We analyzed HIV-1 partial pol sequences (HXB2 position 2550–3356), obtained as part of clinical routine care, and sociodemographic data of PLWH who received HIV care at the university hospitals of Bonn, Cologne, Frankfurt, Hamburg, and Hannover and at 2 hospitals in Munich during 2001–2018. Patients could participate in the study if they had recently received their diagnosis of HIV-1 and were ART naive; this conservative approach excluded participants for whom the exact start date of ART or history of prior ART was not accurately documented.

Blood samples were collected before ART initiation. We sequenced partial HIV-1 pol region as previously described (7,9). We set the mixed mutation calling threshold at >10%, consistent with Sanger sequencing sensitivity (10). We identified major DRMs by using the Stanford University Genotypic Resistance Interpretation HVdb version 8.9, (https://hivdb.stanford.edu). We inferred the genetic transmission network as previously described (7,9,11); we inferred putative linkage for genetic distances <1.5% (12) (Appendix).

We performed statistical analyses by using Stata version 14 (StataCorp LP, https://www.stata.com). We applied Fisher exact or χ² test and univariable and multivariable logistic regression models, as appropriate, to determine characteristics that are associated with shared DRM and clustering. A shared DRM was defined as any DRM present in >2 genetically linked persons.

Overall, 1,397 HIV-1 infected participants were included. Most were male (82.9%; 1,158/1,397), originated from Germany (69.6%; 972/1,397), and infected with HIV-1 subtype B (72.8%; 1,017/1,397). The most commonly reported transmission risk was men who have sex with men (MSM) (56.7%; 792/1,397) (Table 1).

We identified an overall prevalence of any DRM at the time of diagnosis, excluding polymorphic mutations, of 17.8% (95% CI 15.7%–19.8%), 248/1,397 participants. The proportion varied significantly (p<0.001) by region, ranging from 9.1% (95% CI 5%–13%; 21/231) in Munich, up to 31.4% (95% CI 24%–38%; 53/169) in Hannover. Resistance mutations associated with nucleoside reverse transcriptase inhibitors (NRTIs) (172/1,397; 12.3%) were most frequent, followed by nonnucleoside reverse transcriptase inhibitors (NNRTIs) (124/1,397; 8.9%). The most common single mutations related to NNRTIs were K103N (31/124; 25.0%), and G190A (8/124; 6.5%). Of the NRTI resistance mutations, M41L (25/172; 14.5%), and T215S (18/172; 10.5%) were most frequently observed (Table 2).

Figure 1

Figure 1. Transmission network analysis by sex and location (A) and by characteristic (B) for 1,397 patients with HIV, Germany, 2001–2018. Endemic, recent immigration from a country with HIV prevalence >1%; HTS, heterosexual...

Transmission network analyses revealed that 20.7% (289/1,397) of participants had a putative linkage forming 102 transmission clusters. The largest cluster included 12 participants, mostly MSM from Bonn, Cologne, Munich, and Frankfurt (Figure 1, panels A, B). Participants <25 years and 25–45 years of age were significantly more likely to cluster compared with participants >45 years (<25 years, adjusted OR [aOR] 4.38, 95% CI 2.55–7.52, p<0.001; 25–45 years, aOR 1.91, 95% CI 1.36–2.678; p<0.001). Participants infected with HIV-1 subtype B were more likely to cluster than those with non-B subtype (aOR 4.05, 95% CI 2.37–6.90; p<0.001). Geospatial distribution differed; participants from Bonn were linked significantly more often than those from Cologne (aOR 1.63; 95% CI 1.06–2.49; p = 0.025), even though the cities are geographically close (Appendix Table).

Figure 2

Figure 2. Presence of drug resistance mutations by location (A) and by risk factor (B) for 1,397 patients with HIV, Germany, 2001–2018. DRM, drug resistance mutation; HTS, heterosexual; MSM, men who have sex...

The prevalence of transmitted DRM was comparable in clustering (47/289, 16.3%) and nonclustering (201/1,108; 18.1%) participants (p = 0.46) (Appendix Table). Of the 47 sequences harboring DRM, 19 (40.4%) were preferentially shared by participants living predominantly in Cologne (14/19, 73.7%) and Bonn (3/19, 15.8%) (Figure 2, panel A) and by participants reporting MSM as main risk factor (15/19; 78.9%) (Figure 2, panel B). Younger age (<25 years) was associated with a higher proportion of shared DRM (3/11; 3.5%) compared with older age (24–45 years, 13/856 [1.5%]; >45 years, 2/430 [0.5%]) (Table 1).

The most frequently observed putatively shared DRM was K103N, detected in 9/19 (47.4%) participants forming 4 distinct clusters, predominantly originating from Cologne (7/9, 77.8%). The second most common shared DRM was D67N, found in 6/19 (31.6%) participants from Cologne and Bonn.

The increasing prevalence of DRMs in PLWH has become a serious matter of concern for clinicians and public health entities (4). In our study, we observed a 17.8% prevalence of DRMs, higher than in previous studies (6,7). The proportion of NNRTI resistance mutations was 8.9%, which is potentially associated with the common use of NNRTI in first-line ART regimens. K103N represented one quarter of NNRTI resistance mutations, reducing susceptibility to the first-generation drugs nevirapine and efavirenz (13). Transmission network analyses revealed that K103N was the most frequently shared DRM.

K65R, K70RT, and M184IV were the most common of the NRTI resistance mutations we observed, particularly among the risk group of MSM living in Cologne and Hannover, indicating potential resistance to preexposure prophylaxis (PrEP) with tenofovir/emtricitabine. Such resistance might be an upcoming challenge as PrEP use increases. Monitoring for HIV infections with these mutations is of utmost importance for preventing an epidemic among high-risk PrEP users; one mitigation is to consider alternative PrEP regimens in regions with high resistance.

Our study had several limitations. First, our sample population could have been biased because participants were not randomly selected; our dataset was limited to ART-naive patients who received an HIV diagnosis at 7 university hospitals during 2001–2018. Although we know no reason why a university hospital setting would not be representative of the region, it is possible that populations treated outside these centers may have different transmission networks and risks; results are not generalizable to the entire regions or nationwide. Second, mixing of heterosexual patients and MSM in clusters may be due to missing single or multiple risk factors. Thus, their role could not be represented in the transmission networks. Third, we have not tested clinical correlates and drug resistance; the clinical relevance was inferred from the Stanford database.

In summary, we found that the overall rate of DRM was high in Germany. Network analysis elucidated cases of shared DRMs among genetically linked persons, mainly in MSM-dominated clusters. Our findings highlight regional differences and illustrate the need to test MSM, especially younger men, for HIV regularly and to evaluate local HIV programs and adapt screening and treatment strategies to local epidemics.