How does loss to follow-up influence cohort findings on HIV infection? A joint analysis of the French hospital database on HIV, MortalitE 2000 survey and death certificates

We aimed to retrieve the vital status of patients lost to follow-up (LFU), with no further visits for at least 12 months, for the 34 835 patients in the Agence Nationale de Recherche sur le SIDA CO4 French Hospital Database on HIV (ANRS CO4 FHDH) seen in 1999 and to examine how loss to follow-up might influence estimates of survival and the impact of delayed access to care (DAC) on survival. The status of LFU patients was established by using the mid-2006 update of the FHDH in which their status 12 months after loss to follow-up was added when available and by matching with the MortalitE 2000-Epidemiological Centre for Medical Causes of Death (CEpiDc) database, which included HIV-infected patients dying in 2000. We compared Kaplan-Meier and hazard ratio (HR) estimates before and after correction for the status of LFU patients. In the mid-2006 updated FHDH, of the patients seen in 1999, 7.5% were LFU: of these, 2.1% later returned for follow-up, with a median time without follow-up in an FHDH centre of 3.5 years, and 5.4% had no further FHDH visits whatsoever, of whom 29.8% died according to MortalitE 2000-CEpiDc. After correction, the estimated 1-year survival rates following enrolment in 1999 differed between the original and updated analyses (97.1 vs. 95.9%, respectively; P=0.017); the estimates of mortality HRs associated with DAC did not differ during the first 6 months, but did differ for the 6-18-month period. Among LFU patients, 28.1% returned to follow-up after several years and at least 21.4% died, which led to a slight overestimation of both survival and the impact of DAC on survival.