU in the programmes. Incorrect or missing telephone numbers and addresses were often the reason why patients could not be located. Transfer to another programme, financial constraints and improving or deteriorating health were common reasons for not returning to the clinic.
We performed a comprehensive search of the literature, including of abstracts presented at three major HIV/AIDS conferences, thus minimizing possible publication bias. We identified studies of over 6,000 patients who were LTFU in ART programmes in 10 low- or middle-income countries. Sites were heterogeneous and included both rural and urban locations. The approach used to trace patients varied and included telephone calls, home visits and social networks. Our findings should therefore be applicable to other ART programmes, particularly in sub-Saharan Africa.
Definitions of LTFU and the assessment of reasons for not returning to the clinic were not standardized across studies, which precluded formal meta-analysis of these data, and information on causes of death was limited. Other limitations include the lack of information, in most studies, on the time of death. The limited information that is available from some studies 
indicates that patients were lost in the first few months of ART, and died soon thereafter. Data from the ART in Lower Income Countries (ART-LINC) collaboration
and other treatment programmes, for example the Médecins Sans Frontières (MSF) programmes in Malawi 
, and South Africa 
, showed that loss to follow-up and death mostly occur in the first six months after ART initiation.
A high risk of death in the first few months after starting ART is characteristic of resource-limited settings where most patients start therapy late with advanced disease 
. However, mortality in patients LTFU is substantially higher than the mortality commonly reported in the first year of ART based on routinely recorded deaths and, censoring of follow-up in patients LTFU 
. ART programmes with high rates of LTFU and poor ascertainment of deaths may therefore seriously underestimate mortality. Furthermore, mortality among patients LTFU differs depending on the rate of LTFU of the treatment programme: mortality declined with increasing rates of LTFU. In programmes with high rates of LTFU those LTFU might thus include a sizeable group of low-risk patients who self-transferred to another programme, for example because of a more convenient location of the new clinic, to avoid stigma or due to work-related reasons. The results of dedicated studies tracing patients LTFU can be used to correct naïve estimates of mortality in a given programme 
. In the absence of such studies, the data from this systematic review provide a sensible range of estimates of mortality in patients LTFU which can be used in sensitivity analyses to adjust overall mortality.
In most studies an important proportion of patients could not be located, and mortality of those whose vital status could be ascertained may not be representative of all patients LTFU. Contact information that is absent, incorrect or out-of-date could be related to the risk of death. For example, healthier individuals may be more mobile than sicker patients, and more likely to leave the catchment area of the clinic in search of work. Conversely, patients providing incorrect details may be part of a vulnerable group, with little social support and low adherence to ART. If results of tracing studies are likely to be affected by selection bias, correction of mortality is again best done in sensitivity analyses, using a range of plausible values. Clearly, the quality and completeness of patient's contact details should be improved and regularly updated during follow-up. Of note, a recent survey 
of electronic medical record systems used in ART programmes in lower-income countries found that well managed databases might contribute to retaining patients in programmes.
An understanding of the reasons for not returning to care is important to the design of effective and cost-effective ART programmes. Outreach teams that routinely trace patients, combined with other measures, can substantially reduce LTFU 
, but such teams are costly, and the emphasis should be on the prevention of LTFU. Transfer to another programme was common among patients found to be alive. Strengthening of referral systems and regular exchange of information between clinics, together with patient education could increase the recording of transfers and ensure continuity of care. Unsurprisingly, financial constraints were another common reason for not returning to the clinic. Direct and indirect costs related to the provision of care have been identified as major obstacles to access to ART, acceptance of ART 
and adherence to treatment 
. Mortality in programmes that charge user fees has been shown to be higher than in those offering free treatment 
. Decentralisation of services, task shifting to lay care providers, longer drug refill periods for stable patients, as well as provision of transport vouchers for those in need are some of the strategies that could address this issue.
Other important reasons for LTFU were improvements in health, adverse effects and feeling too sick to come to the clinic or being hospitalised. Reports of stopping care as a result of perceived improved health reflect a poor understanding of the chronic nature of the disease and the need for continued, life-long ART. The experience or fear of toxicities has been found to be associated with poor adherence in previous studies 
. These issues need to be addressed through training of care givers and preparing patients for ART. Interventions that are aimed at the individual (rather than groups) and provided over longer time periods (>12 weeks) have been shown to be effective in improving adherence to ART 
Stigma and social problems were also repeatedly mentioned. Fear of disclosure, social isolation or the exposure to a discouraging social network have being identified as barriers to treatment adherence in studies conducted in high and low-income settings 
. In a study conducted in Botswana, Tanzania and Uganda, patients reported difficulties in taking their drugs when they were among employers, co-workers or friends to whom they had not disclosed their HIV status 
. The development of practical medication management skills in open discussions with patients could be beneficial in this context 
In conclusion, a substantial minority of patients LTFU cannot be traced and among those traced on average 46% of patients have died. Transfer to another programme, financial constraints and improving or deteriorating health were common reasons for not returning to the clinic. These findings have important implications both for patient care and the monitoring and evaluation of ART programmes in resource-limited settings.