To count lost to-follow-up (LTFU) as part of adherence denominator?
Started by Ziad El-Khatib on 21 Aug 2008
Hi!
We know already that LTFU is a problem, and adherence level has been reported to be high.
My question should we count LTFU as part of "patients failing treatment" under the umbrella of 'intent to treat' or..?
thanks
ziad
Keywords: Adherence, LTFU, Monitoring & Measurement, intent to treat, intent to treat analysis, lost-to-follow-up, treatment failure
David Bangsberg, MD, MPH
This is an important question which is often confused in the literature. I think treatment access, treatment adherence, treatment failure, and LTFU should be treated differently because they have different root causes and require different interventions. The definitions below may help distinguish these concepts, and hopefully guide strategies to improve each component of successful treatment. These definitions are loosely based on work by Giordano[1] who has conceptualized adherence into a range of behaviors from HIV testing, entry into care, visit adherence, and medication adherence.
1. Treatment access. Separating treatment access from treatment adherence is important. Treatment adherence presumes someone has medications to take. There are several studies that find low rates of treatment adherence because of patients had an interruption in supply of medications, sometimes pharmacy stock-outs. We have argued that it doesn't make sense to combine incomplete access with incomplete adherence[2]. How can you miss medications that do exist? Incomplete access is often related to logistical, economic or structural barriers, rather than behavioral or social barriers, and, therefore, require logistical, economic, or structural interventions.
2. Treatment adherence. There are many definitions and nuances of treatment adherence (see [3]), but the most common definition is the number of doses taken/number of doses prescribed over a defined interval of time. This is often expressed as a percent. As above, this presumes access to medication. Incomplete adherence in the setting of a ready supply of medications is often related to memory, side effects, depression, needing to conceal medications due to HIV non-disclosure, or other social-behavioral factors.
3. Loss-to-follow up or visit adherence. Visit adherence is different from, though related to treatment adherence. While there are many nuances to this definition as well (does a rescheduled visit count as a missed visit?), it is often defined as the percent of scheduled visits that are attended. It is related to treatment access and treatment adherence because a treatment visit is required to pick up medication refills in many settings. Visit adherence or LFTU has emerged as an important problem in resource-limited settings. Literature to date suggests that LTFU in RLS may be due to economic/structural barriers of transportation, transfer of care to a different clinic, mortality between visits, and less commonly behavioral-social factors. Please see Rosen, Brinkoff, and Geng for references below for a more detailed discussion[4-6].
4. Treatment failure. This is often defined as virologic failure and is related to either incomplete adherence, suboptimal pharmacokinetics related to medication absorption and metabolism, and possibly overwhelming disease.
Please feel free to email me at if you have any trouble finding these references. I would be pleased to send you a pdf.
David Bangsberg
1. Giordano, T.P., M.E. Suarez-Almazor, and R.M. Grimes, The population effectiveness of highly active antiretroviral therapy: are good drugs good enough? Curr HIV/AIDS Rep, 2005. 2(4): p. 177-83.
2. Bangsberg, D.R., N. Ware, and J.M. Simoni, Adherence without access to antiretroviral therapy in sub-Saharan Africa? Aids, 2006. 20(1): p. 140-1; author reply 141-2.
3. Kerr, T., et al., Measuring adherence to highly active antiretroviral therapy: implications for research and practice. Curr HIV/AIDS Rep, 2005. 2(4): p. 200-5.
4. Rosen, S., M.P. Fox, and C.J. Gill, Patient retention in antiretroviral therapy programs in sub-Saharan Africa: a systematic review. PLoS Med, 2007. 4(10): p. e298.
5. Geng, E.H., et al., Sampling-based approach to determining outcomes of patients lost to follow-up in antiretroviral therapy scale-up programs in Africa. Jama, 2008. 300(5): p. 506-7.
6. Brinkhof, M.W., et al., Early loss of HIV-infected patients on potent antiretroviral therapy programmes in lower-income countries. Bull World Health Organ, 2008. 86(7): p. 559-67.
8:02 AM, 23 Aug 2008 | Permalink
Mona Haidar, MD
Thank you a lot for your reply!
I wonder how well visit adherence correlates with treatment adherence in resources-poor settings.
I spent the last year working with PIH in Lesotho/HIV rural initiative.Outreach activities/home visits and community health workers are essential components of the care delivery.
If a patient on ART misses his appointment , this is usually followed by a home visit by the clinic staff to assess the situation, address the barriers and refill his ART.Also, sometimes the community health workers present to the clinic on behalf of the patient she/he usually accompanies and discuss the patient's situation with the clinic, the reason he had to miss his appointment and refill her/his ART (another visit is usually scheduled).These scenarios are exceptions, but they show possible dicrepancy between visit adherence and treatment adherence.
11:50 PM, 26 Aug 2008 | Permalink
John Kraemer
Ziad, two additional wrinkles to add to David and Mona's excellent responses.
First, in some instances you may want to distinguish between visit adherence (or non-adherence) and true loss to follow-up. Visit adherence being the percentage of visits attended, and loss to follow-up being a sustained non-attendance (essentially dropping entirely out of care). There isn't much data published on this, but we suspect that poor visit adherence is a predictor of loss to follow-up, and we're hoping to do a basic retrospective chart review case-control study to see if this is born out.
The other thing is that, depending on the purposes for which you're reporting data, you may want to think about how to treat patients who transferred to other clinics. We've generally considered transfers to be censored data when doing survival analysis, since, in our experience, they tend to represent patients successfully under care for the length of time that we were treating them. However, if patients are being transferred to clinics that can provide more specialized care because they are not faring well, they might be better classified as treatment failures for program evaluation purposes.
Feel free to email me if I can provide more detailed help: . We've grappled with this denominator question quite a lot as we've been trying to figure out how best to analyze and evaluate our treatment programs.
John Kraemer
Tiyatien Health
8:03 PM, 9 Sep 2008 | Permalink
Ziad El-Khatib
Thank you all!
Question to the moderators, can the person refer to these discussions when writing a report/article?
best regards
ziad
5:35 AM, 11 Sep 2008 | Permalink
Jerry Okal
I have been following your discussion about adherence and LTFU for some time now. Really its not easy drawing the lines but I thought clear definations are needed for what constitutes a LTFU and visit adherence. We conducted an adherence study [from 2003] in Mombasa,Kenya comparing patients on standard care [once monthly visit to facility] and those on DAART arm [twice weekly visit to facility] for which we had to contend with non visit and LTFU. I thought it would be usefull if you link up with the study PI who would be happy to provide you with more information on this. Kindly let me know if you would like to contact her.
9:50 AM, 11 Sep 2008 | Permalink
Ziad El-Khatib
Dear Jerry, thank you.
If there is any document(s) to share it will be also good? ()
best regards
ziad
10:58 AM, 11 Sep 2008 | Permalink
Chorongo Salee
My experience in the adherence study in mombasa has it that LTFU has no clear definition but my opinion is that it should be classified into two division whenever you refer to it and this are
1) Service provider driven ie logistics in availing medication
2) Patient driven ie personal reasons for not adherering to medication
incase there are other arms you are using to evaluate your treatment programme i will be happy to here about them meanwhile i feel those two are the main ones
chorongo salee
mombasa kenya
3:50 AM, 12 Sep 2008 | Permalink
Jerry Okal
Hi Ziad,
Let me find out if there is already an article on the adherence study then i will let you know.
Thanks
Jerry
10:28 AM, 12 Sep 2008 | Permalink
Mona Haidar, MD
Dear Ziad, All,
It is a great idea!
Please feel free to refer to this discussion in reports and articles. You can use the direct link (http://www.ghdonline.org/adherence/discussion/to-count-lost-to-follow-up-ltfu-as-part-of-adheren/) and mention the mission of GHDonline if you wish:
GHDonline is where health care implementers share proven practices, connect with colleagues, and find resources they need to improve health outcomes in resource-limited settings.
We are working hard to develop the Global Health Delivery communities of practice so spreading the word and making sure colleagues and peers know about it would be a tremendous help. If it is published, we would welcome receiving the link to the publication, or even better, you could add it as a new resource in the community.
Many thanks in advance.
5:14 PM, 14 Sep 2008 | Permalink
Joia Mukherjee, MD, MPH
While i agree with David that LTFU is different than access or adherence, I believe that in the field of HIV, we should develop a standard way to analyze a the cohort, LTFU should be an integral part of the analysis.
from the TB world, the standard cohort definition are cured, failed, died and defaulted; in this way, defaulted takes away from the % cure, as it should. believe we should count lost to follow up as "failing" in an intention to treat analysis.
i think it is critical to make retention, not only adherence a critical part of our evaluation of programs
Joia Mukherjee
3:03 PM, 26 Sep 2008 | Permalink
Ziad El-Khatib
Hi Joia, thank you...
With my naive way of understanding TB treatment/adherence world, does the period of time of treatment (6-9 months) makes it more convenient to include LTFU as part of adherence, in compare to a HIV that a life time treatment program?
Best regards
ziad
6:03 PM, 26 Sep 2008 | Permalink
David Bangsberg, MD, MPH
Hi Joia,
Thanks for bringing up the TB concept of defaulting. I am not sure that all "defaulting" is a bad outcome in rural HIV treatment programs. Elvin Geng at UCSF (CROI 2008; JAMA 2008) has set up a outreach program in Mbarara where a motorcycle tracker find as many people as possible from a formally generated random list of all people lost for >6 months. Our tracker is able to determine vital status on 87%. 62% are alive and 83% of those alive transferred care to another clinic, often because a clinic closer to where they live has started providing HIV antiretroviral therapy as part of successful scale-up. These "defaulters" are better described as "transfers" and I interpret them as "positive" outcomes. This makes LTFU as seen from the clinic perspective such a murky issue for program evaluation in that it represents both system failures and system successes. It is impossible to know how much is "success" and how much is "failure" without an active outreach program.
Best regards,
David
9:02 AM, 27 Sep 2008 | Permalink
Ziad El-Khatib
Comment inline with David last reply. A defaulter survey in Johannesburg found part of them went to clinics in their areas without informing their main HIV clinic.
This makes the definition of LTFU as patients confirmed to be alive and not exposed to ARVs somewhere else.
Best regards
ziad
9:51 AM, 27 Sep 2008 | Permalink
Greg Bisson
In a sample of 410 patients consecutively registering for treatment in a public clinic in Gaborone, Botswana, we found that ~60% of patients who were categorized as lost to follow-up before tracing were dead. In addition, lower baseline CD4 count was associated with remaining lost to follow-up after tracing, suggesting that some of the patients whose status could not be ascertained also had died. In this study, we excluded transfers out to other clinics, and were careful not to count this usually good outcome as a bad event.
This is the link to the paper presenting the data above:
http://www.plosone.org/article/info:doi%2F10.1371%2Fjournal.pone.0001725
My opinion is that LTFU should be handled by performing a sensitivity analysis, wherein the proportion of LTFUs that are estimated to have died is varied around plausible (and previously reported by studies where patients have been traced) estimates of deaths in this group. Going further, you could vary the deaths among those lost to follow-up as being anywhere from 0 to 100%, and see if this affected your primary results. Similarly, one could vary likely adherence among those lost to follow-up prior to having adherence measured in the same manner. These types of sensitivity analyses give you an idea of how robust your conclusions are.
This is likely not only important for determining overall adherence rates, but also for determining unbiased estimates for risk factors for poor adherence, as we showed in the PLoS ONE paper linked above.
Greg
10:32 AM, 28 Sep 2008 | Permalink