Malaria prevalence and performance of diagnostic tests among hospitalized 1 fever patients in Zanzibar

25 Background: Control efforts in Zanzibar reduced the burden of malaria substantially from 26 2000 to 2015, but re-emergence of falciparum malaria has been observed the last years. This 27 study evaluated the prevalence of malaria and performance of routine diagnostic tests among 28 hospitalized fever patients in a 1.5 years period in 2015 and 2016. 29 Methods: From March 2015 to October 2016, pediatric and adult patients hospitalized with 30 acute undifferentiated fever at Mnazi Mmoja Hospital, Zanzibar were included. The malaria 31 prevalence was assessed by polymerase chain reaction (PCR), rapid diagnostic test (RDT) and 32 routine microscopy. 33 Results: Malaria prevalence was 8% (67/820). All cases identified by PCR had Plasmodium 34 falciparum infection, except for one P. ovale. Compared to PCR, the RDT had sensitivity of 35 64% (36/56), specificity 98% (561/575), positive predictive value (PPV) 72% (36/50) and 36 negative predictive value (NPV) 97% (561/581). Microscopy had a sensitivity of 50% 37 (18/36), specificity 99% (251/254), PPV 86% (18/21) and NPV 93% (251/269) compared to 38 PCR. 39 Conclusions: A high malaria prevalence was identified compared to previous studies in 40 Zanzibar. Microscopy showed higher PPV than RDT in this study. Both RDT and microscopy 41 had low sensitivity compared to PCR. However, low parasitemia identified only by PCR in a 42 semi-immune individual could be coincidental and may not be the cause of the presenting 43 symptoms. To achieve malaria elimination in Zanzibar, PCR-based surveillance is suitable to 44 guide control and elimination efforts.

Commercially available malaria rapid diagnostic tests (RDTs) differ widely in sensitivity and 61 specificity (7), and accurate microscopy depends on high quality technical equipment and 62 experience. PCR has high sensitivity and detect parasitemia lower than 1 p/µL, while the 63 detection limits for microscopy and sensitive RDTs are around 50-200 p/µL and 100 p/µL, 64 respectively (8). 65 The main objective of this study was to evaluate the prevalence of malaria identified by 66 routine tests and PCR, and the performance of RDT and microscopy in assessing febrile 67 patients admitted to Mnazi Mmoja Hospital (MMH), Zanzibar. Neonates under 15 days of age were excluded. Demographic and clinical information was 77 obtained using a standardized case-report form. 78 We obtained blood for on-site RDT and microscopy, and blood in EDTA tubes stored at -79 20ºC and shipped on dry ice to Norway for malaria-PCR to be done later. Malaria microscopy 80 was performed if requested by attending clinician, while PCR and RDT was performed on all 81 patients for the sake of the study.

82
A positive result from one or more of the three tests was used to determine the prevalence of 83 malaria. In case of discrepancy between routine diagnostics and PCR, PCR results were 84 defined as final results. In case of discrepancy and PCR was missing, a positive result from 85 RDT or microscopy was defined as a malaria case. In order to evaluate the performance of 86 routine diagnostic tests compared to a more sensitive and specific reference method, results 87 from microscopy and RDT were compared to PCR. Plasmodium DNA was assessed applying a genus-specific PCR, and quantitative analysis (q-100 PCR) was performed using a customized plasmid (10). For quality assurance, results with   (Table 1).

152
Monthly variations in malaria prevalence are shown in figure 4. An increase of malaria cases 153 was observed at the end and shortly after the rainy season.

155
In this study on patients hospitalized for acute febrile illness, a high malaria prevalence was 156 identified by PCR and routine malaria diagnostic tests compared to previous reports from  There are no previous PCR-based malaria prevalence studies in hospitalized patients in 161 Zanzibar. Previous community-based studies in Zanzibar report a prevalence below 3% up to 162 2015, including PCR-based studies (15)(16)(17). In 2015, a PCR-based study documented a 2 % 163 malaria prevalence in out-patients from rural areas of the two main islands of Zanzibar (5).

164
In line with previous studies (15), malaria prevalence in the present study was higher in 165 children, teenagers and young adults. This may possibly reflect semi-immunity in older 166 individuals exposed to malaria prior to implementation of the comprehensive malaria control 167 program.

168
Malaria was strongly associated with travel to mainland Tanzania within the past six months.

169
This is also shown in recent molecular studies substantiating malaria import from the  The RDT showed a slightly poorer performance compared to PCR than reported previously 187 from Zanzibar (15). The lower PPV of RDTs in the present study (72%) compared to 97% in 188 the study of Shakely et al. (15) 195 However, the high NPV indicates that using RDT or microscopy is relatively safe since less 196 than 4% (RDT) and 7% (microscopy) of malaria cases were missed. Since PCR has very high 197 sensitivity, it is possible that some of the discrepancy between PCR and RDT/microscopy 198 could be due to coincidental non-significant low-level parasitemia in semi-immune 199 individuals suffering from febrile illness of other causes. Indeed, patients positive only by 200 PCR had significantly lower parasitemia than those who also had positive RDT and/or 201 microscopy ( Figure 2).

202
The superior sensitivity of PCR compared to microscopy is well known (8), and may, apart 203 from inherent methodological issues, be due to suboptimal staining of blood slides, 204 malfunctioning microscopes and deficient training of the laboratory technician (21). In the 205 present study, sensitivity of microscopy is still substantially higher than in several other 206 surveys (22). Our findings are in line with a review comparing PCR and microscopy for 207 malaria diagnosis in endemic areas, which found that PCR identified on average twice the 208 number of malaria infections compared to microscopy (23). While PCR is highly sensitive, 209 the level of parasitemia detected by RDT and microscopy corresponds well with clinically 210 relevant malaria (24). Our study could not evaluate whether low-level parasitemia was 211 associated with other causes of fever.

212
With its documented high sensitivity, PCR has an obvious role in malaria surveillance.

213
However, for clinical diagnosis of acute undifferentiated febrile illness, PCR has 214 disadvantages of high cost and technical requirements on laboratories, as well as the potential 215 for detecting non-significant low level malaria parasitemia, and detecting DNA remains of 216 non-viable parasites several weeks after parasite clearance (25). In a study from 2015, PCR 217 was positive in 2% of asymptomatic individuals in Zanzibar (26). PCR's excellent ability to 218 detect low-level parasitemia makes this method particularly suitable for surveillance in 219 settings of near-elimination of malaria, such as in Zanzibar (27).