The onset of the coronavirus disease 2019 (COVID-19) pandemic led to severe restrictions on public mobility and interaction, national or state-wide lockdowns, school and workplace closures, and curfews. The result was widespread economic hardship, in addition to the toll on health.
A December 2020 preprint research paper published on the medRxiv* server describes such lockdowns as possibly ineffective, based on data from 11 Danish municipalities that were or were not locked down during the pandemic.
As the spring wave set in within many countries of the Northern hemisphere, many researchers have found themselves perplexed about the real effectiveness of many non-pharmaceutical interventions (NPIs), as they are called. The difficulty is compounded by the absence of real controls, making most deductions the result of indirect modeling studies or empirical evaluations.
Understandably, many such models concluded that NPIs reduce the rates of new infections and, thus of deaths. The problem is that counterfactual scenarios are purely hypothetical.
On the other hand, observational studies are empirical or based on actual data but are subject to many confounding factors which could cause many of the conclusions to be biased. Again, many countries need to be included, rather than the many comparison countries, including small groups of countries. The most significant chance of errors occurs when the curve is influenced by factors relating to natural differences in individual susceptibility and transmission. Especially, the age of an individual and the presence of other comorbidities may be more important than NPIs in determining the risk and outcome of infection.
The current study is based on data from Northern Jutland, in Denmark, which is a major mink farming region. Mink is a significant export of Denmark, and mink farms here were found to harbor several mutations of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that arose de novo in the infected animals.
The finding that such mutations were present led to the imposition of stringent lockdowns in the early part of November 2020 in seven of eleven municipalities in the region. In the other four, the moderate lockdown already underway was continued.
Data from both groups, comprising seven and four municipalities, respectively, showed similar numbers of infections. The large number of infections ensures the study has adequate power to detect a difference in outcome due to NPIs, if present while preventing confounding factors. The researchers comment, “This situation provides us with a unique data set of an intra-country (homogenous population) comparison with a direct case-control, highly focused in both time and space.”
In the locked-down municipalities, the number of positives was 0.15 per 1,000 per day, relative to 0.14 per 1,000 in the other group, prior to the lockdown. The next seven days are unlikely to show any effect due to the lockdown since the median incubation period of the virus is 5-6 days. During this period, the infection rates were 0.11 per 1,000 in both groups.
The spring wave led to a peak of 0.69 positive tests per 1,000 residents in the first group of seven, compared to 0.82 per 1,000 in the group of four municipalities, which lacks statistical significance. The high comparability of these rates at different time points prior to the lockdown seems to indicate that this provided a true and almost natural, experiment on lockdown efficacy.
The NPIs lasted for about one or two weeks and thus offered a setting in which the short-term effects of the lockdown could be evaluated without other confounders. The lockdown is estimated to have affected 126,000 employees and 13,600 students.
The researchers found no significant difference in the number of infections before and after the lockdown was intensified, compared to the other group of four. Secondly, the number of infections began to go down a week or more before the lockdown could have been expected to have any effect. Thirdly, the number of infections also began to go down in the four municipalities where no change was enforced.
Mass testing was carried out for several weeks in the seven locked-down municipalities but did not appear to change the infection dynamics. On the other hand, the intensive testing ensured that the vast majority of cases were detected. In fact, two of every three people in these seven municipalities were estimated to have been tested.
The reasons for the decline in infections may include efficient surveillance measures already put in place, along with a high degree of compliance on the part of the citizens. The combination of these factors appears to be at least as effective as a complete lockdown.
Lockdown was ineffective
After the lockdown was enforced, there was no significant change in infection rates, suggesting its ineffectiveness. The effect of the lockdown was not found to change within the next one week, which was taken to be the incubation period.
They also found that the lockdown had no significant effect at any level of infection in terms of a difference in infection rates between those municipalities that were or were not locked down. On the other hand, the number of tests went up steeply after the lockdown began. However, this did not result in any significant change in the number of infections relative to other municipalities.
This study thus demonstrates that a full lockdown appears to have no significant gains on reducing the number of infections, compared to voluntary compliance with behavioral changes and other NPIs of a less stringent nature. “In our own analysis, we did not find evidence of any effects of the lockdown on the development of infection rates across Northern Jutland, relative to the control.” The incidence of infection did fall, but the starting point fell before the lockdown was announced, and to a comparable extent relative to the control municipalities. In December 2020, both groups of municipalities showed a fivefold rise in test positives in a short period of time, showing that the overall pattern of infection were not significantly changed over time.
What did cause the reduction in cases, if not the lockdown? The researcher suggests that testing saturation occurred so that the number of cases identified plateaued much before mass testing occurred. Thus, cases were being efficiently identified and isolated even before the lockdown, and this scenario was an important factor in containing the pandemic.
Secondly, voluntary compliance with recommended behavior changes, as well as trust in the government and public health interventions, coupled with the availability of high-quality information about the pandemic, indicates that this acted along with the efficient testing and contact tracing program to produce a markedly effective containment of the virus.
Finally, lockdown may not be as effective as thought due to heterogeneous susceptibility, disease activity, individual infectivity and compliance with public health interventions. “Complete lockdown is not necessarily the most effective measure in the context of other, less costly NPIs, in particular, if considering the effects of infection tracing and voluntary compliance,” said the researchers. Instead, limiting gatherings to less than a hundred people and closing academic institutions may be adequate to contain the viral spread, as many earlier studies have suggested. Further analysis of this dataset may confirm this conclusion.