We aimed to evaluate the outcomes of patients treated for KS in a sub-Saharan African context, where a paucity of information on KS therapy exists. This is partly attributed to the scarcity of resources available for provision of KS treatment [
18]. The study indicated that more patients were accessing treatment for KS than ever before and about one out of two patients had responded objectively to treatment. However, poor outcomes were also common as a result of late presentation and dropping out of care.
The number of patients seeking hospital care for KS at the Dermatology Unit of DNH has increased exponentially within the last few years. Between 2000 and 2009, about 134 patients were documented to have been enrolled in care [
11], but we identified over 250 in the last four years. Elsewhere, the International Epidemiologic Databases to Evaluate AIDS (IeDEA) Consortia in the period 2009–2012 identified 677 cases from Kenya, 172 from Uganda, 57 from Nigeria, 67 from Cameroon and 355 from Malawi [
3]. While East Africa is traditionally known as the hotbed of KS in the world [
19,
20], the rising burden of KS in Guinea is consistent with the prevalent situation in the region where KS is the most commonly diagnosed malignancy associated with HIV. This is especially the case in West and Central Africa, where ART coverage remains low [
6,
18,
21]. In Guinea, an increased ascertainment of KS and referral by physicians may be a contributing factor, together with an increased access to KS treatment, due to the support provided by MSF. Still, most patients presented not only with advanced HIV disease but also with an advanced KS, probably due to poor awareness or reliance on traditional medicine. The distance to travel to the capital city for those in remote areas might not only be a factor for late presentation, but also a contributor to early drop out from care and an obstacle for accessing treatment. This double burden of HIV/KS explains the high rates of attrition from care. Most late presenters could only benefit from palliative care to improve their quality of life. Under such circumstances, the goal of treatment may not be to achieve a high complete response rate but a high objective response rate (ORR = CR + PR), which is a surrogate endpoint for survival [
22]. Thus, attaining an ORR of 52.4% in this cohort may be considered a satisfactory result for the patients, care givers and the programme. However, due to issues related to under- or overestimation of treatment effects, an ORR may not fully capture the net benefits of survival and should thus be interpreted with some caution [
23]. We performed survival analyses based on Kaplan-Meier curves and Cox regression to demonstrate a correlation between treatment response and survival and, we observed that complete and partial responders lived longer than unknown or non-responders. Therefore, ORR could be a reliable marker of survival in advanced KS. This corroborates with findings from an earlier study on metastatic breast cancer [
22]. We believe that defaulters with an unknown response and unknown vital status were probably non-responders and had died. This is because on one hand, they received only a median of two cycles of chemotherapy (treatment response was dose-dependent and vice versa); and on the other hand, the rate of attrition in the HIV cohort in Guinea at the same time was about 25% overall, which is much lower than the rate observed by the KS patients. Conversely, it was possible that some defaulters had responded at least partially to treatment before dropping out from care and thus resulting to an underestimation of the true overall response rate. This uncertainty might also explain the lack of precision (wide 95% CI) around the survival estimates between partial responders and non-responders (Table
3). Our overall findings seem to correlate with results from most other African countries where loss to follow up is very high and generally very poor KS treatment responses, though excellent outcomes have also been described in some African settings [
3,
4,
21,
24]. Therefore, causes of loss to follow up should be investigated and strategies to mitigate them implemented.
The presence of woody oedema was a predictor of a poor treatment response. This form of skin thickening has been linked to long-standing skin oedema, and the resulting reactional inflammatory fibrosis is indicative of advanced disease or stage T1 [
15]. Poor treatment outcomes have been reported in children and adolescents who presented with skin oedema in Malawi [
25]. Clinicians need to identify KS at an early stage by raising their index of suspicion amongst other differential diagnoses in the presence of skin oedema, especially in patients infected with HIV.
Previous exposure to chemotherapy was also predictive of poor treatment response. It is very likely that this group of patients were restarting a treatment regimen that failed them earlier on, given that the country had few options for KS treatment. We recommend such patients to be candidates for second line therapy, which is now available in the form of pegylated liposomal doxorubicin (PLD). Although it is not certain that this will be successful because there have been no significant differences between various chemotherapy regimens in the treatment of severe or progressive Kaposi’s sarcoma in HIV-infected adults [
26]. However, it is reasonable alternative to remove patients from a failed regimen and attempt other options.
Our study had some limitations. The diagnosis of KS was clinical rather than pathologic; while clinical diagnosis is very typical of SSA settings, and probably reasonably accurate in the correct context of HIV with classical skin lesions, it is clearly inaccurate with a recent study finding only 77% positive predictive value for clinically suspected KS compared to final histologic diagnosis in some East African HIV clinics [
19]. The ACTG staging was incomplete and thus we could not account for disease severity in our attrition analyses. Treatment response was not ascertained for patients who dropped out of care, although we assumed with some degree of uncertainty that they were likely to have a progressive disease. This assumption must have led to an underestimation of the overall response rate. Treatment regimens varied between and within patients from one cycle to another, and it was therefore difficult to assess the effect of a single or combination therapy. The observed dose-response effect of chemotherapy appeared plausible but not causal because the temporal sequence of treatment duration and response could not be determined given the retrospective and incomplete nature of our data. A better response could as well have led to longer chemotherapy durations or cycles (and vice versa).