Fielder_02.qxp 16/7/08 10:38 am Page 55
Statistical Analysis of HIV-patient Data from Central Europe – Lessons from a Three-centre Study
Table 3: Descriptive Statistics of CD4
+
Counts
CD4
+
count (cells ml
-1
)
Number Total number Mean Number Inter-quartile Standard
of Patients of CD4
+
readings of Readings Mode Median Range Mean Deviation Min Max*
Gender Male 9,072 164,857 18.17 290 360 335 402.21 274.32 1 5,000
Female 2,944 51,813 17.6 360 380 323 424.99 278.07 1 4,876
NK** 399 3,733 9.36 320 360 326 403.26 269.30 2 1,820
Risk Group IVDA 2,417 46,124 19.08 240 350 313 392.37 261.97 1 3,258
Homo/Bi 3,839 78,090 20.34 360 375 344 411.46 274.35 1 4,567
Het 3,049 52,841 17.33 290 373 320 415.19 264.95 1 3,676
Bld 322 7,117 22.10 380 339 294 373.72 256.74 1 2,543
M/C 63 1,297 20.59 598 590 509 746.72 619.20 2 4,876
NK 2,725 34,934 12.82 170 352 345 401.90 283.66 1 5,000
Database ARCA 5,204 94,402 18.14 420 382 343 427.38 287.53 1 4,876
Arevir 3,448 43,921 12.74 290 316 324 359.30 262.95 1 5,000
Karolinska 3,763 81,080 21.55 290 370 320 410.12 263.62 1 3,000
All 12,415 220,403 17.75 290 365 335 407.6 275.29 1 5,000
* A CD4
+
count upper threshold of 5,000 cells ml
-1
was used to remove spurious outliers.
IVDA = intravenous drug abuse; Homo/Bi = homosexual/bisexual; Het = heterosexual; Bld = blood products; M/C = mother-to-child; NK = not known.
This contrasts with the alternative definition, which has an additional Kruskal-Wallis tests to contrast the log
10
(VL) values for different risk
category of ‘success and then failure’ that gives treatment success/ groups, genders and databases showed highly statistically significant
failure/success-then-failure proportions as 16.2, 81.2 and 2.6% for differences in all three factors: risk group (H=304.2, n=33,421; p<0.0001),
ARCA, 10.2, 82.3 and 7.4% for Arevir and 24.7, 72.2 and 3.1% for gender (H=56.9, n=38,355; p<0.0001) and originating database
Karolinska. The alternative definition classifies more therapies, but is (H=2,861.7, n=44,457; p<0.0001). Furthermore, it can be concluded from
biased towards failures since, in the main, every successful HIV therapy the results of a Friedman’s test, given in Table 4, that at a 5% significance
will eventually fail as resistance to a particular drug combination level there are statistically significant differences in log
10
(VL) for each of the
develops or as a consequence of the effects of long-term drug gender, risk group and database factors, either singly or as interactions. For
administration; also, in some cases, debilitating side effects have an all interactions and main effects of gender and database, the results are
influence on patients. highly statistically significant.
Descriptive Statistics and Analyses Relating to The post hoc analysis in Table 5 (with α=0.01) clearly shows differences in
Patient Prognostic Indicators log
10
(VL) by risk and database. By database, there is no difference between
Tables 2 and 3 show the descriptive statistics for the log
10
(VL) ARCA and Arevir, whereas Karolinska differs significantly from both.
and CD4
+
counts, respectively, by risk group, gender and database. Furthermore, the IVDA group differs from the others, but there are no
The modal value for log
10
(VL) is 2.48 – up to 2dp, equivalent significant differences between the Homo/Bi, Bld or Het groups.
to a VL of 300 copies ml
-1
– in all cases; this is due to the
predominance of values at or below the lower threshold value. Discussion
Average values of VL or CD4
+
values showed little difference between
males and females. Implications for Integrated Database Development
The EuResist project aims to produce a viable prognostic system for the
The data suggest that there are differences in the way in which the successful treatment of patients with HIV in Europe. Consequently, the data
various groups are monitored and in the practice methods of the various repository has to be large and representative of the European HIV-positive
groups of clinics – the database factor represents an aggregation of population. The contributing databases located in northern, central and
clinics in ARCA and Arevir. The mean number of VL and CD4
+
southern Europe do not have the same patient characteristics, as
measurements is highest for the Homo/Bi and M/C groups, as well as demonstrated in this article. However, collecting data from across Europe
being generally higher in the Karolinska data than Arevir or ARCA. It is from regions with different immigration flows and different HIV epidemic
lowest in all of the NK groups, which may be indicative of the fact that features provides a unique opportunity to analyse patient characteristics
these groups contain partial patient records and records from patients within these population groups.
who change clinics. For these reasons, the NK groups were excluded
from statistical tests (see below). All three of the databases were originally designed for clinical purposes, with
patient data generally being entered into the database with limited
There is a clear difference in average values between those patients screening. While the clinical origin of the data has resulted in a large
infected by vertical transmission and patients in other risk groups. The database, with data spanning 20 years, patient data have not been recorded
vertically infected patients have the highest mean and median log
10
(VL) in a consistent manner. This produced a number of analytical problems –
values and CD4
+
counts, as well as the largest standard deviation of any such as inconsistent VL and CD4 count frequencies – but it also resulted in
of the groups. This may be influenced by the small size of this group, but a database with a natural variability that is representative of that
it is more likely to reflect the immaturity of the immune system within encountered by clinicians. These fluctuations should allow EuResist’s
these patients,
25,26
many of whom would have been infected before or prediction engines to be trained on data that closely replicate the clinical
during birth.
27
environment, leading to more appropriate models.
EUROPEAN INFECTIOUS DISEASE 55
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132