HIV and cancer registry linkage identifies a substantial burden of cancers in persons with HIV in India

Abstract

We utilized computerized record-linkage methods to link HIV and cancer databases with limited unique identifiers in Pune, India, to determine feasibility of linkage and obtain preliminary estimates of cancer risk in persons living with HIV (PLHIV) as compared with the general population.Records of 32,575 PLHIV were linked to 31,754 Pune Cancer Registry records (1996-2008) using a probabilistic-matching algorithm. Cancer risk was estimated by calculating standardized incidence ratios (SIRs) in the early (4-27 months after HIV registration), late (28-60 months), and overall (4-60 months) incidence periods. Cancers diagnosed prior to or within 3 months of HIV registration were considered prevalent.Of 613 linked cancers to PLHIV, 188 were prevalent, 106 early incident, and 319 late incident. Incident cancers comprised 11.5% AIDS-defining cancers (ADCs), including cervical cancer and non-Hodgkin lymphoma (NHL), but not Kaposi sarcoma (KS), and 88.5% non-AIDS-defining cancers (NADCs). Risk for any incident cancer diagnosis in early, late, and combined periods was significantly elevated among PLHIV (SIRs: 5.6 [95% CI 4.6-6.8], 17.7 [95% CI 15.8-19.8], and 11.5 [95% CI 10-12.6], respectively). Cervical cancer risk was elevated in both incidence periods (SIRs: 9.6 [95% CI 4.8-17.2] and 22.6 [95% CI 14.3-33.9], respectively), while NHL risk was elevated only in the late incidence period (SIR: 18.0 [95% CI 9.8-30.20]). Risks for NADCs were dramatically elevated (SIR > 100) for eye-orbit, substantially (SIR > 20) for all-mouth, esophagus, breast, unspecified-leukemia, colon-rectum-anus, and other/unspecified cancers; moderately elevated (SIR > 10) for salivary gland, penis, nasopharynx, and brain-nervous system, and mildly elevated (SIR > 5) for stomach. Risks for 6 NADCs (small intestine, testis, lymphocytic leukemia, prostate, ovary, and melanoma) were not elevated and 5 cancers, including multiple myeloma not seen.Our study demonstrates the feasibility of using probabilistic record-linkage to study cancer/other comorbidities among PLHIV in India and provides preliminary population-based estimates of cancer risks in PLHIV in India. Our results, suggesting a potentially substantial burden and slightly different spectrum of cancers among PLHIV in India, support efforts to conduct multicenter linkage studies to obtain precise estimates and to monitor cancer risk in PLHIV in India.

Figure 1

The flowchart depicts the methodology and outcomes of the computerized HIV database and Cancer Registry Match. Steps for pre-match database review, cleaning, and preparation resulting in match-compatible files are shown. Details of the 4 match passes for the computerized linkage are outlined. It ends with details of the HIV records linked to cancer by time since registration in HIV database including the early incident (EI) and late incident (LI) periods. Cancers diagnosed before or within 3 months of NARI registration were considered prevalent and were excluded from incidence analysis.

Figure 2

The graphical representation of standardized incidence ratios (SIR) for AIDS-defining cancers along with the 95% Poisson confidence intervals. SIRs were calculated by dividing the observed number of cancers linked to persons living with HIV (PLHIV) by the expected number of cancers in the period. SIRs with lower bound of CI > 1 are considered significant. SIRs for cervical cancer were significantly elevated in both early and late periods and only in late period for non-Hodgkin lymphoma (NHL).

Figure 3

The figure includes four panels each showing graphical representation of standardized incidence ratio (SIR) in the early (EI) and late incidence (LI) periods with 95% Poisson confidence intervals for different non-AIDS-defining cancers among persons living with HIV (PLHIV) in Pune. SIRs are plotted on a logarithmic scale. (A) Depicts SIR for cancers that are significantly associated with HIV in both early and late incidence periods and additionally shows a significantly increasing trend from early to late periods. ^∗P (trend) values are: breast (P < 0.001), mouth (P < 0.001), esophagus (P = 0.007), colon (P = 0.002), unspecified cancers (P = 0.001). (B) Depicts SIRs for cancers that are also significantly associated with HIV in both early and late incidence periods; however, they do not show increasing trend from early to late periods. Note: SIR for eye and orbit (shown in table 2) is not shown in this figure. (C) Depicts SIRs in early and late incidence periods for cancers that are significantly raised in late and combined incidence periods and therefore considered to be possibly associated with HIV (data for combined period is shown in Table 2). However, they do not show increasing trend from early to late periods. (D) Depicts the SIRs for cancers which are not significantly elevated in either the early or late periods and are hence considered “not associated” with HIV.